gnm/0000755000176000001440000000000012515671462011074 5ustar ripleyusersgnm/inst/0000755000176000001440000000000012515664670012054 5ustar ripleyusersgnm/inst/CITATION0000644000176000001440000000140012515147523013175 0ustar ripleyusers## R >= 2.8.0 passes package metadata to citation(). if(!exists("meta") || is.null(meta)) meta <- packageDescription("gnm") year <- sub("-.*", "", meta$Date) note <- sprintf("R package version %s", meta$Version) url <- sprintf("http://CRAN.R-project.org/package=gnm", meta$Version) citEntry(entry="Manual", title = {"Generalized nonlinear models in R: An overview of the gnm package"}, author = personList(as.person("Heather Turner"), as.person("David Firth")), year = year, note = note, url = url, textVersion = paste( "Heather Turner and David Firth", sprintf("(%s).", year), "Generalized nonlinear models in R: An overview of the gnm package.", sprintf("(%s).", note), sprintf("(%s).", url)) ) gnm/inst/NEWS0000644000176000001440000006431312515534674012562 0ustar ripleyusersChanges in gnm 1.0-8 ==================== Improvements ------------- o now use lazy data loading. o improvements to vignette (with thanks to Michael Friendly). o copyright notices added to source files to clarify authorship, with appropriate credit given to contributors in Rd and DESCRIPTION files. Changes in behaviour -------------------- o predict.gnm now includes eliminate term in predictions on new data. Bug fixes --------- o expandCategorical now works when there are no covariates in the data. o better handling of single-column model matrices. o predict.gnm now works with se.fit = TRUE for models with eliminated terms; correctly handles new data without all levels of homogeneous factors present, and respects contrasts settings. o environment of formula preserved when using instances. Changes in gnm 1.0-7 ==================== Bug fixes --------- o corrected use of anova.glmlist in anova.gnm. Changes in gnm 1.0-6 ==================== Bug fixes --------- o added catch for when deviance becomes NaN. Changes in gnm 1.0-5 ==================== Improvements ------------ o eliminated coefficients now returned as named vector. o step-quartering introduced to start-up iterations to avoid increasing deviance. Changes in behaviour -------------------- o gnm() no longer restarts if algorithm fails - better to provide improved starting values in this case. Bug fixes --------- o fixed bug in the way 'etastart' is used to initialise the linear parameters when 'eliminate' is used as well. Changes in gnm 1.0-4 ==================== Bug fixes --------- o restarting mechanism now reinitialises correctly. o removed call to external C function that is no longer available. o gnm() now works with 'eliminate' argument when remaining linear part of predictor only involves one parameter. Changes in gnm 1.0-3 ==================== Improvements ------------ o pickCoef() extended to allow fixed pattern matching and to optionally return actual coefficients rather than their indices. o gnm() now looks for exact match in coefficient names when a single character string is passed to the 'constrain' argument before treating as regular expression. o hatvalues.gnm() has been reimplemented to work more efficiently for large model matrices. o "nonlin" terms defined for homogeneous factors will now accept factors specified as an interaction (using ':'). Bug fixes --------- o results now returned in original order for models fitted with 'eliminate' argument. o bug introduced into residSVD reverted so that now correctly aggregates working residuals. o anova.gnm() now works when model is a single "nonlin" term. Changes in gnm 1.0-2 ==================== Improvements ------------ o factors specified as homogenous in nonlin functions can now be specified as interactions of factors. Bug fixes --------- o fixed bug so that variables handled correctly in nonlinTerms. o corrected rank calculation for constrained models. o removed calls to Internal Changes in gnm 1.0-1 ==================== New Features ------------ o added meanResiduals() function o 'check' argument added to getContrasts. Improvements ------------ o added example SVD calculation to wheat.Rd; also in vignette. Bug fixes --------- o added update.gnm() so that nonlinear terms were no ordered as linear, first order terms. Changes in gnm 1.0-0 ==================== Improvements ------------ o eliminated coefficients now treated entirely separately, in particular the design matrix no longer has columns for these coefficients, making the algorithm far more efficient for models with many eliminated coefficients. o more reliable calculation of rank Changes in Behaviour -------------------- o ofInterest and constrain now index non-eliminated coefficients only. o eliminated coefficients now returned as attribute of returned coefficient vector. o "lsMethod" argument to gnm() removed as now the LAPACK routines are always used to determine the least squares solution at the heart of the fitting algorithm. Hence qrSolve() and cholInv() deprecated. o the "eliminate", "onlyFirstCol" and "onlyNonElim" arguments to MPinv() have been removed as no longer used. Bug fixes --------- o etastart now works for models with no linear parameters. o anova now ignores terms that are completely constrained. Changes in gnm 0.10-0 ===================== Improvements ------------ o mustart/etastart now used to obtain starting values for linear and nonlinear parameters separately, improving performance. Changes in Behaviour -------------------- o expandCategorical now groups together individuals with common covariate values, by default. New 'group' argument added to switch this behaviour. Bug fixes --------- o print.profile.gnm() now prints full result. o data now read in correctly for Lee-Carter example in vignette. Changes in gnm 0.9-9 ==================== New Features ------------ o 'etastart' and 'mustart' arguments added to gnm(). Changes in gnm 0.9-8 ==================== Improvements ------------ o gnm() now returns 'data' argument as glm() does. Changes in gnm 0.9-7 ==================== Bug fixes --------- o more minor corrections in documentation. Changes in gnm 0.9-6 ==================== Bug fixes --------- o minor corrections in documentation. Changes in gnm 0.9-5 ==================== Improvements ------------ o getContrasts() can now estimate _scaled_ contrasts with more flexibility in how the reference level is defined. o changed tolerance level in checkEstimable to 1e6 * .Machine$double.eps as previous tolerance too strict for some examples. Changes in Behaviour -------------------- o getContrasts() now only handles one set of parameters at a time. o use of Const() is now restricted to the symbolic predictors of "nonlin" functions. o Nonlin() - the wrapper function for plug-in functions - is now defunct. Use "nonlin" functions to specify custom nonlinear terms. Bug fixes --------- o plot.gnm now uses standardised Pearson residuals for plot 'which = 5' so that the Cook's distance contours are correct Changes in gnm 0.9-4 ==================== New Features ------------ o predict() now implemented for "gnm" objects Improvements ------------ o results formatted as contingency tables where appropriate by extractor functions (fitted(), etc), rather than gnm() Changes in Behaviour -------------------- o default for `match' argument of nonlinTerms() now zero vector (i.e. no matching to arguments of `call' by default) Bug fixes --------- o termPredictors() now works on "gnm" objects fitted with glm.fit() o intercept removed when `eliminate' argument of gnm() is non-NULL o models with all parameters eliminated now summarised sensibly o Diag() and Symm() now work for factors of length 1 o asGnm() now returns object with "gnm"-type terms component o print method for "profile.gnm" objects now exported Changes in gnm 0.9-3 ==================== New Features ------------ o added DrefWeights() for computing the weights in a diagonal reference term and the corresponding standard errors. Improvements ------------ o "assign" attribute now attached to the parameter vector when passed to start functions defined by "nonlin" functions, specifying the correspondence between parameters and predictors in the nonlinear term. Bug fixes --------- o start function in Dref() now identifies weight parameters correctly. o can now evaluate term predictors for "nonlin" terms that depend on covariates. Changes in gnm 0.9-2 ==================== Improvements ------------ o Calls to "nonlin" functions now evaluated in the same environment and enclosure as call to create model frame, so "nonlin" functions should be able to find variables in gnm calls - potentially useful for setting starting values. Bug fixes --------- o gnm() algorithm now reinitiates correctly when restarting after non-convergence. o gnm() now works correctly when a model is specified with nonlinear terms inbetween linear terms. Changes in gnm 0.9-1 ==================== New Features ------------ o introduction of functions of class "nonlin" for the unified specification of nonlinear terms. Mult(), Exp(), Dref() and MultHomog() have all been converted to functions of this class. o added Inv() to specify the reciprocal of a predictor. o added Const() to specify a constant in a predictor. o added instances() to specify multiple instances of a nonlinear term. Improvements ------------ o nonlinear terms can now be nested. o Exp() can now be used outside of Mult() or to exponentiate part of a constituent multiplier. Changes in Behaviour -------------------- o to accommodate the increased functionality introduced by "nonlin" functions, new labelling conventions have been introduced. In particular, most "nonlin" functions use argument-matched parameter labels. o in the new implementation of Dref() the 'formula' argument has been re-named 'delta' to provide more informative parameter labels under the new conventions. Bug fixes --------- o specifying 'ofInterest = "[?]"' in gnm() now works as documented. Changes in gnm 0.8-5 ==================== New Features ------------ none in this release Improvements ------------ o added a new `ridge' argument to gnm, to allow some control over the Levenberg-Marquardt regularization of the internal least squares calculation o changed the default ridge constant to 1e-8 (from 1e-5), to increase speed of convergence (especially in cases where there are infinite parameter estimates) o modified the "qr" method so that it no longer checks for rank deficiency (it was both unreliable, and not necessary since the matrix is regularized prior to solving) o substantial speed improvements in model fitting when there are large numbers of eliminated parameters, achieved mainly via a new internal function cholInv1(). Corresponding example timings changed in the Overview document (vignette). o speed improvements in vcov.gnm() when there are eliminated parameters; new logical argument `use.eliminate' gives control over this o in getContrasts(), added new arguments `dispersion' and `use.eliminate', both of which are passed on to vcov() o implemented faster alternatives to ifelse in gnmFit() o speed gains from use of tcrossprod(). Because of this the gnm package now requires R 2.3.0 or later. Changes in Behaviour -------------------- o in gnm(), changed the default value of argument x to TRUE (it was previously FALSE) o in checkEstimable(), changed the name of the first argument from `coefMatrix' to `combMatrix' (to reflect better that it is a matrix of coefficient *combinations*); and changed the default tolerance value to one which should give more reliable results. Also, more fundamentally, changed the check to be whether combinations are in the column span of crossprod(X) instead of the row span of X; the results should be the same, but the new version is much faster for large n. o model.matrix.gnm() no longer passes extra arguments to gnm as it's unlikely to be useful/sensible. For the same reasons it will not pass extra arguments to model.frame(), unlike model.matrix.lm() o getContrasts() now results in a list only when the `sets' argument itself is a list; otherwise (ie, normally) the result is a single object (rather than a list of objects) of class `qv' Bug fixes --------- o fixed a bug in internal function quick.glm.fit(), which greatly improves its performance. Also changed the default value of the nIter argument from 3 to 2. o fixed a small bug in demo(gnm) o fixed a bug in vcov.gnm(), which previously gave an error when data were of class "table") o fixed summary.gnm() so that it now takes proper account of the dispersion argument o in se(), added new arguments `Vcov' and `dispersion'; the latter fixes a bug, while the former minimizes wasted computation in summary.gnm o fixed bug in model.matrix.gnm() so that it can compute the model matrix even when original data is not available - unless model frame has not been saved. Original data still needed to update model frame - this is the same as for glms, etc. o fixed bug in gnm() so that reconstructing "table"-class data works for models with weights/offsets Changes in gnm 0.8-4 ==================== New Features ------------ o added "gnm" methods for profile() and confint(). Use of 'alpha' argument differs slightly from "glm" methods: see help files. o 'constrain' argument to gnm() now supplemented by 'constrainTo' argument, allowing specification of values to which parameters should be constrained. o gnm() now has 'ofInterest' argument to specify a subset of coefficients which are of interest - returned in 'ofInterest' component of "gnm" object as named numeric vector. print() summaries of model object/its components extracted by accessor functions only print coefficients of interest and (where appropriate) methods for "gnm" objects select coefficients of interest by default. o added ofInterest() and ofInterest<-() to extract/replace 'ofInterest' component of "gnm" object. o added parameters() to which returns coefficient vector with constrained parameters replaced by their constrained value. o added pickCoef() function to aid selection of coefficients - returns numeric indices of coefficients selected by Tk dialog or regular expression matching. Improvements ------------ o 'constrain' argument to gnm() now accepts a regular expression to match against coefficient names. Changes in Behaviour -------------------- o 'constrain' component of "gnm" objects is now a numeric, rather than logical, vector of indices. o all "gnm" methods for which a subset of the coefficients may be specified by numeric indices now interpret those indices as referencing the full coefficient vector (not just non-eliminated parameters). o gnm() now preserves order of terms rather than moving all linear terms to the start (this fixes bug in anova.gnm()). o the "pick" option for the 'constrain' argument to gnm() and the 'estimate' argument to se() has been replaced by "[?]" to avoid possible conflict with coefficient names/regular expressions. Bug Fixes --------- o fixed bug in se() so will now work for single parameter. o fixed bug in summary.gnm() so will now work for models with one parameter. o fixed bug in anova() so that rows of returned table are correct for models with eliminated terms. o fixed bug in eliminate() so that it now accepts interactions. o fixed bug in MPInv() so that it works for models in which all parameters are eliminated. Changes in gnm 0.8-3 ==================== Improvements ------------ o improved use of functions from other packages Bug Fixes --------- o fixed bug in asGnm.lm() where object not fully identified o corrected maintainer address in DESCRIPTION! Changes in gnm 0.8-2 ==================== New Features ------------ o added demonstration script to run using demo() o added package help file, opened by package?gnm Improvements ------------ o improved existing documentation Changes in gnm 0.8-1 ==================== New Features ------------ o added the `method' argument to MPinv(), to allow the method of calculation to be specified. Permitted values are "svd" to compute the pseudo-inverse by singular value decomposition, and "chol" to use the Cholseky decomposition instead. The latter is valid only for symmetric matrices, but is usually faster and more accurate. o added the `lsMethod' argument to gnm(), to allow specification of the numerical method used for least-squares calculations in the core of the iterative algorithm. Permitted options are "chol" and "qr". o added new function qrSolve(), which behaves like base::qr.coef but in the non-full-rank case gives the minimum-length solution rather than an arbitrary solution determined by pivoting. o added .onUnload() so that compiled code is unloaded when namespace of package is unloaded using unloadNamespace(). o added 'coef' argument to model.matrix.gnm() so that the model matrix can be evaluated at any specified value of the parameter vector. o added asGnm() generic to coerce linear model objects to gnm objects. o added exitInfo() for printing numerical details of last iteration on non-convergence of gnm(). o added new dataset, friend, to illustrate a workaround to fit a homogeneous RC(2) using gnm() - documented in help file for MultHomog(). Improvements ------------ o gnm() now takes less time per (main) iteration, due to improvements made internally in the iterative algorithm. These include pre-scaling of the local design matrix, and Levenberg-Marquardt adjustment of the least-squares solvers so that rank determination is no longer necessary. o the default convergence tolerance has been tightened (from 1e-4 to 1e-6) o modified model.matrix.gnm() so it can be used when only the namespace of gnm is loaded. Bug Fixes --------- o fixed bug in gnm() so that 'subset' now works with table data. o fixed bug in model.matrix.gnm() so can construct model matrix from "gnm" object even when original call not made in .GlobalEnv. o fixed bug in the examples on help page for House2001 data. o fixed bug so that 'formula' in gnm() now accepts '.' in formulae even when 'eliminate = NULL'. o fixed bug in getContrasts(), so that the first two columns of the qvframe component of each element of the result list are correctly named as "estimate" and "SE", as required for objects of class "qv". Changes in gnm 0.8-0 ======================= New Features ------------ o added "model.matrix" option for 'method' argument of gnm() so that model matrix can be obtained much faster. The new method is used in model.matrix.gnm() and vcov.gnm(). o added new utility function residSVD(), to facilitate the calculation of good starting values for parameters in certain Mult() terms. o added new dataset House2001, to illustrate the use of gnm in Rasch-type scaling of legislator votes. o added new utility function expandCategorical() for expanding data frame on the basis of a categorical variable. o added formula.gnm() method - returns formula from "gnm" object excluding the 'eliminate'd factor where necessary. Improvements ------------ o gnm() now takes less time to run due to improvements made in internal functions. o the fitting algorithm used by gnm() now copes better with zero-valued residuals. o output given by gnm() when 'trace' = TRUE or 'verbose' = TRUE is now displayed as it is generated on console-based versions of R. o plot.gnm() now includes option 'which' = 5 as in plot.lm() in R >= 2.2.0. Now has separate help page. o the 'constrain' argument to gnm() now accepts the names of parameters. o the 'formula' argument to gnm() now accepts '.' as described in ?terms.formula, ignoring eliminated factor if in 'data'. o interface for se() extended - can now use to find standard errors for all parameters or (a selection of) individual parameters in a gnm model. o made it possible to use gnm() with alternative fitting function. o ".Environment" attribute now attached to "gnm" objects so that gnm package loaded when workspace containing "gnm" objects is loaded. Changes in Behaviour -------------------- o start-up iterations now only update column of design matrix required in next iteration. Therefore plug-in functions using the default start-up procedure for nonlinear parameters need a localDesignFunction() with the argument 'ind' specifying the column that should be returned. o modified output given by gnm() when 'trace' = TRUE: now prints initial deviance and the deviance at the end of each iteration. o modified updates of linear parameters in starting procedure: now offset contribution of fully specified terms only. o results of summary.gnm(), vcov.gnm() and coef.gnm() now include any eliminated parameters. Print methods have been added for vcov.gnm and coef.gnm objects so that any eliminated parameters are not shown. o Mult() terms are no longer split into components by anova.gnm(), termPredictors.gnm(), labels.gnm() or the "assign" attribute of the model matrix - consistent with terms() output. o the 'eliminate' argument to gnm() must now be an expression that evaluates to a factor - this reverts the extension of 0.7-2. o when using gnm() with 'constrain' = "pick", the name(s) of the chosen parameter(s) will replace "pick" in the reurned model call. o getContrasts() now uses first level of a factor as the reference level (by default). o gnmControl() replaced by arguments to gnm(). o gnm() now uses glm.fit() for linear models (with control parameters at the gnm() defaults).unless 'eliminate'is non-NULL. o vcov.gnm() and summary.gnm() now return variance-covariance matrices including any aliased parameters. o summary.gnm() now returns standard errors with test statistics etc, where estimated parameters are identified. Bug Fixes --------- o fixed bug in summary.gnm(), anova.gnm(), termPredictors.gnm() and model.matrix.gnm() where search for model variables was incorrect. o fixed bug preventing estimation of weight parameters in Dref() terms and changed default starting values so that these parameters no longer sum to one or appear to be estimable. o corrected options for 'method' argument in gnm() help file: replaced 'method' = "coef" with 'method' = "coefNames". o fixed bug in gnm() so that it can handle tables with missing values when formatting components of fit. o hatvalues.gnm() now works for objects produced from table data. o residuals.gnm() now returns table not matrix when 'type' = "deviance" for "gnm" objects produced from table data. o hatvalues.gnm(), cooks.distance.gnm() and plot.gnm() now handle cases which are fitted exactly (giving a hat value of 1). o example fitting proportional odds model in backPain help file now works. o fixed bug in Mult() terms so that an offset can be added to a constituent multiplier without an unspecified intercept being added also. o gnm() argument 'constrain' = "pick" now allows selection of more than one constraint and is compatible with use of 'eliminate'. o gnm() can now fit models which only have the term specified by 'eliminate'. Changes in gnm 0.7-2 ======================= Improvements ------------ o Extended use of the 'eliminate' argument of gnm() to allow crossed factors - this also fixes bug which occurred when interactions were eliminated in the presence of lower order terms involving other factors Changes in Behaviour -------------------- o 'vcov' returned by gnm() now has no rank attribute (as before, the rank is returned as the separate component 'rank'). Bug Fixes --------- o Changed the calculation of 'df.residual' returned by gnm() to correctly take account of zero-weighted observations (as in glm()). o When gnm() is called with arguments 'x' = TRUE or 'VCOV' = TRUE, the returned matrices now include columns of zeros for constrained parameters. o Corrected evaluation of model frame in gnm() so that if data is missing, variables are taken from environment(formula), as documented. Modified evaluation of plug-in functions to be consistent with this, i.e. objects are taken from environment(formula) if not in model frame. o MPinv() now checks that the diagonal elements of an 'eliminate'd submatrix are all non-zero and reports an error otherwise. Changes in gnm 0.7-1 ======================= New Features ------------ o Topo() introduced for creating topological interaction factors. o anova() implemented for objects of class c("gnm", "glm"). Improvements ------------ o Diagnostic messages given by gnm() have been improved. o Step-halving introduced in main iterations of gnm() to ensure deviance is reduced at every iteration. o getContrasts() now (additionally) reports quasi standard errors, when available. o Calls to gnm() plug-in functions are now evaluated in the environment of the model frame and the enclosing environment of the parent frame of the call to gnm(). This means that variables can be found in a more standard fashion. Changes in Behaviour -------------------- o The 'data' argument of Nonlin() is defunct: Nonlin() now identifies variables to be added to the model frame as those passed to unspecified arguments of the plug-in function or those identified by a companion function to the plug-in, which is of a specified format. o The (optional) 'start' object returned by a plug-in function can no longer be a function, only a vector. However it may now include NA values, to indicate parameters which may be treated as linear for the purpose of finding starting values, given the non-NA values. Bug Fixes --------- o The 'eliminate' argument of gnm() now handles functions of variables in the given formula e.g. ~ strata(A, B), ~ as.factor(A):as.factor(B), etc. o gnm() was giving an error for models with either no linear parameters, or none specified by the 'start' argument, this is now fixed. o Long calls to plug-in functions caused problems in parsing the model formula: now fixed. o gnm() now only restarts after failing if there are unspecified nonlinear parameters. o gnm() now returns NULL if model fails. o Bug fixed in calculation of starting values for gnm() that occurred when some parameters were constrained. gnm/inst/doc/0000755000176000001440000000000012515664670012621 5ustar ripleyusersgnm/inst/doc/gnmOverview.Rnw0000644000176000001440000037057012515664670015635 0ustar ripleyusers%\VignetteIndexEntry{Generalized nonlinear models in R: An overview of the gnm package} %\VignetteKeywords{Generalized Nonlinear Models} %\VignettePackage{gnm} \documentclass[a4paper]{article} \usepackage[english]{babel} % to avoid et~al with texi2pdf \usepackage{Sweave} %\usepackage{alltt} % now replaced by environments Sinput, Soutput, Scode \usepackage{amsmath} %\usepackage{times} %\usepackage[scaled]{couriers} \usepackage{txfonts} % Times, with Belleek math font and txtt for monospaced \usepackage[scaled=0.92]{helvet} %\usepackage[T1]{fontenc} %\usepackage[expert,altbullet,lucidasmallerscale]{lucidabr} \usepackage{booktabs} \usepackage[round,authoryear]{natbib} \usepackage[left=2cm,top=2.5cm,nohead]{geometry} \usepackage{hyperref} \usepackage{array} % for paragraph columns in tables %\usepackage{moreverb} \setkeys{Gin}{width=0.6\textwidth} %% The next few definitions from "Writing Vignettes for Bioconductor Packages" %% by R Gentleman \newcommand{\Robject}[1]{{\emph{\texttt{#1}}}} \newcommand{\Rfunction}[1]{{\emph{\texttt{#1}}}} \newcommand{\Rcode}[1]{{\emph{\texttt{#1}}}} \newcommand{\Rpackage}[1]{{\textsf{#1}}} \newcommand{\Rclass}[1]{{\emph{#1}}} \newcommand{\Rmethod}[1]{{\emph{\texttt{#1}}}} \newcommand{\Rfunarg}[1]{{\emph{\texttt{#1}}}} \newcommand{\R}{\textsf{R}} \newcommand\twiddle{{\char'176}} %\setlength{\oddsidemargin}{0.5in} %\setlength{\evensidemargin}{0.5in} %\setlength{\textwidth}{5.5in} \setlength{\itemindent}{1cm} \title{Generalized nonlinear models in \R: An overview of the \Rpackage{gnm} package} \author{Heather Turner and David Firth\footnote{ This work was supported by the Economic and Social Research Council (UK) through Professorial Fellowship RES-051-27-0055.}\\ \emph{University of Warwick, UK} } \date{For \Rpackage{gnm} version \Sexpr{packageDescription("gnm")[["Version"]]} , \Sexpr{Sys.Date()}} \begin{document} \maketitle {\small \tableofcontents } <>= options(SweaveHooks = list(eval = function() options(show.signif.stars = FALSE))) @ \section{Introduction} The \Rpackage{gnm} package provides facilities for fitting \emph{generalized nonlinear models}, i.e., regression models in which the link-transformed mean is described as a sum of predictor terms, some of which may be non-linear in the unknown parameters. Linear and generalized linear models, as handled by the \Rfunction{lm} and \Rfunction{glm} functions in \R, are included in the class of generalized nonlinear models, as the special case in which there is no nonlinear term. This document gives an extended overview of the \Rpackage{gnm} package, with some examples of applications. The primary package documentation in the form of standard help pages, as viewed in \R\ by, for example, \Rcode{?gnm} or \Rcode{help(gnm)}, is supplemented rather than replaced by the present document. We begin below with a preliminary note (Section \ref{sec:glms}) on some ways in which the \Rpackage{gnm} package extends \R's facilities for specifying, fitting and working with generalized \emph{linear} models. Then (Section \ref{sec:nonlinear} onwards) the facilities for nonlinear terms are introduced, explained and exemplified. The \Rpackage{gnm} package is installed in the standard way for CRAN packages, for example by using \Rfunction{install.packages}. Once installed, the package is loaded into an \R\ session by <>= library(gnm) @ \section{Generalized linear models} \label{sec:glms} \subsection{Preamble} Central to the facilities provided by the \Rpackage{gnm} package is the model-fitting function \Rfunction{gnm}, which interprets a model formula and returns a model object. The user interface of \Rfunction{gnm} is patterned after \Rfunction{glm} (which is included in \R's standard \Rpackage{stats} package), and indeed \Rfunction{gnm} can be viewed as a replacement for \Rfunction{glm} for specifying and fitting generalized linear models. In general there is no reason to prefer \Rfunction{gnm} to \Rfunction{glm} for fitting generalized linear models, except perhaps when the model involves a large number of incidental parameters which are treatable by \Rfunction{gnm}'s \emph{eliminate} mechanism (see Section \ref{sec:eliminate}). While the main purpose of the \Rpackage{gnm} package is to extend the class of models to include nonlinear terms, some of the new functions and methods can be used also with the familiar \Rfunction{lm} and \Rfunction{glm} model-fitting functions. These are: three new data-manipulation functions \Rfunction{Diag}, \Rfunction{Symm} and \Rfunction{Topo}, for setting up structured interactions between factors; a new \Rclass{family} function, \Rfunction{wedderburn}, for modelling a continuous response variable in $[0,1]$ with the variance function $V(\mu) = \mu^2(1-\mu)^2$ as in \citet{Wedd74}; and a new generic function \Rfunction{termPredictors} which extracts the contribution of each term to the predictor from a fitted model object. These functions are briefly introduced here, before we move on to the main purpose of the package, nonlinear models, in Section \ref{sec:nonlinear}. \subsection{\Rfunction{Diag} and \Rfunction{Symm}} When dealing with \emph{homologous} factors, that is, categorical variables whose levels are the same, statistical models often involve structured interaction terms which exploit the inherent symmetry. The functions \Rfunction{Diag} and \Rfunction{Symm} facilitate the specification of such structured interactions. As a simple example of their use, consider the log-linear models of \emph{quasi-independence}, \emph{quasi-symmetry} and \emph{symmetry} for a square contingency table. \citet{Agre02}, Section 10.4, gives data on migration between regions of the USA between 1980 and 1985: <>= count <- c(11607, 100, 366, 124, 87, 13677, 515, 302, 172, 225, 17819, 270, 63, 176, 286, 10192 ) region <- c("NE", "MW", "S", "W") row <- gl(4, 4, labels = region) col <- gl(4, 1, length = 16, labels = region) @ The comparison of models reported by Agresti can be achieved as follows: <>= independence <- glm(count ~ row + col, family = poisson) quasi.indep <- glm(count ~ row + col + Diag(row, col), family = poisson) symmetry <- glm(count ~ Symm(row, col), family = poisson) quasi.symm <- glm(count ~ row + col + Symm(row, col), family = poisson) comparison1 <- anova(independence, quasi.indep, quasi.symm) print(comparison1, digits = 7) comparison2 <- anova(symmetry, quasi.symm) print(comparison2) @ The \Rfunction{Diag} and \Rfunction{Symm} functions also generalize the notions of diagonal and symmetric interaction to cover situations involving more than two homologous factors. \subsection{\Rfunction{Topo}} More general structured interactions than those provided by \Rfunction{Diag} and \Rfunction{Symm} can be specified using the function \Rfunction{Topo}. (The name of this function is short for `topological interaction', which is the nomenclature often used in sociology for factor interactions with structure derived from subject-matter theory.) The \Rfunction{Topo} function operates on any number ($k$, say) of input factors, and requires an argument named \Rfunarg{spec} which must be an array of dimension $L_1 \times \ldots \times L_k$, where $L_i$ is the number of levels for the $i$th factor. The \Rfunarg{spec} argument specifies the interaction level corresponding to every possible combination of the input factors, and the result is a new factor representing the specified interaction. As an example, consider fitting the `log-multiplicative layer effects' models described in \citet{Xie92}. The data are 7 by 7 versions of social mobility tables from \citet{Erik82}: <>= ### Collapse to 7 by 7 table as in Erikson et al. (1982) erikson <- as.data.frame(erikson) lvl <- levels(erikson$origin) levels(erikson$origin) <- levels(erikson$destination) <- c(rep(paste(lvl[1:2], collapse = " + "), 2), lvl[3], rep(paste(lvl[4:5], collapse = " + "), 2), lvl[6:9]) erikson <- xtabs(Freq ~ origin + destination + country, data = erikson) @ From sociological theory --- for which see \citet{Erik82} or \citet{Xie92} --- the log-linear interaction between origin and destination is assumed to have a particular structure: \begin{Sinput} > levelMatrix <- matrix(c(2, 3, 4, 6, 5, 6, 6, + 3, 3, 4, 6, 4, 5, 6, + 4, 4, 2, 5, 5, 5, 5, + 6, 6, 5, 1, 6, 5, 2, + 4, 4, 5, 6, 3, 4, 5, + 5, 4, 5, 5, 3, 3, 5, + 6, 6, 5, 3, 5, 4, 1), 7, 7, byrow = TRUE) \end{Sinput} The models of table 3 of \citet{Xie92} can now be fitted as follows: \begin{Sinput} > ## Null association between origin and destination > nullModel <- gnm(Freq ~ country:origin + country:destination, + family = poisson, data = erikson, verbose = FALSE) > > ## Interaction specified by levelMatrix, common to all countries > commonTopo <- update(nullModel, ~ . + + Topo(origin, destination, spec = levelMatrix), + verbose = FALSE) > > ## Interaction specified by levelMatrix, different multiplier for each country > multTopo <- update(nullModel, ~ . + + Mult(Exp(country), Topo(origin, destination, spec = levelMatrix)), + verbose = FALSE) > > ## Interaction specified by levelMatrix, different effects for each country > separateTopo <- update(nullModel, ~ . + + country:Topo(origin, destination, spec = levelMatrix), + verbose = FALSE) > > anova(nullModel, commonTopo, multTopo, separateTopo) \end{Sinput} \begin{Soutput} Analysis of Deviance Table Model 1: Freq ~ country:origin + country:destination Model 2: Freq ~ Topo(origin, destination, spec = levelMatrix) + country:origin + country:destination Model 3: Freq ~ Mult(country, Topo(origin, destination, spec = levelMatrix)) + country:origin + country:destination Model 4: Freq ~ country:origin + country:destination + country:Topo(origin, destination, spec = levelMatrix) Resid. Df Resid. Dev Df Deviance 1 108 4860.0 2 103 244.3 5 4615.7 3 101 216.4 2 28.0 4 93 208.5 8 7.9 \end{Soutput} Here we have used \Rfunction{gnm} to fit all of these log-link models; the first, second and fourth are log-linear and could equally well have been fitted using \Rfunction{glm}. \subsection{The \Rfunction{wedderburn} family} In \citet{Wedd74} it was suggested to represent the mean of a continuous response variable in $[0,1]$ using a quasi-likelihood model with logit link and the variance function $\mu^2(1-\mu)^2$. This is not one of the variance functions made available as standard in \R's \Rfunction{quasi} family. The \Rfunction{wedderburn} family provides it. As an example, Wedderburn's analysis of data on leaf blotch on barley can be reproduced as follows: <>= ## data from Wedderburn (1974), see ?barley logitModel <- glm(y ~ site + variety, family = wedderburn, data = barley) fit <- fitted(logitModel) print(sum((barley$y - fit)^2 / (fit * (1-fit))^2)) @ This agrees with the chi-squared value reported on page 331 of \citet{McCu89}, which differs slightly from Wedderburn's own reported value. \subsection{\Rfunction{termPredictors}} \label{sec:termPredictors} The generic function \Rfunction{termPredictors} extracts a term-by-term decomposition of the predictor function in a linear, generalized linear or generalized nonlinear model. As an illustrative example, we can decompose the linear predictor in the above quasi-symmetry model as follows: <>= print(temp <- termPredictors(quasi.symm)) rowSums(temp) - quasi.symm$linear.predictors @ Such a decomposition might be useful, for example, in assessing the relative contributions of different terms or groups of terms. \section{Nonlinear terms} \label{sec:nonlinear} The main purpose of the \Rpackage{gnm} package is to provide a flexible framework for the specification and estimation of generalized models with nonlinear terms. The facility provided with \Rfunction{gnm} for the specification of nonlinear terms is designed to be compatible with the symbolic language used in \Rclass{formula} objects. Primarily, nonlinear terms are specified in the model formula as calls to functions of the class \Rclass{nonlin}. There are a number of \Rclass{nonlin} functions included in the \Rpackage{gnm} package. Some of these specify simple mathematical functions of predictors: \Rfunction{Exp}, \Rfunction{Mult}, and \Rfunction{Inv}. %\Rfunction{Log}, \Rfunction{Raise} (to raise to a constant power), and \Rfunction{Logit}. Others specify more specialized nonlinear terms, in particular \Rfunction{MultHomog} specifies homogeneous multiplicative interactions and \Rfunction{Dref} specifies diagonal reference terms. Users may also define their own \Rclass{nonlin} functions. \subsection{Basic mathematical functions of predictors} \label{sec:Basic} Most of the \Rclass{nonlin} functions included in \Rpackage{gnm} are basic mathematical functions of predictors: \begin{description} \setlength{\itemindent}{-0.5cm} \item[\Rfunction{Exp}:] the exponential of a predictor \item[\Rfunction{Inv}:] the reciprocal of a predictor %\item[\Rfunction{Log}:] the natural logarithm of a predictor %\item[\Rfunction{Logit}:] the logit of a predictor \item[\Rfunction{Mult}:] the product of predictors %\item[\Rfunction{Raise}:] a predictor raised to a constant power \end{description} Predictors are specified by symbolic expressions that are interpreted as the right-hand side of a \Rclass{formula} object, except that an intercept is \textbf{not} added by default. The predictors may contain nonlinear terms, allowing more complex functions to be built up. For example, suppose we wanted to specify a logistic predictor with the same form as that used by \Rfunction{SSlogis} (a selfStart model for use with \Rfunction{nls} --- see section~\ref{sec:gnmVnls} for more on \Rfunction{gnm} vs.\ \Rfunction{nls}): \[\frac{\text{Asym}}{1 + \exp((\text{xmid} - x)/\text{scal})}.\] This expression could be simplified by re-parameterizing in terms of xmid/scal and 1/scal, however we shall continue with this form for illustration. We could express this predictor symbolically as follows \begin{Scode} ~ -1 + Mult(1, Inv(Const(1) + Exp(Mult(1 + offset(-x), Inv(1))))) \end{Scode} where \Rfunction{Const} is a convenience function to specify a constant in a \Rclass{nonlin} term, equivalent to \Rcode{offset(rep(1, nObs))} where \Robject{nObs} is the number of observations. However, this is rather convoluted and it may be preferable to define a specialized \Rclass{nonlin} function in such a case. Section \ref{sec:nonlin.functions} explains how users can define custom \Rclass{nonlin} functions, with a function to specify logistic terms as an example. One family of models usefully specified with the basic functions is the family of models with multiplicative interactions. For example, the row-column association model \[ \log \mu_{rc} = \alpha_r + \beta_c + \gamma_r\delta_c, \] also known as the Goodman RC model \citep{Good79}, would be specified as a log-link model (for response variable \Robject{resp}, say), with formula \begin{Scode} resp ~ R + C + Mult(R, C) \end{Scode} where \Robject{R} and \Robject{C} are row and column factors respectively. In some contexts, it may be desirable to constrain one or more of the constituent multipliers\footnote{ A note on terminology: the rather cumbersome phrase `constituent multiplier', or sometimes the abbreviation `multiplier', will be used throughout this document in preference to the more elegant and standard mathematical term `factor'. This will avoid possible confusion with the completely different meaning of the word `factor' --- that is, a categorical variable --- in \R. } in a multiplicative interaction to be nonnegative . This may be achieved by specifying the multiplier as an exponential, as in the following `uniform difference' model \citep{Xie92, Erik92} \[ \log \mu_{rct} = \alpha_{rt} + \beta_{ct} + e^{\gamma_t}\delta_{rc}, \] which would be represented by a formula of the form \begin{Scode} resp ~ R:T + C:T + Mult(Exp(T), R:C) \end{Scode} \subsection{\Rfunction{MultHomog}} \Rfunction{MultHomog} is a \Rclass{nonlin} function to specify multiplicative interaction terms in which the constituent multipliers are the effects of two or more factors and the effects of these factors are constrained to be equal when the factor levels are equal. The arguments of \Rfunction{MultHomog} are the factors in the interaction, which are assumed to be objects of class \Rclass{factor}. As an example, consider the following association model with homogeneous row-column effects: \[\log \mu_{rc} = \alpha_r + \beta_c + \theta_{r}I(r=c) + \gamma_r\gamma_c.\] To fit this model, with response variable named \Robject{resp}, say, the formula argument to \Rfunction{gnm} would be \begin{Scode} resp ~ R + C + Diag(R, C) + MultHomog(R, C) \end{Scode} If the factors passed to \Rfunction{MultHomog} do not have exactly the same levels, a common set of levels is obtained by taking the union of the levels of each factor, sorted into increasing order. \subsection{\Rfunction{Dref}} \label{sec:Dref function} \Rfunction{Dref} is a \Rclass{nonlin} function to fit diagonal reference terms \citep{Sobe81, Sobe85} involving two or more factors with a common set of levels. A diagonal reference term comprises an additive component for each factor. The component for factor $f$ is given by \[ w_f\gamma_l \] for an observation with level $l$ of factor $f$, where $w_f$ is the weight for factor $f$ and $\gamma_l$ is the ``diagonal effect'' for level $l$. The weights are constrained to be nonnegative and to sum to one so that a ``diagonal effect'', say $\gamma_l$, is the value of the diagonal reference term for data points with level $l$ across the factors. \Rfunction{Dref} specifies the constraints on the weights by defining them as \[ w_f = \frac{e^{\delta_f}}{\sum_i e^{\delta_i}} \] where the $\delta_f$ are the parameters to be estimated. Factors defining the diagonal reference term are passed as unspecified arguments to \Rfunction{Dref}. For example, the following diagonal reference model for a contingency table classified by the row factor \Robject{R} and the column factor \Robject{C}, \[ \mu_{rc} =\frac{e^{\delta_1}}{e^{\delta_1} + e^{\delta_2}}\gamma_r + \frac{e^{\delta_2}}{e^{\delta_1} + e^{\delta_2}}\gamma_c, \] would be specified by a formula of the form \begin{Scode} resp ~ -1 + Dref(R, C) \end{Scode} The \Rfunction{Dref} function has one specified argument, \Rfunarg{delta}, which is a formula with no left-hand side, specifying the dependence (if any) of $\delta_f$ on covariates. For example, the formula \begin{Scode} resp ~ -1 + x + Dref(R, C, delta = ~ 1 + x) \end{Scode} specifies the generalized diagonal reference model \[ \mu_{rci} = \beta x_i + \frac{e^{\xi_{01} + \xi_{11}x_i}}{e^{\xi_{01} + \xi_{11}x_i} + e^{\xi_{02} + \xi_{12}x_i}}\gamma_r + \frac{e^{\xi_{02} + \xi_{12}x_i}}{e^{\xi_{01} + \xi_{11}x_i} + e^{\xi_{02} + \xi_{12}x_i}}\gamma_c. \] The default value of \Rfunarg{delta} is \Robject{\twiddle 1}, so that constant weights are estimated. The coefficients returned by \Rfunction{gnm} are those that are directly estimated, i.e. the $\delta_f$ or the $\xi_{.f}$, rather than the implied weights $w_f$. However, these weights may be obtained from a fitted model using the \Rfunction{DrefWeights} function, which computes the corresponding standard errors using the delta method. \subsection{\Rfunction{instances}} \label{sec:instances} Multiple instances of a linear term will be aliased with each other, but this is not necessarily the case for nonlinear terms. Indeed, there are certain types of model where adding further instances of a nonlinear term is a natural way to extend the model. For example, Goodman's RC model, introduced in section \ref{sec:Basic} \[ \log \mu_{rc} = \alpha_r + \beta_c + \gamma_r\delta_c, \] is naturally extended to the RC(2) model, with a two-component interaction \[ \log \mu_{rc} = \alpha_r + \beta_c + \gamma_r\delta_c + \theta_r\phi_c. \] Currently all of the \Rclass{nonlin} functions in \Rpackage{gnm} except \Rpackage{Dref} have an \Rfunarg{inst} argument to allow the specification of multiple instances. So the RC(2) model could be specified as follows \begin{Scode} resp ~ R + C + Mult(R, C, inst = 1) + Mult(R, C, inst = 2) \end{Scode} The convenience function \Rfunction{instances} allows multiple instances of a term to be specified at once \begin{Scode} resp ~ R + C + instances(Mult(R, C), 2) \end{Scode} The formula is expanded by \Rfunction{gnm}, so that the instances are treated as separate terms. The \Rfunction{instances} function may be used with any function with an \Rfunarg{inst} argument. \subsection{Custom \Rclass{nonlin} functions} \label{sec:nonlin.functions} \subsubsection{General description} Users may write their own \Rclass{nonlin} functions to specify nonlinear terms which can not (easily) be specified using the \Rclass{nonlin} functions in the \Rpackage{gnm} package. A function of class \Rclass{nonlin} should return a list of arguments for the internal function \Rfunction{nonlinTerms}. The following arguments must be specified in all cases: \begin{description} \setlength{\itemindent}{-0.5cm} \item[\Robject{predictors}:] a list of symbolic expressions or formulae with no left hand side which represent (possibly nonlinear) predictors that form part of the term. \item[\Robject{term}:] a function that takes the arguments \Rfunarg{predLabels} and \Rfunarg{varLabels}, which are labels generated by \Rfunction{gnm} for the specified predictors and variables (see below), and returns a deparsed mathematical expression of the nonlinear term. Only functions recognised by \Rfunction{deriv} should be used in the expression, e.g. \Rfunction{+} rather than \Rfunction{sum}. \end{description} Intercepts are added by default to predictors that are specified by formulae. If predictors are named, these names are used as a prefix for parameter labels or as the parameter label itself in the single-parameter case. The following arguments of \Rfunction{nonlinTerms} must be specified whenever applicable to the nonlinear term: \begin{description} \setlength{\itemindent}{-0.5cm} \item[\Robject{variables}:] a list of expressions representing variables in the term (variables with a coefficient of 1). \item[\Robject{common}:] a numeric index of \Rfunarg{predictors} with duplicated indices identifying single factor predictors for which homologous effects are to be estimated. \end{description} The arguments below are optional: \begin{description} \setlength{\itemindent}{-0.5cm} \item[\Robject{call}:] a call to be used as a prefix for parameter labels. \item[\Robject{match}:] (if \Robject{call} is non-\Rcode{NULL}) a numeric index of \Robject{predictors} specifying which arguments of \Robject{call} the predictors match to --- zero indicating no match. If \Rcode{NULL}, predictors will not be matched to the arguments of \Robject{call}. \item[\Robject{start}:] a function which takes a named vector of parameters corresponding to the predictors and returns a vector of starting values for those parameters. This function is ignored if the term is nested within another nonlinear term. \end{description} Predictors which are matched to a specified argument of \Robject{call} should be given the same name as the argument. Matched predictors are labelled using ``dot-style'' labelling, e.g. the label for the intercept in the first constituent multiplier of the term \Rcode{Mult(A, B)} would be \Rcode{"Mult(.\ + A, 1 + B).(Intercept)"}. It is recommended that matches are specified wherever possible, to ensure parameter labels are well-defined. The arguments of \Rclass{nonlin} functions are as suited to the particular term, but will usually include symbolic representations of predictors in the term and/or the names of variables in the term. The function may also have an \Rfunarg{inst} argument to allow specification of multiple instances (see \ref{sec:instances}). \subsubsection{Example: a logistic function} As an example, consider writing a \Rclass{nonlin} function for the logistic term discussed in \ref{sec:Basic}: \[\frac{\text{Asym}}{1 + \exp((\text{xmid} - x)/\text{scal})}.\] We can consider \emph{Asym}, \emph{xmid} and \emph{scal} as the parameters of three separate predictors, each with a single intercept term. Thus we specify the \Rfunarg{predictors} argument to \Rfunction{nonlinTerms} as \begin{Scode} predictors = list(Asym = 1, xmid = 1, scal = 1) \end{Scode} The term also depends on the variable $x$, which would need to be specified by the user. Suppose this is specified to our \Rclass{nonlin} function through an argument named \Rfunarg{x}. Then our \Rclass{nonlin} function would specify the following \Rfunarg{variables} argument \begin{Scode} variables = list(substitute(x)) \end{Scode} We need to use \Rfunction{substitute} here to list the variable specified by the user rather than the variable named \Rcode{``x''} (if it exists). Our \Rclass{nonlin} function must also specify the \Rfunarg{term} argument to \Rfunction{nonlinTerms}. This is a function that will paste together an expression for the term, given labels for the predictors and the variables: \begin{Scode} term = function(predLabels, varLabels) { paste(predLabels[1], "/(1 + exp((", predLabels[2], "-", varLabels[1], ")/", predLabels[3], "))") } \end{Scode} We now have all the necessary ingredients of a \Rclass{nonlin} function to specify the logistic term. Since the parameterization does not depend on user-specified values, it does not make sense to use call-matched labelling in this case. The labels for our parameters will be taken from the labels of the \Rfunarg{predictors} argument. Since we do not anticipate fitting models with multiple logistic terms, our \Rclass{nonlin} function will not specify a \Rfunarg{call} argument with which to prefix the parameter labels. We do however, have some idea of useful starting values, so we will specify the \Rfunarg{start} argument as \begin{Scode} start = function(theta){ theta[3] <- 1 theta } \end{Scode} which sets the initial scale parameter to one. Putting all these ingredients together we have \begin{Scode} Logistic <- function(x){ list(predictors = list(Asym = 1, xmid = 1, scal = 1), variables = list(substitute(x)), term = function(predLabels, varLabels) { paste(predLabels[1], "/(1 + exp((", predLabels[2], "-", varLabels[1], ")/", predLabels[3], "))") }, start = function(theta){ theta[3] <- 1 theta }) } class(Logistic) <- "nonlin" \end{Scode} \subsubsection{Example: \Rfunction{MultHomog}} The \Rfunction{MultHomog} function included in the \Rpackage{gnm} package provides a further example of a \Rclass{nonlin} function, showing how to specify a term with quite different features from the preceding example. The definition is \begin{Scode} MultHomog <- function(..., inst = NULL){ dots <- match.call(expand.dots = FALSE)[["..."]] list(predictors = dots, common = rep(1, length(dots)), term = function(predLabels, ...) { paste("(", paste(predLabels, collapse = ")*("), ")", sep = "")}, call = as.expression(match.call())) } class(MultHomog) <- "nonlin" \end{Scode} Firstly, the interaction may be based on any number of factors, hence the use of the special ``\Rfunarg{...}'' argument. The use of \Rfunction{match.call} is analogous to the use of \Rfunction{substitute} in the \Rfunction{Logistic} function: to obtain expressions for the factors as specified by the user. The returned \Rfunarg{common} argument specifies that homogeneous effects are to be estimated across all the specified factors. The term only depends on these factors, but the \Rfunarg{term} function allows for the empty \Robject{varLabels} vector that will be passed to it, by having a ``\Rfunarg{...}'' argument. Since the user may wish to specify multiple instances, the \Rfunarg{call} argument to \Rfunction{nonlinTerms} is specified, so that parameters in different instances of the term will have unique labels (due to the \Rfunarg{inst} argument in the call). However as the expressions passed to ``\Rfunarg{...}'' may only represent single factors, rather than general predictors, it is not necessary to use call-matched labelling, so the \Rfunarg{match} argument is not specified here. % Dref starting values as example of ensuring the arbitrariness of the final % parameterization is emphasised (see old plug-in section)? \section{Controlling the fitting procedure} The \Rfunction{gnm} function has a number of arguments which affect the way a model will be fitted. Basic control parameters can be set using the arguments %\Rfunarg{checkLinear}, \Rfunarg{lsMethod}, \Rfunarg{ridge}, \Rfunarg{tolerance}, \Rfunarg{iterStart} and \Rfunarg{iterMax}. Starting values for the parameter estimates can be set by \Rfunarg{start} or they can be generated from starting values for the predictors on the link or response scale via \Rfunarg{etastart} or \Rfunarg{mustart} respectively. Parameters can be constrained via \Rfunarg{constrain} and \Rfunarg{constrainTo} arguments, while parameters of a stratification factor can be handled more efficiently by specifying the factor in an \Rfunarg{eliminate} argument. These options are described in more detail below. \subsection{Basic control parameters} %By default, \Rfunction{gnm} will use \Rfunction{glm.fit} to fit models where the %predictor is linear and \Rfunarg{eliminate} is \Rcode{NULL}. This behaviour can %be overridden by setting \Rfunarg{checkLinear} to \Rcode{FALSE}. %%% At present there is no advantage to doing this! Parameterization would be %%% the same. The arguments \Rfunarg{iterStart} and \Rfunarg{iterMax} control respectively the number of starting iterations (where applicable) and the number of main iterations used by the fitting algorithm. The progress of these iterations can be followed by setting either \Rfunarg{verbose} or \Rfunarg{trace} to \Robject{TRUE}. If \Rfunarg{verbose} is \Robject{TRUE} and \Rfunarg{trace} is \Robject{FALSE}, which is the default setting, progress is indicated by printing the character ``.'' at the beginning of each iteration. If \Rfunarg{trace} is \Robject{TRUE}, the deviance is printed at the beginning of each iteration (over-riding the printing of ``.'' if necessary). Whenever \Rfunarg{verbose} is \Robject{TRUE}, additional messages indicate each stage of the fitting process and diagnose any errors that cause that cause the algorithm to restart. Prior to solving the (typically rank-deficient) least squares problem at the heart of the \Rfunction{gnm} fitting algorithm, the design matrix is standardized and regularized (in the Levenberg-Marquardt sense); the \Rfunarg{ridge} argument provides a degree of control over the regularization performed (smaller values may sometimes give faster convergence but can lead to numerical instability). The fitting algorithm will terminate before the number of main iterations has reached \Rfunarg{iterMax} if the convergence criteria have been met, with tolerance specified by \Rfunarg{tolerance}. Convergence is judged by comparing the squared components of the score vector with corresponding elements of the diagonal of the Fisher information matrix. If, for all components of the score vector, the ratio is less than \Robject{tolerance\^{}2}, or the corresponding diagonal element of the Fisher information matrix is less than 1e-20, the algorithm is deemed to have converged. \subsection{Specifying starting values} \label{sec:start} \subsubsection{Using \Rfunarg{start}} In some contexts, the default starting values may not be appropriate and the fitting algorithm will fail to converge, or perhaps only converge after a large number of iterations. Alternative starting values may be passed on to \Rfunction{gnm} by specifying a \Rfunarg{start} argument. This should be a numeric vector of length equal to the number of parameters (or possibly the non-eliminated parameters, see Section \ref{sec:eliminate}), however missing starting values (\Robject{NA}s) are allowed. If there is no user-specified starting value for a parameter, the default value is used. This feature is particularly useful when adding terms to a model, since the estimates from the original model can be used as starting values, as in this example: \begin{Scode} model1 <- gnm(mu ~ R + C + Mult(R, C)) model2 <- gnm(mu ~ R + C + instances(Mult(R, C), 2), start = c(coef(model1), rep(NA, 10))) \end{Scode} The \Rfunction{gnm} call can be made with \Rcode{method = "coefNames"} to identify the parameters of a model prior to estimation, to assist with the specification of arguments such as \Rfunarg{start}. For example, to get the number \Rcode{10} for the value of \Rfunarg{start} above, we could have done \begin{Scode} gnm(mu ~ R + C + instances(Mult(R, C), 2), method = "coefNames") \end{Scode} from whose output it would be seen that there are 10 new coefficients in \Robject{model2}. When called with \Rcode{method = "coefNames"}, \Rfunction{gnm} makes no attempt to fit the specified model; instead it returns just the names that the coefficients in the fitted model object would have. The starting procedure used by \Rfunction{gnm} is as follows: \begin{enumerate} \item Begin with all parameters set to \Rcode{NA}. \item \label{i:nonlin} Replace \Rcode{NA} values with any starting values set by \Rclass{nonlin} functions. \item \label{i:start} Replace current values with any (non-\Rcode{NA}) starting values specified by the \Rfunarg{start} argument of \Rfunction{gnm}. \item \label{i:constrain} Set any values specified by the \Rfunarg{constrain} argument to the values specified by the \Rfunarg{constrainTo} argument (see Section \ref{sec:constrain}). \item \label{i:gnmStart} Categorise remaining \Rcode{NA} parameters as linear or nonlinear, treating non-\Rcode{NA} parameters as fixed. Initialise the nonlinear parameters by generating values $\theta_i$ from the Uniform($-0.1$, $0.1$) distribution and shifting these values away from zero as follows \begin{equation*} \theta_i = \begin{cases} \theta_i - 0.1 & \text{if } \theta_i < 1 \\ \theta_i + 0.1 & \text{otherwise} \end{cases} \end{equation*} \item Compute the \Rfunction{glm} estimate of the linear parameters, offsetting the contribution to the predictor of any terms fully determined by steps \ref{i:nonlin} to \ref{i:gnmStart}. \item \label{i:iter} Run starting iterations: update nonlinear parameters one at a time, jointly re-estimating linear parameters after each round of updates. \end{enumerate} Note that no starting iterations (step \ref{i:iter}) will be run if all parameters are linear, or if all nonlinear parameters are specified by \Rfunarg{start}, \Rfunarg{constrain} or a \Rclass{nonlin} function. \subsubsection{Using \Rfunarg{etastart} or \Rfunarg{mustart}} An alternative way to set starting values for the parameters is to specify starting values for the predictors. If there are linear parameters in the model, the predictor starting values are first used to fit a model with only the linear terms (offsetting any terms fully specified by starting values given by \Rfunarg{start}, \Rfunarg{constrain} or a \Rclass{nonlin} function). In this case the parameters corresponding to the predictor starting values can be computed analytically. If the fitted model reproduces the predictor starting values, then these values contain no further information and they are replaced using the \Rfunction{initialize} function of the specified \Rfunarg{family}. The predictor starting values or their replacement are then used as the response variable in a nonlinear least squares model with only the unspecified nonlinear terms, offsetting the contribution of any other terms. Since the model is over-parameterized, the model is approximated using \Rfunarg{iterStart} iterations of the ``L-BFGS-B'' algorithm of \Rfunction{optim}, assuming parameters lie in the range (-10, 10). Starting values for the predictors can be specified explicitly via \Rfunarg{etastart} or implicitly by passing starting values for the fitted means to \Rfunarg{mustart}. For example, when extending a model, the fitted predictors from the first model can be used to find starting values for the parameters of the second model: \begin{Scode} model1 <- gnm(mu ~ R + C + Mult(R, C)) model2 <- gnm(mu ~ R + C + instances(Mult(R, C), 2), etastart = model1$predictors) \end{Scode} %$ Using \Rfunction{etastart} avoids the one-parameter-at-a-time starting iterations, so is quicker than using \Rfunction{start} to pass on information from a nested model. However \Rfunction{start} will generally produce better starting values so should be used when feasible. For multiplicative terms, the \Rfunction{residSVD} functions provides a better way to avoid starting iterations. \subsection{Using \Rfunarg{constrain}} \label{sec:constrain} By default, \Rfunction{gnm} only imposes identifiability constraints according to the general conventions used by \Robject{R} to handle linear aliasing. Therefore models that have any nonlinear terms will be typically be over-parameterized, and \Rfunction{gnm} will return a random parameterization for unidentified coefficients (determined by the randomly chosen starting values for the iterative algorithm, step 5 above). To illustrate this point, consider the following application of \Rfunction{gnm}, discussed later in Section \ref{sec:RCmodels}: <>= set.seed(1) RChomog1 <- gnm(Freq ~ origin + destination + Diag(origin, destination) + MultHomog(origin, destination), family = poisson, data = occupationalStatus, verbose = FALSE) @ Running the analysis again from a different seed <>= set.seed(2) RChomog2 <- update(RChomog1) @ gives a different representation of the same model: <>= compareCoef <- cbind(coef(RChomog1), coef(RChomog2)) colnames(compareCoef) <- c("RChomog1", "RChomog2") round(compareCoef, 4) @ Even though the linear terms are constrained, the parameter estimates for the main effects of \Robject{origin} and \Robject{destination} still change, because these terms are aliased with the higher order multiplicative interaction, which is unconstrained. Standard errors are only meaningful for identified parameters and hence the output of \Rmethod{summary.gnm} will show clearly which coefficients are estimable: <>= summary(RChomog2) @ Additional constraints may be specified through the \Rfunarg{constrain} and \Rfunarg{constrainTo} arguments of \Rfunction{gnm}. These arguments specify respectively parameters that are to be constrained in the fitting process and the values to which they should be constrained. Parameters may be specified by a regular expression to match against the parameter names, a numeric vector of indices, a character vector of names, or, if \Rcode{constrain = "[?]"} they can be selected through a \emph{Tk} dialog. The values to constrain to should be specified by a numeric vector; if \Rfunarg{constrainTo} is missing, constrained parameters will be set to zero. In the case above, constraining one level of the homogeneous multiplicative factor is sufficient to make the parameters of the nonlinear term identifiable, and hence all parameters in the model identifiable. Figure~\ref{fig:Tk} illustrates how the coefficient to be constrained may be specified via a \emph{Tk} dialog, an approach which can be helpful in interactive R sessions. % here illustrate TclTk dialog, but explain other methods better for reproducibility \begin{figure}[tp] \centering \begin{tabular}[!h]{m{0.6\linewidth}m{0.4\linewidth}} \scalebox{0.9}{\includegraphics{screenshot1.png}} & When \Rfunction{gnm} is called with \Rcode{constrain = "[?]"}, a \emph{Tk} dialog is shown listing the coefficients in the model.\\ \scalebox{0.9}{\includegraphics{screenshot2.png}} & Scroll through the coefficients and click to select a single coefficient to constrain. To select multiple coefficients, hold down the \texttt{Ctrl} key whilst clicking. The \texttt{Add} button will become active when coefficient(s) have been selected.\\ \scalebox{0.9}{\includegraphics{screenshot3.png}} & Click the \texttt{Add} button to add the selected coefficients to the list of coefficients to be constrained. To remove coefficients from the list, select the coefficients in the right pane and click \texttt{Remove}. Click \texttt{OK} when you have finalised the list.\\ \end{tabular} \caption{Selecting coefficients to constrain with the \emph{Tk} dialog.} \label{fig:Tk} \end{figure} However for reproducible code, it is best to specify the constrained coefficients directly. For example, the following code specifies that the last level of the homogeneous multiplicative factor should be constrained to zero, <>= set.seed(1) RChomogConstrained1 <- update(RChomog1, constrain = length(coef(RChomog1))) @ Since all the parameters are now constrained, re-fitting the model will give the same results, regardless of the random seed set beforehand: <>= set.seed(2) RChomogConstrained2 <- update(RChomogConstrained1) identical(coef(RChomogConstrained1), coef(RChomogConstrained2)) @ It is not usually so straightforward to constrain all the parameters in a generalized nonlinear model. However use of \Rfunarg{constrain} in conjunction with \Rfunarg{constrainTo} is usually sufficient to make coefficients of interest identifiable . The functions \Rfunction{checkEstimable} or \Rfunction{getContrasts}, described in Section \ref{sec:Methods}, may be used to check whether particular combinations of parameters are estimable. \subsection{Using \Rfunarg{eliminate}} \label{sec:eliminate} When a model contains the additive effect of a factor which has a large number of levels, the iterative algorithm by which maximum likelihood estimates are computed can usually be accelerated by use of the \Rfunarg{eliminate} argument to \Rfunction{gnm}. A factor passed to \Rfunarg{eliminate} specifies the first term in the model, replacing any intercept term. So, for example \begin{Scode} gnm(mu ~ A + B + Mult(A, B), eliminate = strata1:strata2) \end{Scode} is equivalent, in terms of the structure of the model, to \begin{Scode} gnm(mu ~ -1 + strata1:strata2 + A + B + Mult(A, B)) \end{Scode} However, specifying a factor through \Rfunarg{eliminate} has two advantages over the standard specification. First, the structure of the eliminated factor is exploited so that computational speed is improved --- substantially so if the number of eliminated parameters is large. Second, eliminated parameters are returned separately from non-eliminated parameters (as an attribute of the \Robject{coefficients} component of the returned object). Thus eliminated parameters are excluded from printed model summaries by default and disregarded by \Rclass{gnm} methods that would not be relevant to such parameters (see Section \ref{sec:Methods}). The \Rfunarg{eliminate} feature is useful, for example, when multinomial-response models are fitted by using the well known equivalence between multinomial and (conditional) Poisson likelihoods. In such situations the sufficient statistic involves a potentially large number of fixed multinomial row totals, and the corresponding parameters are of no substantive interest. For an application see Section \ref{sec:Stereotype} below. Here we give an artificial illustration: 1000 randomly-generated trinomial responses, and a single predictor variable (whose effect on the data generation is null): <>= set.seed(1) n <- 1000 x <- rep(rnorm(n), rep(3, n)) counts <- as.vector(rmultinom(n, 10, c(0.7, 0.1, 0.2))) rowID <- gl(n, 3, 3 * n) resp <- gl(3, 1, 3 * n) @ The logistic model for dependence on \Robject{x} can be fitted as a Poisson log-linear model\footnote{For this particular example, of course, it would be more economical to fit the model directly using \Rfunction{multinom} (from the recommended package \Rpackage{nnet}). But fitting as here via the `Poisson trick' allows the model to be elaborated within the \Rpackage{gnm} framework using \Rfunction{Mult} or other \Rclass{nonlin} terms.}, using either \Rfunction{glm} or \Rfunction{gnm}: \begin{Sinput} > ## Timings on a Xeon 2.33GHz, under Linux > system.time(temp.glm <- glm(counts ~ rowID + resp + resp:x, family = poisson))[1] \end{Sinput} \begin{Soutput} user.self 37.126 \end{Soutput} \begin{Sinput} > system.time(temp.gnm <- gnm(counts ~ resp + resp:x, eliminate = rowID, family = poisson, verbose = FALSE))[1] \end{Sinput} \begin{Soutput} user.self 0.04 \end{Soutput} \begin{Sinput} > c(deviance(temp.glm), deviance(temp.gnm)) \end{Sinput} \begin{Soutput} [1] 2462.556 2462.556 \end{Soutput} Here the use of \Rfunarg{eliminate} causes the \Rfunction{gnm} calculations to run much more quickly than \Rfunction{glm}. The speed advantage increases with the number of eliminated parameters (here 1000). By default,the eliminated parameters do not appear in printed model summaries as here: \begin{Sinput} > summary(temp.gnm) \end{Sinput} \begin{Soutput} Call: gnm(formula = counts ~ resp + resp:x, eliminate = rowID, family = poisson, verbose = FALSE) Deviance Residuals: Min 1Q Median 3Q Max -2.852038 -0.786172 -0.004534 0.645278 2.755013 Coefficients of interest: Estimate Std. Error z value Pr(>|z|) resp2 -1.961448 0.034007 -57.678 <2e-16 resp3 -1.255846 0.025359 -49.523 <2e-16 resp1:x -0.007726 0.024517 -0.315 0.753 resp2:x -0.023340 0.037611 -0.621 0.535 resp3:x 0.000000 NA NA NA (Dispersion parameter for poisson family taken to be 1) Std. Error is NA where coefficient has been constrained or is unidentified Residual deviance: 2462.6 on 1996 degrees of freedom AIC: 12028 Number of iterations: 4 \end{Soutput} although the \Rmethod{summary} method has a logical \Rfunarg{with.eliminate} that can toggled so that the eliminated parameters are included if desired. The \Rfunarg{eliminate} feature as implemented in \Rpackage{gnm} extends the earlier work of \cite{Hatz04} to a broader class of models and to over-parameterized model representations. \section{Methods and accessor functions} \label{sec:Methods} \subsection{Methods} \label{sec:specificMethods} The \Rfunction{gnm} function returns an object of class \Robject{c("gnm", "glm", "lm")}. There are several methods that have been written for objects of class \Rclass{glm} or \Rclass{lm} to facilitate inspection of fitted models. Out of the generic functions in the \Rpackage{base}, \Rpackage{stats} and \Rpackage{graphics} packages for which methods have been written for \Rclass{glm} or \Rclass{lm} objects, Figure \ref{fig:glm.lm} shows those that can be used to analyse \Rclass{gnm} objects, whilst Figure \ref{fig:!glm.lm} shows those that are not implemented for \Rclass{gnm} objects. \begin{figure}[!tbph] \centering \begin{fbox} { \begin{tabular*}{7.5cm}{@{\extracolsep{\fill}}lll@{\extracolsep{\fill}}} add1$^*$ & family & print \\ anova & formula & profile \\ case.names & hatvalues & residuals \\ coef & labels & rstandard \\ cooks.distance & logLik & summary \\ confint & model.frame & variable.names \\ deviance & model.matrix & vcov \\ drop1$^*$ & plot & weights \\ extractAIC & predict & \\ \end{tabular*} } \end{fbox} \caption{Generic functions in the \Rpackage{base}, \Rpackage{stats} and \Rpackage{graphics} packages that can be used to analyse \Rclass{gnm} objects. Starred functions are implemented for models with linear terms only.} \label{fig:glm.lm} \end{figure} \begin{figure}[!tbph] \centering \begin{fbox} { \begin{tabular*}{4.5cm}{@{\extracolsep{\fill}}ll@{\extracolsep{\fill}}} alias & effects \\ dfbeta & influence \\ dfbetas & kappa \\ dummy.coef & proj \\ \end{tabular*} } \end{fbox} \caption{Generic functions in the \Rpackage{base}, \Rpackage{stats} and \Rpackage{graphics} packages for which methods have been written for \Rclass{glm} or \Rclass{lm} objects, but which are \emph{not} implemented for \Rclass{gnm} objects.} \label{fig:!glm.lm} \end{figure} In addition to the accessor functions shown in Figure \ref{fig:glm.lm}, the \Rpackage{gnm} package provides a new generic function called \Rfunction{termPredictors} that has methods for objects of class \Rclass{gnm}, \Rclass{glm} and \Rclass{lm}. This function returns the additive contribution of each term to the predictor. See Section \ref{sec:termPredictors} for an example of its use. Most of the functions listed in Figure \ref{fig:glm.lm} can be used as they would be for \Rclass{glm} or \Rclass{lm} objects, however care must be taken with \Rmethod{vcov.gnm}, as the variance-covariance matrix will depend on the parameterization of the model. In particular, standard errors calculated using the variance-covariance matrix will only be valid for parameters or contrasts that are estimable! Similarly, \Rmethod{profile.gnm} and \Rmethod{confint.gnm} are only applicable to estimable parameters. The deviance function of a generalized nonlinear model can sometimes be far from quadratic and \Rmethod{profile.gnm} attempts to detect assymetry or asymptotic behaviour in order to return a sufficient profile for a given parameter. As an example, consider the following model, described later in Section \ref{sec:Unidiff}: \begin{Scode} unidiff <- gnm(Freq ~ educ*orig + educ*dest + Mult(Exp(educ), orig:dest), constrain = "[.]educ1", family = poisson, data = yaish, subset = (dest != 7)) prof <- profile(unidiff, which = 61:65, trace = TRUE) \end{Scode} If the deviance is quadratic in a given parameter, the profile trace will be linear. We can plot the profile traces as follows: \begin{figure}[!tbph] \begin{center} \scalebox{1.1}{\includegraphics{fig-profilePlot.pdf}} \end{center} \caption{Profile traces for the multipliers of the orig:dest association} \label{fig:profilePlot} \end{figure} From these plots we can see that the deviance is approximately quadratic in \Robject{Mult(Exp(.), orig:dest).educ2}, assymetric in \Robject{Mult(Exp(.), orig:dest).educ3} and \Robject{Mult(Exp(.), orig:dest).educ4} and asymptotic in \Robject{Mult(Exp(.), orig:dest).educ5}. When the deviance is approximately quadratic in a given parameter, \Rmethod{profile.gnm} uses the same stepsize for profiling above and below the original estimate: \begin{Sinput} > diff(prof[[2]]$par.vals[, "Mult(Exp(.), orig:dest).educ2"]) \end{Sinput} \begin{Soutput} [1] 0.1053072 0.1053072 0.1053072 0.1053072 0.1053072 0.1053072 0.1053072 [8] 0.1053072 0.1053072 0.1053072 \end{Soutput} When the deviance is asymmetric, \Rmethod{profile.gnm} uses different step sizes to accommodate the skew: \begin{Sinput} > diff(prof[[4]]$par.vals[, "Mult(Exp(.), orig:dest).educ4"]) \end{Sinput} \begin{Soutput} [1] 0.2018393 0.2018393 0.2018393 0.2018393 0.2018393 0.2018393 0.2018393 [8] 0.2018393 0.2018393 0.2243673 0.2243673 0.2243673 0.2243673 0.2243673 \end{Soutput} Finally, the presence of an asymptote is recorded in the \Robject{"asymptote"} attribute of the returned profile: \begin{Sinput} > attr(prof[[5]], "asymptote") \end{Sinput} \begin{Soutput} [1] TRUE FALSE \end{Soutput} This information is used by \Rmethod{confint.gnm} to return infinite limits for confidence intervals, as appropriate: \begin{Sinput} > confint(prof, level = 0.95) \end{Sinput} \begin{Soutput} 2.5 % 97.5 % Mult(Exp(.), orig:dest).educ1 NA NA Mult(Exp(.), orig:dest).educ2 -0.5978901 0.1022447 Mult(Exp(.), orig:dest).educ3 -1.4836854 -0.2362378 Mult(Exp(.), orig:dest).educ4 -2.5792398 -0.2953420 Mult(Exp(.), orig:dest).educ5 -Inf -0.7006889 \end{Soutput} \subsection{\Rfunction{ofInterest} and \Rfunction{pickCoef}} \label{sec:ofInterest} It is quite common for a statistical model to have a large number of parameters, but for only a subset of these parameters be of interest when it comes to interpreting the model. The \Rfunarg{ofInterest} argument to \Rfunction{gnm} allows the user to specify a subset of the parameters which are of interest, so that \Rclass{gnm} methods will focus on these parameters. In particular, printed model summaries will only show the parameters of interest, whilst methods for which a subset of parameters may be selected will by default select the parameters of interest, or where this may not be appropriate, provide a \emph{Tk} dialog for selection from the parameters of interest. Parameters may be specified to the \Rfunarg{ofInterest} argument by a regular expression to match against parameter names, by a numeric vector of indices, by a character vector of names, or, if \Rcode{ofInterest = "[?]"} they can be selected through a \emph{Tk} dialog. The information regarding the parameters of interest is held in the \Robject{ofInterest} component of \Rclass{gnm} objects, which is a named vector of numeric indices, or \Robject{NULL} if all parameters are of interest. This component may be accessed or replaced using \Rfunction{ofInterest} or \Rfunction{ofInterest<-} respectively. The \Rfunction{pickCoef} function provides a simple way to obtain the indices of coefficients from any model object. It takes the model object as its first argument and has an optional \Rfunarg{regexp} argument. If a regular expression is passed to \Rfunarg{regexp}, the coefficients are selected by matching this regular expression against the coefficient names. Otherwise, coefficients may be selected via a \emph{Tk} dialog. So, returning to the example from the last section, if we had set \Robject{ofInterest} to index the education multipliers as follows \begin{Scode} ofInterest(unidiff) <- pickCoef(unidiff, "[.]educ") \end{Scode} then it would not have been necessary to specify the \Rfunarg{which} argument of \Rfunction{profile} as these parameters would have been selected by default. \subsection{\Rfunction{checkEstimable}} \label{sec:checkEstimable} The \Rfunction{checkEstimable} function can be used to check the estimability of a linear combination of parameters. For non-linear combinations the same function can be used to check estimability based on the (local) vector of partial derivatives. The \Rfunction{checkEstimable} function provides a numerical version of the sort of algebraic test described in \citet{CatcMorg97}. Consider the following model, which is described later in Section \ref{sec:Unidiff}: <>= doubleUnidiff <- gnm(Freq ~ election:vote + election:class:religion + Mult(Exp(election), religion:vote) + Mult(Exp(election), class:vote), family = poisson, data = cautres) @ The effects of the first constituent multiplier in the first multiplicative interaction are identified when the parameter for one of the levels --- say for the first level --- is constrained to zero. The parameters to be estimated are then the differences between each other level and the first. These differences can be represented by a contrast matrix as follows: <>= coefs <- names(coef(doubleUnidiff)) contrCoefs <- coefs[grep(", religion:vote", coefs)] nContr <- length(contrCoefs) contrMatrix <- matrix(0, length(coefs), nContr, dimnames = list(coefs, contrCoefs)) contr <- contr.sum(contrCoefs) # switch round to contrast with first level contr <- rbind(contr[nContr, ], contr[-nContr, ]) contrMatrix[contrCoefs, 2:nContr] <- contr contrMatrix[contrCoefs, 2:nContr] @ Then their estimability can be checked using \Rfunction{checkEstimable} <>= checkEstimable(doubleUnidiff, contrMatrix) @ which confirms that the effects for the other three levels are estimable when the parameter for the first level is set to zero. However, applying the equivalent constraint to the second constituent multiplier in the interaction is not sufficient to make the parameters in that multiplier estimable: <>= coefs <- names(coef(doubleUnidiff)) contrCoefs <- coefs[grep("[.]religion", coefs)] nContr <- length(contrCoefs) contrMatrix <- matrix(0, length(coefs), length(contrCoefs), dimnames = list(coefs, contrCoefs)) contr <- contr.sum(contrCoefs) contrMatrix[contrCoefs, 2:nContr] <- rbind(contr[nContr, ], contr[-nContr, ]) checkEstimable(doubleUnidiff, contrMatrix) @ \subsection{\Rfunction{getContrasts}, \Rfunction{se}} \label{sec:getContrasts} To investigate simple ``sum to zero'' contrasts such as those above, it is easiest to use the \Rfunction{getContrasts} function, which checks the estimability of possibly scaled contrasts and returns the parameter estimates with their standard errors. Returning to the example of the first constituent multiplier in the first multiplicative interaction term, the differences between each election and the first can be obtained as follows: <>= myContrasts <- getContrasts(doubleUnidiff, pickCoef(doubleUnidiff, ", religion:vote")) myContrasts @ %def Visualization of estimated contrasts using `quasi standard errors' \citep{Firt03,FirtMene04} is achieved by plotting the resulting object: <>= plot(myContrasts, main = "Relative strength of religion-vote association, log scale", xlab = "Election", levelNames = 1:4) @ \begin{figure}[!tbph] \begin{center} \includegraphics{gnmOverview-qvplot.pdf} \end{center} \caption{Relative strength of religion-vote association, log scale} \label{fig:qvplot} \end{figure} %Attempting to obtain the equivalent contrasts for the second %(religion-vote association) multiplier produces the %following result: %<>= %coefs.of.interest <- grep("[.]religion", names(coef(doubleUnidiff))) %getContrasts(doubleUnidiff, coefs.of.interest) %@ %def By default, \Rfunction{getContrasts} uses the first parameter of the specified set as the reference level; alternatives may be set via the \Rfunarg{ref} argument. In the above example, the simple contrasts are estimable without scaling. In certain other applications, for example row-column association models (see Section~\ref{sec:RCmodels}), the contrasts are identified only after fixing their scale. A more general family of \emph{scaled} contrasts for a set of parameters $\gamma_r, r = 1, \ldots, R$ is given by \begin{equation*} \gamma^*_r = \frac{\gamma_r - \overline{\gamma}_w}{ \sqrt{\sum_r v_r (\gamma_r - \overline{\gamma}_u)^2}} \end{equation*} where $\overline{\gamma}_w = \sum w_r \gamma_r$ is the reference level against which the contrasts are taken, $\overline{\gamma}_u = \sum u_r \gamma_r$ is a possibly different weighted mean of the parameters to be used as reference level for a set of ``scaling contrasts'', and $v_r$ is a further set of weights. Thus, for example, the choice \[ w_r= \begin{cases} 1&(r=1)\\ 0&\hbox{(otherwise)} \end{cases}, \qquad u_r=v_r=1/R \] specifies contrasts with the first level, with the coefficients scaled to have variance 1\null. This general type of scaling can be obtained by specifying the form of $\overline{\gamma}_u$ and $v_r$ via the \Rfunarg{scaleRef} and \Rfunarg{scaleWeights} arguments of \Rfunction{getContrasts}. As an example, consider the following model, described in Section~\ref{sec:RCmodels}: @ <>= mentalHealth$MHS <- C(mentalHealth$MHS, treatment) mentalHealth$SES <- C(mentalHealth$SES, treatment) RC1model <- gnm(count ~ SES + MHS + Mult(SES, MHS), family = poisson, data = mentalHealth) @ %def The effects of the constituent multipliers of the multiplicative interaction are identified when both their scale and location are constrained. A simple way to achieve this is to set the first parameter to zero and the last parameter to one: @ <>= RC1model2 <- gnm(count ~ SES + MHS + Mult(1, SES, MHS), constrain = "[.]SES[AF]", constrainTo = c(0, 1), ofInterest = "[.]SES", family = poisson, data = mentalHealth) summary(RC1model2) @ %def Note that a constant multiplier must be incorporated into the interaction term, i.e., the multiplicative term \Rcode{Mult(SES, MHS)} becomes \Rcode{Mult(1, SES, MHS)}, in order to maintain equivalence with the original model specification. The constraints specified for \Robject{RC1model2} result in the estimation of scaled contrasts with level \Rcode{A} of \Rcode{SES}, in which the scaling fixes the magnitude of the contrast between level \Rcode{F} and level \Rcode{A} to be equal to 1\null. The equivalent use of \Rfunction{getContrasts}, together with the \emph{unconstrained} fit (\Robject{RC1model}), in this case is as follows: @ <>= getContrasts(RC1model, pickCoef(RC1model, "[.]SES"), ref = "first", scaleRef = "first", scaleWeights = c(rep(0, 5), 1)) @ %def Quasi-variances and standard errors are not returned here as they can not (currently) be computed for scaled contrasts. When the scaling uses the same reference level as the contrasts, equal scale weights produce ``spherical'' contrasts, whilst unequal weights produce ``elliptical'' contrasts. Further examples are given in Sections~\ref{sec:RCmodels} and \ref{sec:GAMMI}. For more general linear combinations of parameters than contrasts, the lower-level \Rfunction{se} function (which is called internally by \Rfunction{getContrasts} and by the \Rmethod{summary} method) can be used directly. See \Rcode{help(se)} for details. \subsection{\Rfunction{residSVD}} \label{sec:residSVD} Sometimes it is useful to operate on the residuals of a model in order to create informative summaries of residual variation, or to obtain good starting values for additional parameters in a more elaborate model. The relevant arithmetical operations are weighted means of the so-called \emph{working residuals}. The \Rfunction{residSVD} function facilitates one particular residual analysis that is often useful when considering multiplicative interaction between factors as a model elaboration: in effect, \Rfunction{residSVD} provides a direct estimate of the parameters of such an interaction, by performing an appropriately weighted singular value decomposition on the working residuals. As an illustration, consider the barley data from \citet{Wedd74}. These data have the following two-way structure: <>= xtabs(y ~ site + variety, barley) @ In Section~\ref{sec:biplot} a biplot model is proposed for these data, which comprises a two-component interaction between the cross-classifying factors. In order to fit this model, we can proceed by fitting a smaller model, then use \Rfunction{residSVD} to obtain starting values for the parameters in the bilinear term: @ <>= emptyModel <- gnm(y ~ -1, family = wedderburn, data = barley) biplotStart <- residSVD(emptyModel, barley$site, barley$variety, d = 2) biplotModel <- gnm(y ~ -1 + instances(Mult(site, variety), 2), family = wedderburn, data = barley, start = biplotStart) @ %def In this instance, the use of purposive (as opposed to the default, random) starting values had little effect: the fairly large number of iterations needed in this example is caused by a rather flat (quasi-)likelihood surface near the maximum, not by poor starting values. In other situations, the use of \Rfunction{residSVD} may speed the calculations dramatically (see for example Section \ref{sec:GAMMI}), or it may be crucial to success in locating the MLE (for example see \Rcode{help(House2001)}, where the number of multiplicative parameters is in the hundreds). The \Rfunction{residSVD} result in this instance provides a crude approximation to the MLE of the enlarged model, as can be seen in Figure \ref{fig:residSVDplot}: @ <>= plot(coef(biplotModel), biplotStart, main = "Comparison of residSVD and MLE for a 2-dimensional biplot model", ylim = c(-2, 2), xlim = c(-4, 4)) abline(a = 0, b = 1, lty = 2) @ %def \begin{figure}[!tbph] \begin{center} \includegraphics{gnmOverview-residSVDplot} \end{center} \caption{Comparison of residSVD and the MLE for a 2-dimensional biplot model} \label{fig:residSVDplot} \end{figure} \section{\Rfunction{gnm} or \Rfunction{(g)nls}?} \label{sec:gnmVnls} The \Rfunction{nls} function in the \Rpackage{stats} package may be used to fit a nonlinear model via least-squares estimation. Statistically speaking, \Rfunction{gnm} is to \Rfunction{nls} as \Rfunction{glm} is to \Rfunction{lm}, in that a nonlinear least-squares model is equivalent to a generalized nonlinear model with \Rcode{family = gaussian}. A \Rfunction{nls} model assumes that the responses are distributed either with constant variance or with fixed relative variances (specified via the \Rfunarg{weights} argument). The \Rfunction{gnls} function in the \Rpackage{nlme} package extends \Rfunction{nls} to allow correlated responses. On the other hand, \Rfunction{gnm} allows for responses distributed with variances that are a specified (via the \Rfunarg{family} argument) function of the mean; as with \Rfunction{nls}, no correlation is allowed. The \Rfunction{gnm} function also differs from \Rfunction{nls}/\Rfunction{gnls} in terms of the interface. Models are specified to \Rfunction{nls} and \Rfunction{gnls} in terms of a mathematical formula or a \Rclass{selfStart} function based on such a formula, which is convenient for models that have a small number of parameters. For models that have a large number of parameters, or can not easily be represented by a mathematical formula, the symbolic model specification used by \Rfunction{gnm} may be more convenient. This would usually be the case for models involving factors, which would need to be represented by dummy variables in a \Rfunction{nls} formula. When working with artificial data, \Rfunction{gnm} has the minor advantage that it does not fail when a model is an exact fit to the data (see \Rcode{help(nls)})\null. Therefore it is not necessary with \Rfunction{gnm} to add noise to artificial data, which can be useful when testing methods. \section{Examples} \label{sec:Examples} \subsection{Row-column association models} \label{sec:RCmodels} There are several models that have been proposed for modelling the relationship between the cell means of a contingency table and the cross-classifying factors. The following examples consider the row-column association models proposed by \citet{Good79}. The examples shown use data from two-way contingency tables, but the \Rpackage{gnm} package can also be used to fit the equivalent models for higher order tables. \subsubsection{RC(1) model} The RC(1) model is a row and column association model with the interaction between row and column factors represented by one component of the multiplicative interaction. If the rows are indexed by $r$ and the columns by $c$, then the log-multiplicative form of the RC(1) model for the cell means $\mu_{rc}$ is given by \[\log \mu_{rc} = \alpha_r + \beta_c + \gamma_r\delta_c. \] We shall fit this model to the \Robject{mentalHealth} data set from \citet[][page 381]{Agre02}, which is a two-way contingency table classified by the child's mental impairment (MHS) and the parents' socioeconomic status (SES). Although both of these factors are ordered, we do not wish to use polynomial contrasts in the model, so we begin by setting the contrasts attribute of these factors to \Rcode{treatment}: <>= set.seed(1) mentalHealth$MHS <- C(mentalHealth$MHS, treatment) mentalHealth$SES <- C(mentalHealth$SES, treatment) @ The \Rclass{gnm} model is then specified as follows, using the poisson family with a log link function: <>= RC1model <- gnm(count ~ SES + MHS + Mult(SES, MHS), family = poisson, data = mentalHealth) RC1model @ %def The row scores (parameters 10 to 15) and the column scores (parameters 16 to 19) of the multiplicative interaction can be normalized as in Agresti's eqn (9.15): <>= rowProbs <- with(mentalHealth, tapply(count, SES, sum) / sum(count)) colProbs <- with(mentalHealth, tapply(count, MHS, sum) / sum(count)) rowScores <- coef(RC1model)[10:15] colScores <- coef(RC1model)[16:19] rowScores <- rowScores - sum(rowScores * rowProbs) colScores <- colScores - sum(colScores * colProbs) beta1 <- sqrt(sum(rowScores^2 * rowProbs)) beta2 <- sqrt(sum(colScores^2 * colProbs)) assoc <- list(beta = beta1 * beta2, mu = rowScores / beta1, nu = colScores / beta2) assoc @ %def Alternatively, the elliptical contrasts \Robject{mu} and \Robject{nu} can be obtained using \Rfunction{getContrasts}, with the advantage that the standard errors for the contrasts will also be computed: @ <>= mu <- getContrasts(RC1model, pickCoef(RC1model, "[.]SES"), ref = rowProbs, scaleWeights = rowProbs) nu <- getContrasts(RC1model, pickCoef(RC1model, "[.]MHS"), ref = colProbs, scaleWeights = colProbs) mu nu @ %def Since the value of \Robject{beta} is dependent upon the particular scaling used for the contrasts, it is typically not of interest to conduct inference on this parameter directly. The standard error for \Robject{beta} could be obtained, if desired, via the delta method. \subsubsection{RC(2) model} The RC(1) model can be extended to an RC($m$) model with $m$ components of the multiplicative interaction. For example, the RC(2) model is given by \[ \log \mu_{rc} = \alpha_r + \beta_c + \gamma_r\delta_c + \theta_r\phi_c. \] Extra instances of the multiplicative interaction can be specified by the \Rfunarg{multiplicity} argument of \Rfunction{Mult}, so the RC(2) model can be fitted to the \Robject{mentalHealth} data as follows <>= RC2model <- gnm(count ~ SES + MHS + instances(Mult(SES, MHS), 2), family = poisson, data = mentalHealth) RC2model @ \subsubsection{Homogeneous effects} If the row and column factors have the same levels, or perhaps some levels in common, then the row-column interaction could be modelled by a multiplicative interaction with homogeneous effects, that is \[\log \mu_{rc} = \alpha_r + \beta_c + \gamma_r\gamma_c.\] For example, the \Robject{occupationalStatus} data set from \citet{Good79} is a contingency table classified by the occupational status of fathers (origin) and their sons (destination). \citet{Good79} fits a row-column association model with homogeneous effects to these data after deleting the cells on the main diagonal. Equivalently we can account for the diagonal effects by a separate \Rfunction{Diag} term: @ <>= RChomog <- gnm(Freq ~ origin + destination + Diag(origin, destination) + MultHomog(origin, destination), family = poisson, data = occupationalStatus) RChomog @ %def To determine whether it would be better to allow for heterogeneous effects on the association of the fathers' occupational status and the sons' occupational status, we can compare this model to the RC(1) model for these data: <>= RCheterog <- gnm(Freq ~ origin + destination + Diag(origin, destination) + Mult(origin, destination), family = poisson, data = occupationalStatus) anova(RChomog, RCheterog) @ In this case there is little gain in allowing heterogeneous effects. \subsection{Diagonal reference models} \label{sec:Dref} Diagonal reference models, proposed by \citet{Sobe81, Sobe85}, are designed for contingency tables classified by factors with the same levels. The cell means are modelled as a function of the diagonal effects, i.e., the mean responses of the `diagonal' cells in which the levels of the row and column factors are the same. \subsubsection*{\Rfunction{Dref} example 1: Political consequences of social mobility} To illustrate the use of diagonal reference models we shall use the \Robject{voting} data from \citet{Clif93}. The data come from the 1987 British general election and are the percentage voting Labour in groups cross-classified by the class of the head of household (\Robject{destination}) and the class of their father (\Robject{origin}). In order to weight these percentages by the group size, we first back-transform them to the counts of those voting Labour and those not voting Labour: @ <>= set.seed(1) count <- with(voting, percentage/100 * total) yvar <- cbind(count, voting$total - count) @ %def The grouped percentages may be modelled by a basic diagonal reference model, that is, a weighted sum of the diagonal effects for the corresponding origin and destination classes. This model may be expressed as \[ \mu_{od} = \frac{e^{\delta_1}}{e^{\delta_1} + e^{\delta_2}}\gamma_o + \frac{e^{\delta_2}}{e^{\delta_1} + e^{\delta_2}}\gamma_d . \] See Section \ref{sec:Dref function} for more detail on the parameterization. The basic diagonal reference model may be fitted using \Rfunction{gnm} as follows @ <>= classMobility <- gnm(yvar ~ Dref(origin, destination), family = binomial, data = voting) classMobility @ %def and the origin and destination weights can be evaluated as below @ <>= DrefWeights(classMobility) @ %def These results are slightly different from those reported by \citet{Clif93}. The reason for this is unclear: we are confident that the above results are correct for the data as given in \citet{Clif93}, but have not been able to confirm that the data as printed in the journal were exactly as used in Clifford and Heath's analysis. \citet{Clif93} suggest that movements in and out of the salariat (class 1) should be treated differently from movements between the lower classes (classes 2 - 5), since the former has a greater effect on social status. Thus they propose the following model \begin{equation*} \mu_{od} = \begin{cases} \dfrac{e^{\delta_1}}{e^{\delta_1} + e^{\delta_2}}\gamma_o + \dfrac{e^{\delta_2}}{e^{\delta_1} + e^{\delta_2}}\gamma_d & \text{if } o = 1\\ \\ \dfrac{e^{\delta_3}}{e^{\delta_3} + e^{\delta_4}}\gamma_o + \dfrac{e^{\delta_4}}{e^{\delta_3} + e^{\delta_4}}\gamma_d & \text{if } d = 1\\ \\ \dfrac{e^{\delta_5}}{e^{\delta_5} + e^{\delta_6}}\gamma_o + \dfrac{e^{\delta_6}}{e^{\delta_5} + e^{\delta_6}}\gamma_d & \text{if } o \ne 1 \text{ and } d \ne 1 \end{cases} \end{equation*} To fit this model we define factors indicating movement in (upward) and out (downward) of the salariat @ <>= upward <- with(voting, origin != 1 & destination == 1) downward <- with(voting, origin == 1 & destination != 1) @ %def Then the diagonal reference model with separate weights for socially mobile groups can be estimated as follows @ <>= socialMobility <- gnm(yvar ~ Dref(origin, destination, delta = ~ 1 + downward + upward), family = binomial, data = voting) socialMobility @ %def The weights for those moving into the salariat, those moving out of the salariat and those in any other group, can be evaluated as below @ <>= DrefWeights(socialMobility) @ %def Again, the results differ slightly from those reported by \citet{Clif93}, but the essence of the results is the same: the origin weight is much larger for the downwardly mobile group than for the other groups. The weights for the upwardly mobile group are very similar to the base level weights, so the model may be simplified by only fitting separate weights for the downwardly mobile group: @ <>= downwardMobility <- gnm(yvar ~ Dref(origin, destination, delta = ~ 1 + downward), family = binomial, data = voting) downwardMobility DrefWeights(downwardMobility) @ %def \subsubsection*{\Rfunction{Dref} example 2: conformity to parental rules} %\SweaveInput{vanDerSlikEg.Rnw} Another application of diagonal reference models is given by \citet{Vand02}. The data from this paper are not publicly available\footnote{ We thank Frans van der Slik for his kindness in sending us the data.}, but we shall show how the models presented in the paper may be estimated using \Rfunction{gnm}. The data relate to the value parents place on their children conforming to their rules. There are two response variables: the mother's conformity score (MCFM) and the father's conformity score (FCFF). The data are cross-classified by two factors describing the education level of the mother (MOPLM) and the father (FOPLF), and there are six further covariates (AGEM, MRMM, FRMF, MWORK, MFCM and FFCF). In their baseline model for the mother's conformity score, \citet{Vand02} include five of the six covariates (leaving out the father's family conflict score, FCFF) and a diagonal reference term with constant weights based on the two education factors. This model may be expressed as \[ \mu_{rci} = \beta_1x_{1i} + \beta_2x_{2i} + \beta_3x_{3i} +\beta_4x_{4i} +\beta_5x_{5i} + \frac{e^{\delta_1}}{e^{\delta_1} + e^{\delta_2}}\gamma_r + \frac{e^{\delta_2}}{e^{\delta_1} + e^{\delta_2}}\gamma_c . \] The baseline model can be fitted as follows: \begin{Sinput} > set.seed(1) > A <- gnm(MCFM ~ -1 + AGEM + MRMM + FRMF + MWORK + MFCM + + Dref(MOPLM, FOPLF), family = gaussian, data = conformity, + verbose = FALSE) > A \end{Sinput} \begin{Soutput} Call: gnm(formula = MCFM ~ -1 + AGEM + MRMM + FRMF + MWORK + MFCM + Dref(MOPLM, FOPLF), family = gaussian, data = conformity, verbose = FALSE) Coefficients: AGEM MRMM FRMF 0.06363 -0.32425 -0.25324 MWORK MFCM Dref(MOPLM, FOPLF)delta1 -0.06430 -0.06043 -0.33731 Dref(MOPLM, FOPLF)delta2 Dref(., .).MOPLM|FOPLF1 Dref(., .).MOPLM|FOPLF2 -0.02505 4.95121 4.86329 Dref(., .).MOPLM|FOPLF3 Dref(., .).MOPLM|FOPLF4 Dref(., .).MOPLM|FOPLF5 4.86458 4.72343 4.43516 Dref(., .).MOPLM|FOPLF6 Dref(., .).MOPLM|FOPLF7 4.18873 4.43378 Deviance: 425.3389 Pearson chi-squared: 425.3389 Residual df: 576 \end{Soutput} The coefficients of the covariates are not aliased with the parameters of the diagonal reference term and thus the basic identifiability constraints that have been imposed are sufficient for these parameters to be identified. The diagonal effects do not need to be constrained as they represent contrasts with the off-diagonal cells. Therefore the only unidentified parameters in this model are the weight parameters. This is confirmed in the summary of the model: \begin{Sinput} > summary(A) \end{Sinput} \begin{Soutput} Call: gnm(formula = MCFM ~ -1 + AGEM + MRMM + FRMF + MWORK + MFCM + Dref(MOPLM, FOPLF), family = gaussian, data = conformity, verbose = FALSE) Deviance Residuals: Min 1Q Median 3Q Max -3.63688 -0.50383 0.01714 0.56753 2.25139 Coefficients: Estimate Std. Error t value Pr(>|t|) AGEM 0.06363 0.07375 0.863 0.38859 MRMM -0.32425 0.07766 -4.175 3.44e-05 FRMF -0.25324 0.07681 -3.297 0.00104 MWORK -0.06430 0.07431 -0.865 0.38727 MFCM -0.06043 0.07123 -0.848 0.39663 Dref(MOPLM, FOPLF)delta1 -0.33731 NA NA NA Dref(MOPLM, FOPLF)delta2 -0.02505 NA NA NA Dref(., .).MOPLM|FOPLF1 4.95121 0.16639 29.757 < 2e-16 Dref(., .).MOPLM|FOPLF2 4.86329 0.10436 46.602 < 2e-16 Dref(., .).MOPLM|FOPLF3 4.86458 0.12855 37.842 < 2e-16 Dref(., .).MOPLM|FOPLF4 4.72343 0.13523 34.929 < 2e-16 Dref(., .).MOPLM|FOPLF5 4.43516 0.19314 22.963 < 2e-16 Dref(., .).MOPLM|FOPLF6 4.18873 0.17142 24.435 < 2e-16 Dref(., .).MOPLM|FOPLF7 4.43378 0.16903 26.231 < 2e-16 --- (Dispersion parameter for gaussian family taken to be 0.7384355) Std. Error is NA where coefficient has been constrained or is unidentified Residual deviance: 425.34 on 576 degrees of freedom AIC: 1507.8 Number of iterations: 15 \end{Soutput} The weights have been constrained to sum to one as described in Section \ref{sec:Dref function}, so the weights themselves may be estimated as follows: \begin{Sinput} > prop.table(exp(coef(A)[6:7])) \end{Sinput} \begin{Soutput} Dref(MOPLM, FOPLF)delta1 Dref(MOPLM, FOPLF)delta2 0.4225638 0.5774362 \end{Soutput} However, in order to estimate corresponding standard errors, the parameters of one of the weights must be constrained. If no such constraints were applied when the model was fitted, \Rfunction{DrefWeights} will refit the model constraining the parameters of the first weight to zero: \begin{Sinput} > DrefWeights(A) \end{Sinput} \begin{Soutput} Refitting with parameters of first Dref weight constrained to zero $MOPLM weight se 0.4225636 0.1439829 $FOPLF weight se 0.5774364 0.1439829 \end{Soutput} giving the values reported by \citet{Vand02}. All the other coefficients of model A are the same as those reported by \citet{Vand02} except the coefficients of the mother's gender role (MRMM) and the father's gender role (FRMF). \citet{Vand02} reversed the signs of the coefficients of these factors since they were coded in the direction of liberal values, unlike the other covariates. However, simply reversing the signs of these coefficients does not give the same model, since the estimates of the diagonal effects depend on the estimates of these coefficients. For consistent interpretation of the covariate coefficients, it is better to recode the gender role factors as follows: \begin{Sinput} > MRMM2 <- as.numeric(!conformity$MRMM) > FRMF2 <- as.numeric(!conformity$FRMF) > A <- gnm(MCFM ~ -1 + AGEM + MRMM2 + FRMF2 + MWORK + MFCM + + Dref(MOPLM, FOPLF), family = gaussian, data = conformity, + verbose = FALSE) > A \end{Sinput} \begin{Soutput} Call: gnm(formula = MCFM ~ -1 + AGEM + MRMM2 + FRMF2 + MWORK + MFCM + Dref(MOPLM, FOPLF), family = gaussian, data = conformity, verbose = FALSE) Coefficients: AGEM MRMM2 FRMF2 0.06363 0.32425 0.25324 MWORK MFCM Dref(MOPLM, FOPLF)delta1 -0.06430 -0.06043 0.08440 Dref(MOPLM, FOPLF)delta2 Dref(., .).MOPLM|FOPLF1 Dref(., .).MOPLM|FOPLF2 0.39666 4.37371 4.28579 Dref(., .).MOPLM|FOPLF3 Dref(., .).MOPLM|FOPLF4 Dref(., .).MOPLM|FOPLF5 4.28708 4.14593 3.85767 Dref(., .).MOPLM|FOPLF6 Dref(., .).MOPLM|FOPLF7 3.61123 3.85629 Deviance: 425.3389 Pearson chi-squared: 425.3389 Residual df: 576 \end{Soutput} The coefficients of the covariates are now as reported by \citet{Vand02}, but the diagonal effects have been adjusted appropriately. \citet{Vand02} compare the baseline model for the mother's conformity score to several other models in which the weights in the diagonal reference term are dependent on one of the covariates. One particular model they consider incorporates an interaction of the weights with the mother's conflict score as follows: \[ \mu_{rci} = \beta_1x_{1i} + \beta_2x_{2i} + \beta_3x_{3i} +\beta_4x_{4i} +\beta_5x_{5i} + \frac{e^{\xi_{01} + \xi_{11}x_{5i}}}{e^{\xi_{01} + \xi_{11}x_{5i}} + e^{\xi_{02} + \xi_{12}x_{5i}}}\gamma_r + \frac{e^{\xi_{02} + \xi_{12}x_{5i}}}{e^{\xi_{01} + \xi_{11}x_{5i}} + e^{\xi_{02} + \xi_{12}x_{5i}}}\gamma_c. \] This model can be fitted as below, using the original coding for the gender role factors for ease of comparison to the results reported by \citet{Vand02}, \begin{Sinput} > F <- gnm(MCFM ~ -1 + AGEM + MRMM + FRMF + MWORK + MFCM + + Dref(MOPLM, FOPLF, delta = ~ 1 + MFCM), family = gaussian, + data = conformity, verbose = FALSE) > F \end{Sinput} \begin{Soutput} Call: gnm(formula = MCFM ~ -1 + AGEM + MRMM + FRMF + MWORK + MFCM + Dref(MOPLM, FOPLF, delta = ~1 + MFCM), family = gaussian, data = conformity, verbose = FALSE) Coefficients: AGEM 0.05818 MRMM -0.32701 FRMF -0.25772 MWORK -0.07847 MFCM -0.01694 Dref(MOPLM, FOPLF, delta = ~ . + MFCM).delta1(Intercept) 1.03515 Dref(MOPLM, FOPLF, delta = ~ 1 + .).delta1MFCM -1.77756 Dref(MOPLM, FOPLF, delta = ~ . + MFCM).delta2(Intercept) -0.03515 Dref(MOPLM, FOPLF, delta = ~ 1 + .).delta2MFCM 2.77756 Dref(., ., delta = ~ 1 + MFCM).MOPLM|FOPLF1 4.82476 Dref(., ., delta = ~ 1 + MFCM).MOPLM|FOPLF2 4.88066 Dref(., ., delta = ~ 1 + MFCM).MOPLM|FOPLF3 4.83969 Dref(., ., delta = ~ 1 + MFCM).MOPLM|FOPLF4 4.74850 Dref(., ., delta = ~ 1 + MFCM).MOPLM|FOPLF5 4.42020 Dref(., ., delta = ~ 1 + MFCM).MOPLM|FOPLF6 4.17957 Dref(., ., delta = ~ 1 + MFCM).MOPLM|FOPLF7 4.40819 Deviance: 420.9022 Pearson chi-squared: 420.9022 Residual df: 575 \end{Soutput} In this case there are two sets of weights, one for when the mother's conflict score is less than average (coded as zero) and one for when the score is greater than average (coded as one). These can be evaluated as follows: \begin{Sinput} > DrefWeights(F) \end{Sinput} \begin{Soutput} Refitting with parameters of first Dref weight constrained to zero $MOPLM MFCM weight se 1 1 0.02974675 0.2277711 2 0 0.74465224 0.2006916 $FOPLF MFCM weight se 1 1 0.9702532 0.2277711 2 0 0.2553478 0.2006916 \end{Soutput} giving the same weights as in Table 4 of \citet{Vand02}, though we obtain a lower standard error in the case where MFCM is equal to one. \subsection{Uniform difference (UNIDIFF) models} \label{sec:Unidiff} Uniform difference models \citep{Xie92, Erik92} use a simplified three-way interaction to provide an interpretable model of contingency tables classified by three or more variables. For example, the uniform difference model for a three-way contingency table, also known as the UNIDIFF model, is given by \[ \mu_{ijk} = \alpha_{ik} + \beta_{jk} + \exp(\delta_k)\gamma_{ij}. \] The $\gamma_{ij}$ represent a pattern of association that varies in strength over the dimension indexed by $k$, and $\exp(\delta_k)$ represents the relative strength of that association at level $k$. This model can be applied to the \Robject{yaish} data set \citep{Yais98,Yais04}, which is a contingency table cross-classified by father's social class (\Robject{orig}), son's social class (\Robject{dest}) and son's education level (\Robject{educ}). In this case, we can consider the importance of the association between the social class of father and son across the education levels. We omit the sub-table which corresponds to level 7 of \Robject{dest}, because its information content is negligible: @ <>= set.seed(1) unidiff <- gnm(Freq ~ educ*orig + educ*dest + Mult(Exp(educ), orig:dest), ofInterest = "[.]educ", family = poisson, data = yaish, subset = (dest != 7)) coef(unidiff) @ %def The \Robject{ofInterest} component has been set to index the multipliers of the association between the social class of father and son. We can contrast each multiplier to that of the lowest education level and obtain the standard errors for these parameters as follows: @ <>= getContrasts(unidiff, ofInterest(unidiff)) @ %def Four-way contingency tables may sometimes be described by a ``double UNIDIFF'' model \[ \mu_{ijkl} = \alpha_{il} + \beta_{jkl} + \exp(\delta_l)\gamma_{ij} + \exp(\phi_l)\theta_{ik}, \] where the strengths of two, two-way associations with a common variable are estimated across the levels of the fourth variable. The \Robject{cautres} data set, from \citet{Caut98}, can be used to illustrate the application of the double UNIDIFF model. This data set is classified by the variables vote, class, religion and election. Using a double UNIDIFF model, we can see how the association between class and vote, and the association between religion and vote, differ between the most recent election and the other elections: @ <>= set.seed(1) doubleUnidiff <- gnm(Freq ~ election*vote + election*class*religion + Mult(Exp(election), religion:vote) + Mult(Exp(election), class:vote), family = poisson, data = cautres) getContrasts(doubleUnidiff, rev(pickCoef(doubleUnidiff, ", class:vote"))) getContrasts(doubleUnidiff, rev(pickCoef(doubleUnidiff, ", religion:vote"))) @ %def \subsection{Generalized additive main effects and multiplicative interaction (GAMMI) models} \label{sec:GAMMI} Generalized additive main effects and multiplicative interaction models, or GAMMI models, were motivated by two-way contingency tables and comprise the row and column main effects plus one or more components of the multiplicative interaction. The singular value corresponding to each multiplicative component is often factored out, as a measure of the strength of association between the row and column scores, indicating the importance of the component, or axis. For cell means $\mu_{rc}$ a GAMMI-K model has the form \begin{equation} \label{eq:GAMMI} g(\mu_{rc}) = \alpha_r + \beta_c + \sum_{k=1}^K \sigma_k\gamma_{kr}\delta_{kc}, \end{equation} in which $g$ is a link function, $\alpha_r$ and $\beta_c$ are the row and column main effects, $\gamma_{kr}$ and $\delta_{kc}$ are the row and column scores for multiplicative component $k$ and $\sigma_k$ is the singular value for component $k$. The number of multiplicative components, $K$, is less than or equal to the rank of the matrix of residuals from the main effects. The row-column association models discussed in Section \ref{sec:RCmodels} are examples of GAMMI models, with a log link and poisson variance. Here we illustrate the use of an AMMI model, which is a GAMMI model with an identity link and a constant variance. We shall use the \Robject{wheat} data set taken from \citet{Varg01}, which gives wheat yields measured over ten years. First we scale these yields and create a new treatment factor, so that we can reproduce the analysis of \citet{Varg01}: @ <>= set.seed(1) yield.scaled <- wheat$yield * sqrt(3/1000) treatment <- interaction(wheat$tillage, wheat$summerCrop, wheat$manure, wheat$N, sep = "") @ %def Now we can fit the AMMI-1 model, to the scaled yields using the combined treatment factor and the year factor from the \Robject{wheat} dataset. We will proceed by first fitting the main effects model, then using \Rfunction{residSVD} (see Section \ref{sec:residSVD}) for the parameters of the multiplicative term: @ <>= mainEffects <- gnm(yield.scaled ~ year + treatment, family = gaussian, data = wheat) svdStart <- residSVD(mainEffects, year, treatment, 3) bilinear1 <- update(mainEffects, . ~ . + Mult(year, treatment), start = c(coef(mainEffects), svdStart[,1])) @ %def We can compare the AMMI-1 model to the main effects model, @ <>= anova(mainEffects, bilinear1, test = "F") @ %def giving the same results as in Table 1 of \citet{Varg01} (up to error caused by rounding). Thus the significance of the multiplicative interaction can be tested without applying constraints to this term. If the multiplicative interaction is significant, we may wish to apply constraints to obtain estimates of the row and column scores. We illustrate this using the \Robject{barleyHeights} data, which records the average height for 15 genotypes of barley over 9 years. For this small dataset the AMMI-1 model is easily estimated with the default settings: @ <>= set.seed(1) barleyModel <- gnm(height ~ year + genotype + Mult(year, genotype), data = barleyHeights) @ %def To obtain the parameterization of Equation \ref{eq:GAMMI} in which $\sigma_k$ is the singular value for component $k$, the row and column scores must be constrained so that the scores sum to zero and the squared scores sum to one. These contrasts can be obtained using \Robject{getContrasts}: @ <>= gamma <- getContrasts(barleyModel, pickCoef(barleyModel, "[.]y"), ref = "mean", scaleWeights = "unit") delta <- getContrasts(barleyModel, pickCoef(barleyModel, "[.]g"), ref = "mean", scaleWeights = "unit") gamma delta @ %def Confidence intervals based on the assumption of asymptotic normality can be computed as follows: @ <>= gamma[[2]][,1] + (gamma[[2]][,2]) %o% c(-1.96, 1.96) delta[[2]][,1] + (delta[[2]][,2]) %o% c(-1.96, 1.96) @ %def which broadly agree with Table 8 of Chadouef and Denis (1991), allowing for the change in sign. On the basis of such confidence intervals we can investigate simplifications of the model such as combining levels of the factors or fitting an additive model to a subset of the data. The singular value $\sigma_k$ may be obtained as follows @ <>= svd(termPredictors(barleyModel)[, "Mult(year, genotype)"])$d @ %def This parameter is of little interest in itself, given that the significance of the term as a whole can be tested using ANOVA. The SVD representation can also be obtained quite easily for AMMI and GAMMI models with interaction rank greater than 1\null. See \Rcode{example(wheat)} for an example of this in an AMMI model with rank 2\null. (The calculation of \emph{standard errors} and \emph{confidence regions} for the SVD representation with rank greater than 1 is not yet implemented, though.) \subsection{Biplot models} \label{sec:biplot} Biplots are graphical displays of two-dimensional arrays, which represent the objects that index both dimensions of the array on the same plot. Here we consider the case of a two-way table, where a biplot may be used to represent both the row and column categories simultaneously. A two-dimensional biplot is constructed from a rank-2 representation of the data. For two-way tables, the generalized bilinear model defines one such representation: \begin{equation*} g(\mu_{ij}) = \eta_{ij} = \alpha_{1i}\beta_{1j} + \alpha_{2i}\beta_{2j} \end{equation*} since we can alternatively write \begin{align*} \boldsymbol{\eta} &= \begin{pmatrix} \alpha_{11} & \alpha_{21} \\ \vdots & \vdots \\ \alpha_{1n} & \alpha_{2n} \\ \end{pmatrix} \begin{pmatrix} \beta_{11} & \dots & \beta_{1p} \\ \beta_{21} & \dots & \beta_{2p} \\ \end{pmatrix} \\ &= \boldsymbol{AB}^T \end{align*} where the columns of $A$ and $B$ are linearly independent by definition. To demonstrate how the biplot is obtained from this model, we shall use the \Robject{barley} data set which gives the percentage of leaf area affected by leaf blotch for ten varieties of barley grown at nine sites \citep{Wedd74,Gabr98}. As suggested by \citet{Wedd74} we model these data using a logit link and a variance proportional to the square of that of the binomial, implemented as the \Rfunction{wedderburn} family in \Rpackage{gnm} (see also Section \ref{sec:glms}): @ <>= set.seed(83) biplotModel <- gnm(y ~ -1 + instances(Mult(site, variety), 2), family = wedderburn, data = barley) @ %def The effect of site $i$ can be represented by the point \[ (\alpha_{1i}, \alpha_{2i}) \] in the space spanned by the linearly independent basis vectors \begin{align*} a_1 = (\alpha_{11}, \alpha_{12}, \ldots \alpha_{19})^T\\ a_2 = (\alpha_{21}, \alpha_{22}, \ldots \alpha_{29})^T\\ \end{align*} and the variety effects can be similarly represented. Thus we can represent the sites and varieties separately as follows \begin{Sinput} sites <- pickCoef(biplotModel, "[.]site") coefs <- coef(biplotModel) A <- matrix(coefs[sites], nc = 2) B <- matrix(coefs[-sites], nc = 2) par(mfrow = c(1, 2)) plot(A, pch = levels(barley$site), xlim = c(-5, 5), ylim = c(-5, 5), main = "Site Effects", xlab = "Component 1", ylab = "Component 2") plot(B, pch = levels(barley$variety), xlim = c(-5, 5), ylim = c(-5, 5), main = "Variety Effects", xlab = "Component 1", ylab = "Component 2") \end{Sinput} \begin{figure}[!tbph] \begin{center} \includegraphics[width = 6in]{fig-Effect_plots.pdf} \end{center} \caption{Plots of site and variety effects from the generalized bilinear model of the barley data.} \label{fig:Effect_plots} \end{figure} Of course the parameterization of the bilinear model is not unique and therefore the scale and rotation of the points in these plots will depend on the random seed. By rotation and reciprocal scaling of the matrices $A$ and $B$, we can obtain basis vectors with desirable properties without changing the fitted model. In particular, if we rotate the matrices $A$ and $B$ so that their columns are orthogonal, then the corresponding plots will display the euclidean distances between sites and varieties respectively. If we also scale the matrices $A$ and $B$ so that the corresponding plots have the same units, then we can combine the two plots to give a conventional biplot display. The required rotation and scaling can be performed via singular value decomposition of the fitted predictors: @ <>= barleyMatrix <- xtabs(biplotModel$predictors ~ site + variety, data = barley) barleySVD <- svd(barleyMatrix) A <- sweep(barleySVD$u, 2, sqrt(barleySVD$d), "*")[, 1:2] B <- sweep(barleySVD$v, 2, sqrt(barleySVD$d), "*")[, 1:2] rownames(A) <- levels(barley$site) rownames(B) <- levels(barley$variety) colnames(A) <- colnames(B) <- paste("Component", 1:2) A B @ %def These matrices are essentially the same as in \citet{Gabr98}. From these the biplot can be produced, for sites $A \ldots I$ and varieties $1 \dots 9, X$: @ <>= barleyCol <- c("red", "blue") plot(rbind(A, B), pch = c(levels(barley$site), levels(barley$variety)), col = rep(barleyCol, c(nlevels(barley$site), nlevels(barley$variety))), xlim = c(-4, 4), ylim = c(-4, 4), main = "Biplot for barley data", xlab = "Component 1", ylab = "Component 2") text(c(-3.5, -3.5), c(3.9, 3.6), c("sites: A-I","varieties: 1-9, X"), col = barleyCol, adj = 0) @ %def \begin{figure}[!tbph] \begin{center} \includegraphics{gnmOverview-Biplot1.pdf} \end{center} \caption{Biplot for barley data} \label{fig:Biplot1} \end{figure} The biplot gives an idea of how the sites and varieties are related to one another. It also allows us to consider whether the data can be represented by a simpler model than the generalized bilinear model. We see that the points in the biplot approximately align with the rotated axes shown in Figure \ref{fig:Biplot2}, such that the sites fall about a line parallel to the ``h-axis'' and the varieties group about two lines roughly parallel to the ``v-axis''. @ <>= plot(rbind(A, B), pch = c(levels(barley$site), levels(barley$variety)), col = rep(barleyCol, c(nlevels(barley$site), nlevels(barley$variety))), xlim = c(-4, 4), ylim = c(-4, 4), main = "Biplot for barley data", xlab = "Component 1", ylab = "Component 2") text(c(-3.5, -3.5), c(3.9, 3.6), c("sites: A-I","varieties: 1-9, X"), col = barleyCol, adj = 0) abline(a = 0, b = tan(pi/3)) abline(a = 0, b = -tan(pi/6)) abline(a = 2.6, b = tan(pi/3), lty = 2) abline(a = 4.5, b = tan(pi/3), lty = 2) abline(a = 1.3, b = -tan(pi/6), lty = 2) text(2.8, 3.9, "v-axis", font = 3) text(3.8, -2.7, "h-axis", font = 3) @ %def %abline(a = 0, b = tan(3*pi/10), lty = 4) %abline(a = 0, b = -tan(pi/5), lty = 4) \begin{figure}[!tbph] \begin{center} \includegraphics{gnmOverview-Biplot2.pdf} \end{center} \caption{Biplot for barley data, showing approximate alignment with rotated axes.} \label{fig:Biplot2} \end{figure} This suggests that the sites could be represented by points along a line, with co-ordinates \begin{equation*} (\gamma_i, \delta_0). \end{equation*} and the varieties by points on two lines perpendicular to the site line: \begin{equation*} (\nu_0 + \nu_1I(i \in \{2, 3, 6\}), \omega_j) \end{equation*} This corresponds to the following simplification of the bilinear model: \begin{align*} &\alpha_{1i}\beta_{1j} + \alpha_{2i}\beta_{2j} \\ \approx &\gamma_i(\nu_0 + \nu_1I(i \in \{2, 3, 6\})) + \delta_0\omega_j \end{align*} or equivalently \begin{equation*} \gamma_i(\nu_0 + \nu_1I(i \in \{2, 3, 6\})) + \omega_j, \end{equation*} the double additive model proposed by \citet{Gabr98}. We can fit this model as follows: @ <>= variety.binary <- factor(match(barley$variety, c(2,3,6), nomatch = 0) > 0, labels = c("rest", "2,3,6")) doubleAdditive <- gnm(y ~ variety + Mult(site, variety.binary), family = wedderburn, data = barley) @ %def Comparing the chi-squared statistics, we see that the double additive model is an adequate model for the leaf blotch incidence: @ <>= biplotModChiSq <- sum(residuals(biplotModel, type = "pearson")^2) doubleAddChiSq <- sum(residuals(doubleAdditive, type = "pearson")^2) c(doubleAddChiSq - biplotModChiSq, doubleAdditive$df.residual - biplotModel$df.residual) @ %def \subsection{Stereotype model for multinomial response} \label{sec:Stereotype} The stereotype model was proposed by \citet{Ande84} for ordered categorical data. It is a special case of the multinomial logistic model, in which the covariate coefficients are common to all categories but the scale of association is allowed to vary between categories such that \[ p_{ic} = \frac{\exp(\beta_{0c} + \gamma_c \boldsymbol{\beta}^T\boldsymbol{x}_{i})}{\sum_{k = 1}^K \exp(\beta_{0k} + \gamma_k \boldsymbol{\beta}^T\boldsymbol{x}_{i})} \] where $p_{ic}$ is the probability that the response for individual $i$ is category $c$ and $K$ is the number of categories. Like the multinomial logistic model, the stereotype model specifies a simple form for the log odds of one category against another, e.g. \begin{equation*} \log\left(\frac{p_{ic}}{p_{ik}}\right) = (\beta_{0c} - \beta_{0k}) + (\gamma_c - \gamma_k)\boldsymbol{\beta}^T\boldsymbol{x}_{i} \end{equation*} In order to model a multinomial response in the generalized nonlinear model framework, we must re-express the data as category counts $Y_i = (Y_{i1}, \ldots, Y_{iK})$ for each individual (or group). Then assuming a Poisson distribution for the counts $Y_{ic}$, the joint distribution of $Y_i$ is Multinomial$(N_i, p_{i1}, \ldots, p_{iK})$ conditional on the total count for each individual $N_i$. The expected counts are then $\mu_{ic} = N_ip_{ic}$ and the parameters of the stereotype model can be estimated through fitting the following model \begin{align*} \log \mu_{ic} &= \log(N_i) + \log(p_{ic}) \\ &= \alpha_i + \beta_{0c} + \gamma_c\sum_r \boldsymbol{\beta}_{r}\boldsymbol{x}_{ir} \\ \end{align*} where the ``nuisance'' parameters $\alpha_i$ ensure that the multinomial denominators are reproduced exactly, as required. The \Rpackage{gnm} package includes the utility function \Rfunction{expandCategorical} to re-express the categorical response as category counts. By default, individuals with common values across all covariates are grouped together, to avoid redundancy. For example, the \Robject{backPain} data set from \citet{Ande84} describes the progress of patients with back pain. The data set consists of an ordered factor quantifying the progress of each patient, and three prognostic variables. We re-express the data as follows: @ <>= set.seed(1) subset(backPain, x1 == 1 & x2 == 1 & x3 == 1) backPainLong <- expandCategorical(backPain, "pain") head(backPainLong) @ %def We can now fit the stereotype model to these data: @ <>= oneDimensional <- gnm(count ~ pain + Mult(pain, x1 + x2 + x3), eliminate = id, family = "poisson", data = backPainLong) oneDimensional @ %def specifying the \Robject{id} factor through \Rfunarg{eliminate} so that the 12 \Robject{id} effects are estimated more efficiently and are excluded from printed model summaries by default. This model is one dimensional since it involves only one function of $\mathbf{x} = (x1, x2, x3)$. We can compare this model to one with category-specific coefficients of the $x$ variables, as may be used for a qualitative categorical response: @ <>= threeDimensional <- gnm(count ~ pain + pain:(x1 + x2 + x3), eliminate = id, family = "poisson", data = backPainLong) threeDimensional @ %def This model has the maximum dimensionality of three (as determined by the number of covariates). The ungrouped multinomial log-likelihoods reported in \citet{Ande84} are given by \begin{equation*} \sum_{i,c} y_{ic}\log(p_{ic}) = \sum_{i,c} y_{ic}\log(\mu_{ic}/n_{ic}) \end{equation*} We write a simple function to compute this and the corresponding degrees of freedom, then compare the log-likelihoods of the one dimensional model and the three dimensional model: @ <>= logLikMultinom <- function(model, size){ object <- get(model) l <- sum(object$y * log(object$fitted/size)) c(nParameters = object$rank - nlevels(object$eliminate), logLikelihood = l) } size <- tapply(backPainLong$count, backPainLong$id, sum)[backPainLong$id] t(sapply(c("oneDimensional", "threeDimensional"), logLikMultinom, size)) @ %def showing that the \Robject{oneDimensional} model is adequate. To obtain estimates of the category-specific multipliers in the stereotype model, we need to constrain both the location and the scale of these parameters. The latter constraint can be imposed by fixing the slope of one of the covariates in the second multiplier to \Robject{1}, which may be achieved by specifying the covariate as an offset: @ <>= ## before constraint summary(oneDimensional) oneDimensional <- gnm(count ~ pain + Mult(pain, offset(x1) + x2 + x3), eliminate = id, family = "poisson", data = backPainLong) ## after constraint summary(oneDimensional) @ %def The location of the category-specific multipliers can constrained by setting one of the parameters to zero, either through the \Rfunarg{constrain} argument of \Rfunction{gnm} or with \Rfunction{getContrasts}: @ <>= getContrasts(oneDimensional, pickCoef(oneDimensional, "[.]pain")) @ %def giving the required estimates. \subsection{Lee-Carter model for trends in age-specific mortality} In the study and projection of population mortality rates, the model proposed by \cite{LeeCart92} forms the basis of many if not most current analyses. Here we consider the quasi-Poisson version of the model \citep{Wilm93, Alho00, BrouDenuVerm02, RensHabe03}, in which the death count $D_{ay}$ for individuals of age $a$ in year $y$ has mean $\mu_{ay}$ and variance $\phi\mu_{ay}$ (where $\phi$ is 1 for Poisson-distributed counts, and is respectively greater than or less than 1 in cases of over-dispersion or under-dispersion). In the Lee-Carter model, the expected counts follow the log-bilinear form \[ \log(\mu_{ay}/e_{ay}) = \alpha_a + \beta_a \gamma_y, \] where $e_{ay}$ is the `exposure' (number of lives at risk). This is a generalized nonlinear model with a single multiplicative term. The use of \Rpackage{gnm} to fit this model is straightforward. We will illustrate by using data downloaded on 2006-11-14 from the Human Mortality Database\footnote{Thanks to Iain Currie for helpful advice relating to this section} (HMD, made available by the University of California, Berkeley, and Max Planck Institute for Demographic Research, at \texttt{http://www.mortality.org}) on male deaths in Canada between 1921 and 2003. The data are not made available as part of \Rpackage{gnm} because of license restrictions; but they are readily available via the web simply by registering with the HMD. We assume that the data for Canadian males (both deaths and exposure-to-risk) have been downloaded from the HMD and organised into a data frame named \Robject{Canada} in \R, with columns \Robject{Year} (a factor, with levels \Rcode{1921} to \Rcode{2003}), \Robject{Age} (a factor, with levels \Rcode{20} to \Rcode{99}), \Robject{mDeaths} and \Robject{mExposure} (both quantitative). The Lee-Carter model may then be specified as \begin{Sinput} LCmodel.male <- gnm(mDeaths ~ Age + Mult(Exp(Age), Year), offset = log(mExposure), family = "quasipoisson", data = Canada) \end{Sinput} Here we have acknowledged the fact that the model only really makes sense if all of the $\beta_a$ parameters, which represent the `sensitivity' of age group $a$ to a change in the level of general mortality \citep[e.g.,][]{BrouDenuVerm02}, have the same sign. Without loss of generality we assume $\beta_a>0$ for all $a$, and we impose this constraint by using \Rcode{Exp(Age)} instead of just \Rcode{Age} in the multiplicative term. Convergence is to a fitted model with residual deviance 32419.83 on 6399 degrees of freedom --- representing clear evidence of substantial overdispersion relative to the Poisson distribution. In order to explore the lack of fit a little further, we plot the distribution of Pearson residuals in Figure \ref{fig:LCresplot}: \begin{Sinput} par(mfrow = c(2,2)) age <- as.numeric(as.character(Canada$Age)) with(Canada,{ res <- residuals(LCmodel.male, type = "pearson") plot(Age, res, xlab="Age", ylab="Pearson residual", main = "(a) Residuals by age") plot(Year, res, xlab="Year", ylab="Pearson residual", main = "(b) Residuals by year") plot(Year[(age>24) & (age<36)], res[(age>24) & (age<36)], xlab = "Year", ylab = "Pearson residual", main = "(c) Age group 25-35") plot(Year[(age>49) & (age<66)], res[(age>49) & (age<66)], xlab = "Year", ylab = "Pearson residual", main = "(d) Age group 50-65") }) \end{Sinput} %$ \begin{figure}[!tbph] \begin{center} \includegraphics[width=6in]{fig-LCall.pdf} \end{center} \caption{Canada, males: plots of residuals from the Lee-Carter model of mortality} \label{fig:LCresplot} \end{figure} Panel (a) of Figure \ref{fig:LCresplot} indicates that the overdispersion is not evenly spread through the data, but is largely concentrated in two age groups, roughly ages 25--35 and 50--65\null. Panels (c) and (d) focus on the residuals in each of these two age groups: there is a clear (and roughly cancelling) dependence on \Robject{Year}, indicating that the assumed bilinear interaction between \Robject{Age} and \Robject{Year} does not hold for the full range of ages and years considered here. A somewhat more satisfactory Lee-Carter model fit is obtained if only a subset of the data is used, namely only those males aged 45 or over: \begin{Sinput} LCmodel.maleOver45 <- gnm(mDeaths ~ Age + Mult(Exp(Age), Year), offset = log(mExposure), family = "quasipoisson", data = Canada[age>44,]) \end{Sinput} The residual deviance now is 12595.44 on 4375 degrees of freedom: still substantially overdispersed, but less severely so than before. Again we plot the distributions of Pearson residuals (Figure \ref{fig:LCresplot2}). \begin{figure}[!tbph] \begin{center} \includegraphics[width=6in]{fig-LCover45.pdf} \end{center} \caption{Canada, males over 45: plots of residuals from the Lee-Carter model of mortality} \label{fig:LCresplot2} \end{figure} Still clear departures from the assumed bilinear structure are evident, especially for age group 81--89; but they are less pronounced than in the previous model fit. The main purpose here is only to illustrate how straightforward it is to work with the Lee-Carter model using \Rfunction{gnm}, but we will take this example a little further by examining the estimated $\beta_a$ parameters from the last fitted model. We can use \Rfunction{getContrasts} to compute quasi standard errors for the logarithms of $\hat\beta_a$ --- the logarithms being the result of having used \Rcode{Exp(Age)} in the model specification --- and use these in a plot of the coefficients: \begin{Sinput} AgeContrasts <- getContrasts(LCmodel.maleOver45, 56:100) ## ages 45 to 89 only \end{Sinput} \begin{figure}[!tbph] \begin{center} \includegraphics{fig-LCqvplot.pdf} \end{center} \caption{Canada, males over 45, Lee-Carter model: relative sensitivity of different ages to change in total mortality.} \label{fig:LCqvplot} \end{figure} The plot shows that sensitivity to the general level of mortality is highest at younger ages, as expected. An \emph{unexpected} feature is the clear outlying positions occupied by the estimates for ages 51, 61, 71 and 81: for each of those ages, the estimated $\beta_a$ coefficient is substantially less than it is for the neighbouring age groups (and the error bars indicate clearly that the deviations are larger than could plausibly be due to chance variation). This is a curious finding. An explanation comes from a look back at the raw death-count data. In the years between 1921 and 1940, the death counts for ages 31, 41, 51, 61, 71 and 81 all stand out as being very substantially lower than those of neighbouring ages (Figure \ref{fig:deaths2140}: the ages concerned are highlighted in solid red). The same does \emph{not} hold for later years: after about 1940, the `1' ages fall in with the general pattern. This apparent `age heaping\footnote{Age heaping is common in mortality data: see \url{http://www.mortality.org/Public/Overview.php}}' explains our finding above regarding the $\beta_a$ coefficients: whilst all age groups have benefited from the general trend of reduced mortality, the `1' age groups appear to have benefited least because their starting point (in the 1920s and 1930s) was lower than would have been indicated by the general pattern --- hence $\hat\beta_a$ is smaller for ages $a=31$, $a=41$,\ldots, $a=81$. \begin{figure}[!tbph] \begin{center} \includegraphics{fig-deaths1921-1940.pdf} \end{center} \caption{Canada, males: Deaths 1921 to 1940 by age} \label{fig:deaths2140} \end{figure} \subsection{Exponential and sum-of-exponentials models for decay curves} A class of nonlinear functions which arise in various application contexts --- a notable one being pharmacokinetic studies -- involves one or more \emph{exponential decay} terms. For example, a simple decay model with additive error is \begin{equation} \label{eq:singleExp} y = \alpha + \exp(\beta + \gamma x) + e \end{equation} (with $\gamma<0$), while a more complex (`sum of exponentials') model might involve two decay terms: \begin{equation} \label{eq:twoExp} y = \alpha + \exp(\beta_1 + \gamma_1 x) + \exp(\beta_2+ \gamma_2 x) + e. \end{equation} Estimation and inference with such models are typically not straightforward, partly on account of multiple local maxima in the likelihood \citep[e.g.,][Ch.3]{Sebe89}. We illustrate the difficulties here, with a couple of artificial examples. These examples will make clear the value of making repeated calls to \Rfunction{gnm}, in order to use different, randomly-generated parameterizations and starting values and thus improve the chances of locating both the global maximum and all local maxima of the likelihood. \subsubsection{Example: single exponential decay term} Let us first construct some data from model (\ref{eq:singleExp}). For our illustrative purposes here, we will use \emph{noise-free} data, i.e., we fix the variance of $e$ to be zero; for the other parameters we will use $\alpha=0$, $\beta = 0$, $\gamma = -0.1$. @ <>= x <- 1:100 y <- exp(- x / 10) set.seed(1) saved.fits <- list() for (i in 1:100) saved.fits[[i]] <- gnm(y ~ Exp(1 + x), verbose = FALSE) table(zapsmall(sapply(saved.fits, deviance))) @ %def The \Robject{saved.fits} object thus contains the results of 100 calls to \Rfunction{gnm}, each using a different, randomly-generated starting value for the vector of parameters $(\alpha, \beta, \gamma)$. Out of 100 fits, 52 reproduce the data exactly, to machine accuracy. The remaining 48 fits are all identical to one another, but they are far from globally optimal, with residual sum of squares 3.61: they result from divergence of $\hat\gamma$ to $+\infty$, and correspondingly of $\hat\beta$ to $-\infty$, such that the fitted `curve' is in fact just a constant, with level equal to $\bar{y}=0.09508$. For example, the second of the 100 fits is of this kind: @ <>= saved.fits[[2]] @ %def The use of repeated calls to \Rfunction{gnm}, as here, allows the local and global maxima to be easily distinguished. \subsubsection{Example: sum of two exponentials} We can conduct a similar exercise based on the more complex model (\ref{eq:twoExp}): @ <>= x <- 1:100 y <- exp(- x / 10) + 2 * exp(- x / 50) set.seed(1) saved.fits <- list() for (i in 1:100) { saved.fits[[i]] <- suppressWarnings(gnm(y ~ Exp(1 + x, inst = 1) + Exp(1 + x, inst = 2), verbose = FALSE)) } table(round(unlist(sapply(saved.fits, deviance)), 4)) @ %def In this instance, only 27 of the 100 calls to \Rfunction{gnm} have successfully located a local maximum of the likelihood: in the remaining 73 cases the starting values generated were such that numerical problems resulted, and the fitting algorithm was abandoned (giving a \Robject{NULL} result). Among the 27 `successful' fits, it is evident that there are three distinct solutions (with respective residual sums of squares equal to 0.1589, 41.64, and essentially zero --- the last of these, the exact fit to the data, having been found 20 times out of the above 27). The two non-optimal local maxima here correspond to the best fit with a single exponential (which has residual sum of squares 0.1589) and to the fit with no dependence at all on $x$ (residual sum of squares 41.64), as we can see by comparing with: @ <>= singleExp <- gnm(y ~ Exp(1 + x), start = c(NA, NA, -0.1), verbose = FALSE) singleExp meanOnly <- gnm(y ~ 1, verbose = FALSE) meanOnly plot(x, y, main = "Two sub-optimal fits to a sum-of-exponentials curve") lines(x, fitted(singleExp)) lines(x, fitted(meanOnly), lty = "dashed") @ %def \begin{figure}[!tbph] \centering \includegraphics{gnmOverview-doubleExp2.pdf} \caption{Two sub-optimal fits to a sum-of-exponentials curve} \label{fig:doubleExp} \end{figure} In this example, it is clear that even a small amount of noise in the data would make it practically impossible to distinguish between competing models containing one and two exponential-decay terms. In summary: the default \Rfunction{gnm} setting of randomly-chosen starting values is useful for identifying multiple local maxima in the likelihood; and reasonably good starting values are needed if the global maximum is to be found. In the present example, knowing that $\gamma_1$ and $\gamma_2$ should both be small and negative, we might perhaps have tried @ <>= gnm(y ~ instances(Exp(1 + x), 2), start = c(NA, NA, -0.1, NA, -0.1), verbose = FALSE) @ %def which reliably yields the (globally optimal) perfect fit to the data. \newpage \appendix \section{User-level functions} We list here, for easy reference, all of the user-level functions in the \Rpackage{gnm} package. For full documentation see the package help pages. \begin{table}[!h] \begin{tabular*}{\textwidth}{@{}p{0.2in}p{1.3in}p{4.5in}@{}} \toprule \multicolumn{3}{l}{\textbf{Model Fitting}} \\ \midrule & \Rfunction{gnm} & fit generalized nonlinear models \\ \midrule \multicolumn{3}{l}{\textbf{Model Specification}} \\ \midrule & \Rfunction{Diag} & create factor differentiating diagonal elements \\ & \Rfunction{Symm} & create symmetric interaction of factors \\ & \Rfunction{Topo} & create `topological' interaction factors \\ & \Rfunction{Const} & specify a constant in a \Rclass{nonlin} function predictor \\ & \Rfunction{Dref} & specify a diagonal reference term in a \Rfunction{gnm} model formula \\ & \Rfunction{Mult} & specify a product of predictors in a \Rfunction{gnm} formula \\ & \Rfunction{MultHomog} & specify a multiplicative interaction with homogeneous effects in a \Rfunction{gnm} formula \\ & \Rfunction{Exp} & specify the exponential of a predictor in a \Rfunction{gnm} formula \\ % & \Rfunction{Log} & specify the natural logarithm of a predictor in a % \Rfunction{gnm} formula \\ % & \Rfunction{Logit} & specify the logit of a predictor in a % \Rfunction{gnm} formula \\ & \Rfunction{Inv} & specify the reciprocal of a predictor in a \Rfunction{gnm} formula \\ % & \Rfunction{Raise} & specify a predictor raised to a constant % power in a \Rfunction{gnm} formula \\ & \Rfunction{wedderburn} & specify the Wedderburn quasi-likelihood family \\ \midrule \multicolumn{3}{l}{\textbf{Methods and Accessor Functions}} \\ \midrule & \Rmethod{confint.gnm} & compute confidence intervals of \Rclass{gnm} parameters based on the profiled deviance \\ & \Rmethod{confint.profile.gnm} & compute confidence intervals of parameters from a \Rclass{profile.gnm} object \\ & \Rmethod{predict.gnm} & predict from a \Rclass{gnm} model \\ & \Rmethod{profile.gnm} & profile deviance for parameters in a \Rclass{gnm} model \\ & \Rmethod{plot.profile.gnm} & plot profile traces from a \Rclass{profile.gnm} object \\ & \Rmethod{summary.gnm} & summarize \Rclass{gnm} fits \\ & \Rfunction{residSVD} & multiplicative approximation of model residuals \\ & \Rfunction{exitInfo} & print numerical details of last iteration when \Rfunction{gnm} has not converged \\ & \Rfunction{ofInterest} & extract the \Robject{ofInterest} component of a \Rclass{gnm} object \\ & \Rfunction{ofInterest<-} & replace the \Robject{ofInterest} component of a \Rclass{gnm} object \\ & \Rfunction{parameters} & get model parameters from a \Rclass{gnm} object, including parameters that were constrained \\ & \Rfunction{pickCoef} & get indices of model parameters \\ & \Rfunction{getContrasts} & estimate contrasts and their standard errors for parameters in a \Rclass{gnm} model \\ & \Rfunction{checkEstimable} & check whether one or more parameter combinations in a \Rclass{gnm} model is identified \\ & \Rfunction{se} & get standard errors of linear parameter combinations in \Rclass{gnm} models \\ & \Rfunction{Dref} & estimate weights and corresponding standard errors for a diagonal reference term in a \Rclass{gnm} model \\ & \Rfunction{termPredictors} & (\emph{generic}) extract term contributions to predictor \\ \midrule \multicolumn{3}{l}{\textbf{Auxiliary Functions}} \\ \midrule & \Rfunction{asGnm} & coerce an object of class \Rclass{lm} or \Rclass{glm} to class \Rclass{gnm} \\ & \Rfunction{expandCategorical} & expand a data frame by re-expressing categorical data as counts \\ & \Rfunction{getModelFrame} & get the model frame in use by \Rfunction{gnm} \\ & \Rfunction{MPinv} & Moore-Penrose pseudoinverse of a real-valued matrix \\ & \Rfunction{qrSolve} & Minimum-length solution of a linear system\\ \end{tabular*} \end{table} \newpage \bibliography{gnm} \bibliographystyle{chicago} \end{document} gnm/inst/doc/gnmOverview.pdf0000644000176000001440000201032212515664670015624 0ustar ripleyusers%PDF-1.5 %ÐÔÅØ 222 0 obj << /Length 1671 /Filter /FlateDecode >> stream xÚí[KsÛ6¾ûWð(Í„(žK§±&í¤“Ær{Hr`(È昕¤ìº¿¾ ¤, Ží¤Ž3S^ĸ÷Ãî~»€^,~x©X@$¢Dð`± $C˜‰@ ·T°Xïg?›ÂTq–þc–ó <+Ê"K WÝe^.MVwçi1ÿ¸ø¤Š€¤˜¢V*¥ ‹;ï\›ž A|Ûäy';–s"f—öÇT—©™S1»rOVݱ97¾ž ·ç´{Vä¾¾CŠŠ´‚‚tÿÝë8¹ˆÏÌöx&D+6äqN ½BJÓ®ý+Ã(œRs.f›ªè¯ãÂ)ï(žS<»LÝåË´jλ.¨ $Ò“và‚ ªHÀãø37µ3Â1b”É‚Wd×ú´H/¡o‚guÚ\ïêêϹ¦³¸ºJ;”‹gÝÝÓ_½_ Á"‚#u±NøK‹DYmŽ‘Ö¶g·µWß[¡¾J  8\ë´thC¡êNÝØ(&"Ä<¤´§ š‰œE’Ê Ò‘ƒã°,S4õð†‚gQ‹7¶ˆ# ãl[’yH0Ƴ×ES³r¹I;x3àœ#…e?Øv€¿»Â÷j*úgœ=ƒñXËm¢[UmGrF¶#бp¦¨2H{[™8ÿ”™y¨)›¡y( ÿéðmD å&Ôƒ‰o€ s.¢Ã´?Jã3â`on¸FÖmì GÔ`d'×yî$æ¼o4!ö fA~HƒWÍvÛ\`^”ëÒ?t¤&¿3°™‹ +#°¹óÌ‹Ü„E7=þ•Y.Mõ h‚|ÞE×teCyœ§ÄrI£ Ž¯ÀO8cj8ÂOÜi¬©rˆ»Ë4iʪö i„Nfûø(Žˆ`ÉšºÃ˘ãe¿íæ-G/‹n'bjwº°ˆ½ˆk Óóiü¤IœuwV›¢%”õ._o'$ÑÿÎÈÇ£Øí*2Añ`'bXì×Òº-QuTPÄÄà[˜<&ÇÀêY49ÐGÏ®2ðqáÙ¢G„žiâÛ·JÉàV‡…Þ›²4’C…3ß|FdS(õ@I:(IÄPÄäKæãmHr€[ z…¸U7Uìͳ8p#©'Î~ýO#«÷šüAhš,ÍÓ"nÌ„æS¡Éï½´ ™}cšórYöúÄIb꺬ƫ>î*ïÍ]ÅÀ]‡^¹˜|Ê #}üVÜcñ¢\½. Ç0uã_Ž¨~ÈVuš\–ÞÒ98!§Òù—á«]´f‹ Ã|÷rFrn’‹ãºIóØnÖš–ŒŸG_˜÷X¯83M[µâì[²’x(}?óEzÄ·{»jßD ÖL'ÆüX3€:Þ­#Ĕπ»÷~€§N—'ù,¤Ï|r°O .óñ6µÉìmëwxЛ–Ççû;—w„‰ÙË‹w¤@Æ‹‡ÚÇ`igðCŠÌ·Ù.góanýØNM&mò=H ­W:8^üu`{Ʊ~&Ò, Šµ¯'ùÁû8XÂ3à ˆi\µ-s¯ÍÓmYprðûÁ »ãŸª@ -eÔ‰$ ¡D„Ò½æìÆ–tß±"¥Ý>ðÅyê(ï•ÝžVV7®b÷¨Þ¬×eÕô{¡?]*½Ç¾–y_ƒ¸óI™¤ýÕ;S›¸Jλ«ÃˆtšõubOµ*îÅVåæ̵|[•«–€’^š,+Û¿ÔçéÚI?> áýÊc!ÐðéÚn&:' ‘ªK°ö;hþœíü£ endstream endobj 2 0 obj << /Type /ObjStm /N 100 /First 806 /Length 1635 /Filter /FlateDecode >> stream xÚ­XßoÓH~÷_1Eº&Þ_öî q‚¸J! ÷ÒöÁ8›ba{#Ûzý}ãlB 4ÎI•âÎ6»óÍÌ7³³ÞJÉJ)#mÈQnH¤ø(ü±$4>ŽDNÂbhI¦)IAR™DJ’!™ç$3à`‘R<$e†¤,Ms*'Ís–´¤‘%@0Ò²y“)Ò™<%˜ÌR|Ç`@™Á2MYn1O91–rÍó”Û4É$Y!°Œ¬ÈÈæ’2G.•<›¹"‡'G)‚€ËˆÓ@2ÓP'!RX | *ŒØ ,XˆÜ©€ ‚$Àð¤“Àa ÄOæ€Ê¥‰€ B§L+¾ÕŠ§€©á¹HÓË0m˜s~Œ3o2€2÷cÀ)‘ñ<¼™V üÁ¡ ä@dŽÀÊ9}’“Æ D&D r!,ÛQœH^¨8«Y¶I´åÀ‘á¤L²"œæäÙåìfƹÏÇr)Ï#;2e-äGŽn"CRrÜÈ‘”H5(¥a7QÒ所»R)Žu£Ø:Ô¯²†*"@U‰1 €œ¥c(h9ãzEâdžf b€ãFÅÉœiAÉ*!`ÉT’ë‰T\‘Âæ(P6j-Á øƒ+Ë&aåaáP¤in“§Oi~Aó×áC ù)õ¾ªÐÎÄzö,9:k‡.,ÖãwO\,7‹_ûÖwE]ýëTW­/:jÂÂ×ýo4ןvÊ[[ï:_4Ÿj?±:;­Š*Ú]Ü6Í„ŠÚ¨|«0±RÇ•Ÿ=}ó‹…ï>­»––ESÕ·ªf£:ø®A$‹ªB×?ÌYôémh#U¬¸Ÿ)µeêEÑW%5ÅðÙãOU5-×í¸¨§°¤Õ>ûwñ"—çëzø;4áfbyôù´óˉ•‘Ǫ퇢-ý”‘»“u?„†Ú‘“!ý^ùžþ–™X‚´ð}ÙU«ÊöíÈÃËïE³ªýŸTPnªÔî¼8Fý³×¨‰: ¼Ñj„}CÈ*-«aàñª ¥_¬»ý›Bß/ŒrF«¢+ºê'´cø+_VË[¶‹´u£_‹zí'3 ìâýNÿ-yWËŨHc½c îª#î~芪PÒ÷lÖUSµÅàÎQ¬Îs?|‹~l:E‰¢îáæD‘n¶ìDŒ‰Å‘”°Dûõï‡Ñæª*¿œ„‰­·+Âò³/¿¼D 7ÅTO5[Fnü0"²ÐÿAý”VäV‹‹N&0‹ðmÙ½:º¹zÒÖý_+ä÷6Ò~ºò-·ï÷ã2Ô릥¹)«‚ç÷A÷Pv8'WGâêÉFñ=ùCOþ½˜©±C o…uO~¹Ä‚©€ï¡E·C?F™ ÏrÞæ»MÓVËÐ5´¨–[õ«£oÏNÏ^½Ú†1¥=ù‹Å¢ª¯ð;qÐXÀ Úaµªq^ów1"Áðëçççg‡š•÷ha8L%Öào¨0Ü®"W´{ kCS\ö+lg?+ô÷Ç'hS¾»7€Lô D_Üøã~ìª|^‡nGÃí´Ý?üðíÀN1{ýº9Ëcÿãë>F>]ø²¸¥rÝ}õ»Ýûî¬âfX{º!ù½äÀŸÏPøËï"÷pô7¾«øª¾Ïžoû²ïŽkÿÕ×{ºë%¿ §ôžæ¯ªášõž>Mæ8¯ów =™s'ƒÍïæ’W&ó÷¾ë›ø²3~uŽ¥âEøN—)_zoŽû3ë® !•¿Ôo–?oÛÄËñõ›­óÛ÷FfQæQÚ(ÝFº4J¥ŒREñ\ÄsÏE<ñܯª)¢”Qª(u”&Ê,Ê> stream xÚí[ß“ä´~ç¯Ø·ÌVíK²e*IÁÁÁ¥€PÇ8<3Þ]ç<ö”í¹cóק[jɲG³TBU¾ìزܒûÇ×_·½ŸÞ~ðáË‚_iV(•^ÝÞ]eŠ©B^劳Dñ«ÛÝÕ«üzÍ“$Y}þs¹?4ÕpýÓíß®Ò4aR¤W‰#2ý𥖰u*Óyv÷³"+Hã(0KV¯»k‘­Þ¯·]sÜ·×kcå0tÛºëŽöÝ®j†ëµÊ“»^gÉóOì‡ËL“®D”'Z7z0ýõ‹7 ~x ãëuž=kó?nŒäÔbfñl‡ÿ®òsvd‡/»}w_µUw¬* f… ÀL+–swo.#ˆ—1©Ý”j;‚8™§ÏÆøuöÒÆ^kÎh<_šMPòø¬.ﻶl¬Åúê®ê«v[-“†zVío1‚LfA£ö%iÿû¶¾ëú½Õö®ŽÄ‹JX.ø¯‰—À~ˆ‹ßóê³W/_.Ð#ô™ ÌLeñm­ã _†KJûð­/›ú_ÕŽˆÖnW5°w×<[¹À)ëö<r$")~5šõZZxlÆúÐÔÛòtùºa—Û‰ý¡#|ñÉ×_¿:uƒ¢à¿‹z ‚#಩.–úÍH¿ŸÂ#uã|‡™zvÔß'RÏñŠ¬ó¸UÕ‡jF´ð`,ð˶Û×SN];À-Zé?ªbU\±9)ö«ªZ¿({PïyÅŽì»Á8¡Á}µÕ¶~“$bëîíG@¨ñ°ÖÿÚÌθ©&m~þ38\ÕŽÞ=hÇýº»[Wˆ•Ó¤a‘C­ïªmùh·ÇÞ‚,ÌÉeö¿§6õDY«}YKý‰€tsÐXÝÞ7•Ëe3µ(ܨLšˆ?L\ggâZûât¡Í#q½îŽTöþ†»¨~±ìù;#Y/ºkk™²$ ô‚"³Jÿ‚2׫ïÀP™¯ÖÑæ;T¯‡Òck8Q´õ–~%ž2™*a{o4k-%Ë8­Å©•÷ªûk®WÝîhÉ–cq°/% ׂ ¥ì·Du“¸ÊÌ$“J·“ûvaËsOåö-€?<—T«Cß™g­wM8r‡]nkHµ3˜+FÈÑ÷´ŠVá)A!üVŒú€ò¢|O{AFÛµMÝV%‰$è<Ý5‡G+’ÜÉ»ÁéŪf£ÃÞ˜éHÂ`Ùê$Í×4öþ¡Þ>ØÃÑêL’¤Œ£pˆ=2 ìêízìËvÀÂÊlWðÕ¾*QŽ«z°# ªm_oì±*i¸´§6NáÜÄ) újWoG£BF¤ð >‡n_ÙIv2÷{Å… 6µMe@qk¯9ÜOKHȱ}ÛÚ†në±%x¾CÙ—û Ö D¤ÎV_9ayFyîgeJž®š`–U/î^sûà0ø÷6ݾy´¿£÷Ö$ôÄ\±B{ibÞ !”¤ÊMÁ}Ê×à½ÎÕâr &rï>S›íÕm$–2ˆ×©¶z‘©˜â:ôHÔ\_9™Ûæ¸sŠ°+€§i€„,'SP,[“!1hÊa˜ãúý²N Æs׬§dàJç춪9ësþeoéÝÕÁ­]]–™§Ûi:w«Û‡Ú¤›lµë¶Ç=$ {vÔ–f‚Za$á%›­F ¥øŒ8Òâµþ]]'¦+wöwŒÂŸH2–çéüÃW)ùÿ@2MWïëñÁ.`ƒqÚ½j1w›(T6ƒ#aº*wë E¤œå:[Ä\_ïËbB`yäŧ,zÑ‚C5ýcâmðà‘Ùü#DÙ£Ö¤^=TÍÁ‰¾¯Ð'¤²€„WŽ­øSJ Tò)§‡£Âkâ[e+³·yE»‰„kž°)®þzÆHã©i|*ÆdÔ¢XÀ6*› Ö´(NE§9ä|†màóÃlZ¡%\¼õ%†† ’²uµè¡)·nÎæqw€òƒq}ClɶŒ @Zó]ÿp­×sÙØDæbtS5€ Yã¡&ØýRM½¯[ãW6\Ç k÷]to‚Wvˆ»¤ƒ’Mªž‚ˆ˜Ë²¦“=ÏrÉ#œÃsÕÑ”¤§ÒÕ,ëŸa?áOׄ.‘4ó#çiÁ²ä'ãŠerF¥³bJ=¹}™n¤Ý”Á^(«¦rîæH˜ŽM‰å ¡6€”cÌ_êbE©MŠ‡Ý柦€²¹°tÎŒ±`ŽÌVÈ.•t,RܤPÜH~Qñiâ«Ãœa‰C9Â2ÍÎï°áQàË9ºì™/€ñAZ52¸Gr¥ >M”"ÁƒžÎ“ÃÅ¢Èåc @èQôT:¨*£aâC7!˜\òv u¤ŸœjVL5U¬NÔšeYq1[dµ¶¦DIlÑ¢g$:!r¿ö2±Bªzüc!‘1=• g Öi#–Z£=7±+ÒC‹y/ÃAƒ÷©y<ÌkI Sÿ@‡ìB¬^µ3cO\‰(45?´-ÆT98ú1ÒØÁ|‹Œ,gB]²€Ö”@fDPGL^sF‹31žâ-Ý/PÌÍÔ)BB:Ó«ÍìÛê€VçP~UýCy0Å"h$j8l 9á_Êà[š@© £]TdžÞㄦ4ÙäžnmûA"NU ÀL¼…\µ#Éœ:;n6âá’¡j86‚ÅÆ”&\üZOäû— 7Åû)ã…bM0•.8U.c,MJðÿ‰·í«-Ô^õ°Ÿ*›¡"v2+qR–"&0W…3%ô¢Öz¨Ícç‚,’szÍŽC‡cèºn4œós‰U&˜Æ/52y’XQ¶_”ÝÙó©ZìšBÓ~„ïk†÷ˆÛ“€¬Y¶Ëhß›CIų»¦*°Ð¶ïT³â@ëÚ/xöÍ`’%ó0îöЙwŠpbP/oH$OšW6Cg¨T W²Iud§ŒÁÀËTv©[(òâb·Pfú2øÎë®ã 5áªú,e¹ï!OMzãJBbä}t½N “ê+3˜NºÅ;Èõœüœ…Ú@0gö®a‰¤ ’Ï­ð¬3 ¯©…ëÈ@:*|œ÷¸z2_Á/ªVM­€ÛîÐÅ·3Õ¦ïFE*a¨HÇøèÇ Žýq ¤Í¶Óð«W¹ rÁ–7Õø¾2Ur*“»~øG”ápÁÙà³ ÉÀ^>w g>Æž'kyg®˜š¥f\zH±«ú pшÐL°Bó§¥/سñÈÆú"œÒ“l»tw´_M¦2ø^ÏÞác–}mAGjR͉ݗÙÌ·RHÞm>ŽuÒHüùOVš lê¹@å™-8ˆ¯#ü~¨UË""4…†,“zSü#J;u–†<7öL2ìB½I„ŠižIÈáþ³«p-1yùŸc A¡*òèBPä/ÝBزråh$ ‚@Î f™/ÅíL¡çd(SÁÆ“¨€ÅoRÀlëVöŽ&“³½—ƒ«æ¬Ám×ÓD ŒsÝ¡„){1}{1¸ék(«ù¸lêÙs,n\kÉuý—\Ó»›°Óœ¬\»È·˜¦o=éš}Ñ°hEÝ\~ÿåäÍ¡dçº7Ú4®©{ã\X„³ðÿe¦€™òë\ Ê#©q. _Þó IvC•|Ö!úÁvz%0„ª$З©ÇÙR‚=ÓC·ïšSß=¦…H°'P&úp’¤i}#Y)G»x톶È×ñ E1gjC9Ï.mðÁš®6ÕôòϾ—Ô5†yÖÎ02@ÕBk`  ÆIþÔ†Ã/?5ËÓÅ{Ÿîn4u–Èc5•½0çzþu+N ¦6º(‡C×ÅîꑦÙB`4´4<€M«±7¯Äðu`Ê©$ô“,NBvK§¼tŽÅ¥L&Ùeþ51½s>–‚þ¤Z´„ƒNcéÞÍ_#›/ƒsq:Ý ¢P¿‹¯‡‰& ˆKacZë y|ðùíÿÌÿ6 endstream endobj 264 0 obj << /Length 2072 /Filter /FlateDecode >> stream xÚÍYÝ㶿¿Â¸—h‘5+’"%ÚWôúP$}èm‘MÈ2×f#‹Ž(ïÖyÈßÞá§D[Þ½®@^,~ ‡Ã™ß ‡ã?=¼ùÝ_*ºªQÍ _=<®JŒ8-W%Ç(çxõ°]ý+{¯ïÖ„á¬1Ÿ<Óòpì„w˜eÿm¦õèˆÆ½ƒ:iqïZ­êµÜŠ!’¸áNíÖìEãÔVtÚ¿?ü„¬fB^£2¯W¹“ï§S£åZö[qðӷ¯IV#^á°ä~kê –ú|8ˆq8/°ã ±*P7ývߺ À2/VkŒQ͘£}g‘y˜>*£Ž²°j‡™áÚ ÈQö;8ªÑÿÙ ŽÍ¦ènM+š½ß BÒM|Ÿ³œä9/6–¨òì£hG©z7sTøñ¼#,{2\…v³ÛfôúƒÜ Í´|#Æg!l‡eƒEÄ&µ9!(Á*ÑÁ $®ñï]câ\W¹k‡!öΫ.Ÿ©.‡  g:fƒ?Þ­œ£U§~tÍ߯ýQƼõÜÞ~û]l~Œ­ïÞ^±N-0¨ç‰/–Ù®3œ Ï!|»fc]Þ´ÿ0—êÑ[Õ½Èþ¢ßû„?柺÷…›¬yIˆ[u86ƒÔª¡Ë~çálG5ŒÂ£ysvßÉ[ ›¦¾àÁßÜÒ¯m<ÇGÕuÊÌ>ë—#U[+»ïºƒÑßÌs~¹°ä×Qó^ÍAvçDG%µ~Õx6È"+Ëç ‘ôÿ,›Y t÷‘‡½_)jŠãµ?MÎ@þ…äXR™‘æ iìÿ&éä 8µéÕSc¶œƒðþ ÷—~e¿ã ûÑé nìylåNŽ©ƒ—“o㜧þB€9®|êÓ7ÝYK¿Øx´…™x’MtsïÉ…Åní·Æï}¬ywu!½`Ÿ…CÎx‘_Å듼c ø³égïx]vG“w@Ž™ôï2DäåõêNÆÄÓ4i~4†M7–x‹–Ô~&¬;‘zIÓÄ3H¤0Cª ⺅…·&a£¨¬–ÖS¿¾€¹ «ôË5á0Å9bd9VϱwåHä†#…àô\†üVüâeäÜÆ*ù­b5ÑÏ«PåHL3RPÄXþP©*­âi4dƒ}3¾Ã`x™EjS“È\ƒ²\ÄG‰ ! ìHï1|ýJU(/"Ñö#† BãèÔÛgŠ¶¾”5V®/14üY¸¾}H˜F¯"=·°4ƒ[Z`=ó–0 ï6²u]ðàéŸE–©ß¬U&ïrÉ×LGô¥/;9ž¿½ÉÌdWä°Ž@*Ø=Ás-¤ƒƒÏŸêÏw¨\g¯ žÄê2<3‚©A#§2ð$·1 ®YAòf$ˆzÅ’9ᨚ޷ê¨Ô¿öTk0¯‚sYy«*hÝt@{ãpjÇÓ`óQ˜œ©O; <˜÷J 7xœ6å6,ÜœÀâpÿ*ìlPyu%ê5ЈӺ6H½ñ"ë£hå÷ð~Ÿ´3fågÆTDXÓ¼F”¥`¹aðÇI\#ÍíóÝ=2ê<뛃oÙçEA†ÍH”Áö¨ÞËÃØ’4~a{À˜l­9 õd¾¯ üQˆ=Ï{ÙîS^cC më€ Îhßíµ-ølWç/îX©·¨VÎá3\¢=xQ‚+y–æ%g+Qƒ®»ƒ¼z'=êà©O›ÿˆv\šqœÕ dõœeh õ>6Îo˜±±BÅo™´@®Ò‹fôSfêgm¥¥æYúªtúÓac¥ƒ„…B¦£it„—JKˆÕt*~}«L7g{f'É£ûÊþxÝn³¸cWÔ¾ Sƒøé$'·•×}‡; QâÐð, Z—¼’.ƒX?3¶ ˆk”×)p,9Í'=º–õUNœ(Ðo†¡9»¦=|·äÒ“‹&dÑ*FÇo¢0Õ%­ CŒIPQ᮶TdΈCVY¬fD&nø| BÀDñ0æ÷ƒœÃEc?<{·ÀSwN Ÿ˜&§¨ vÓË3TéŠäKPÄÓùLiÄzÞ‹A,ìJᙜàƶª“ û2”ƒ‰çDÚÎhDWC«v³ÊHç¸XâW.›{[æ·e!¨ä £GÛ,21f‚3-ª¼IùYð¾p"Fg÷v¾lJӔŠ$ÚpC­ÀQªßú „ÚÄÆ|gÄÃÙ •ÖrÓ‰°ö°‘}3ñw%­D.½„¡jíp™Á‚\§nôaÞwÿCoe{^¾q„¥Â–À/ Éé]^è:º³Ù¤pÑöý«aŽBêù?¶žš“ÑïËÜî0þƒù'ãg’Ç®Öqv9Ê®965O® bÿ0ˆnŽé¢£ÔeD/\hú+Ç<ÖMÉCèv«˜’þ`ÿ”^[¯kûÜ3HΉ¯c_ðw×6§¨"ü ^×eZÄô]'=Ýסjn|›nØ$x#;9ú¥öŸ }y_䱘*<¼ù ƒÚP endstream endobj 280 0 obj << /Length 1339 /Filter /FlateDecode >> stream xÚÍÉŽÛ6ô>_atz°›©Ý]€´ñ  2=¤.P ™-Óc"²ä’=¾ôÛËMR–<3‰ôôÈÇÇ·/”~_ýp“xƒ¤ŠóÕ Œ@”úƒ8‚À‹à`¾|þ2š„(^__ëÅoežã-'zW•Æ,ζ‹Üb®!-4œ1ú‰—fC*C”½øè…L$ ÝÍß¼ÁB†¡£±™ü4i$%®0X1¼!’‘¡z‚W¾Ë]>9Ù‘œ[ ¾/½§§|"—ϹkOX^ÑW´tÉ;´~5š Hœfò>#[ ¶˜WÊbaË8EwcÍ!sBö³ßiðÊì¤ÝøQ»„ÓðB£iz÷%Yð rOùö†‘Ïõ¯:Ž–ß|VÖEÅF™UŽ­ÄÅä[Å$•AÆZÙVnF˜FC^f´ÌË{šá\bD¬IÉúô#B±^­J¦÷kš‰Õp­·œMp¬!¹‘5$OU )Fv )LÃïJl’Ô"i¤+4Jh:ÉiA0Ó{ZT„áL;K"¤ÚRtÕDãkzC\,õÂñ¶DPn(8¯7ÄPÉÎ"á7Ü ð·!Ó`‹YE³:—zÉ=¯XU5#S ¯?v­ÞâŠÑ7{6 'Ý£ê ™ ð Œ [û¨¯` -6}ì‚S¶çØ­k¨u½v„êÈ.ê1öàÑË´kûÊ?=?Ç.ìa¶ØùÏc×g¬ßÚ7â Õžâ\X¹wÄüýß³çu‡ùÚ$÷¦\Š¡ ×åÊ”‚£ré»'ª’å­d‰É°©®¢ Uöz³8Ò{¨ªšZÃFàJôjMÍÏQÒzhøgçÍ–-Îj§M‡8鿪'´;p§/„ˆ[é ·Zï‹Mg§7í{ÚÑñ›#Kn_¢¤ô ÞÐüàw[RagñÄp0Ç;ÂekÞ¼~÷׬¶õ?1ÿawb‰à[’Ñ%K÷5fµ¹ãPÞlšKÍ7áÓ¾¡„Ÿ‰„f0/·¥Šz+¡žÇX˜€& =U CámÉWòП¤ÈøÑ\Ç£–¡ºNÏ ùÊ„É’®V„‘Â<…7u^ÑmN Ó{õPy…³µ“ÃgE2ù17½á­à-yÌÔóÏ(cµÇ ñl •&ÿï@±Ë*þ¢(º*q"ž:"b—Š¤5ðQ$ü×ôÂo]uJØEƒ…‹r‡»ÌlOc·Ƨ.}ÐЋZÓ.¾ïk‘¯ œ85”XÂ7dGq‘{æjP}‘Ró mÍ,856Ý°ê ¨%õI¹PŒ;Õs5— "¤º¡ë3fÒEl™ä÷™Ô4"÷cí‚Ýç¹½ rO\(KºÎÛžé3ë ²B™=Dnf¾'œ.M/zc*ÇÁ‘næÇÓc]=²l©Íïè%æWHDðºÚ:ÒŠ÷³…@¬a(°¡BDâ®ûþñ>lîÃæ2òQÒ-]ÐH´¾ô(ÜK@#41¸¤=_ Vú]ÌE?Ü+´¾ŽåIÍň:}»'@‚§‰s‡4!*ŠQC#ß…’¡úTÐÒÔQâdû“ˆjM¸‘k~|Ò;ýó£Ø±þ• É%+Æ+‘N¾EÎJùþ—ôX/1!kV­ ’™{ö/‹Zåt¾Ô9’ ;¼Ðõ<ù\ •úãfO¤öÝ¿êëäôè洸ïð& <?ºê>ïtgB?ih€R3 @¢þS$I“âäj6¿ú¥š endstream endobj 286 0 obj << /Length 1836 /Filter /FlateDecode >> stream xÚXێܸ}÷W4°V½¶hÞImnHkÀñ~°7€¦›Ý#¬ZêèâÙÉC¾=E¥–4šq{ðV<,ž*V•úo7/Þ¼Í؆1’)Å77‡ÒDgbc4#T³ÍÍ~ó)áDnSF)MnîÜö—›¿¿ykùt“̈efCQüÞí÷®¹í› eçhML– ²‡üT”Sl2’i®½XÊhÂ4t,Ñ¥ßUÛTpš|ÜZšà)[ú“Âügª(ËŒ„–Á Ë’¢Ã•û-£IÞâ íG×vnîFÙÆ׺*né஡sryį(™ãpWW]Qõuqa÷¹®Ú¸íË–«$oŠü¶tQ™ó‰â…™P“KÆr=0ó‡R$aŠ ìDëÛ¢:záM*3†H‘o”óÊJ–ü§‡KÚ"-‹_=®,îêz«§zïJßåÉ}ÑÝádY0YÕ¯8™W{œB~`f¼hµs¸tè«]WÔÕÚ5¹áĨÑY>S®Ç›j’aƒ!6"8ºà ‚dVmÒ‰T°?Jr;‘:˜µ—#éª2¹^Õë"ôE[Q]&øfÝÉ6UVÁ +Z¤[žTu‡ué nÑ|nà ‚)˜{Â`æ:KNùÞùžýµ(Ñ_½@ÛŒž7{.¢Y­çŽi… kJ¨}ó_+WöÏy´Á÷[¥’6JÑ™çƒøl[¬@ñŒP5ÚáÞvˆ#Zy…Ö“H5E‡@¥ {*PÍ-­ˆVvåŒ@ɹ©·L%_Š½k#-]8Ù$>hxQçËOçÒ½ŒC ×bnœÇ1-r”rGWyùÐqäÀ·û¼ËãL…méò¸v[ÖÝîn¾z›7ePçÇ»|XpØúXïûáØxà¡.ËÚûÇ}ûã ­®#]ùÏàÊ ûÝw>kŒjú©CSŸ°÷ñÂ}OC÷kœjÃÎ_‚Þ!K„£&¬Å£B¤z1Ìÿ˜b{,OøGÿ‹¸E_aóžŽëâ±1%…þŸ°¹8J”¹Ü(J †CDxB˃¦Sõ`²×ðr¯`œ›¢êüŽ¶Wóÿxø˨óÛCý›ãðÍ…æQ&Ô§Q÷ ½&©ž›ƒ^,æäO>yAÖHÊVÞÒbKŒqZ$ù±q.öcÞÆ5èì… ЄT £Ê>®ƒ»úœV7Ý à=Ý·çüe„`qé€íûÝO}YæÇxæ3èüÕ`dì£7Ú,z£È -Þ»¸ecR6¥E •Ó$0±š?23×´1o C´‘sƒ·eq¼ëÊøTñåøÞó_i5>çÈ䑵ºÌÛÈ5؈µVéqMì¥zë\sú'§ØuuÓ®UqQ~ZÄÝë›]åšb‡ƒyÉ0‹.rcWŸi á—$Ñ·kò]×âI96'½}H}‹3{·«OçÂCÜÈô/<è|N^hFE5œ¬Ê ÔUraU(¤\Þl%M^£U yYüw0S”@k6OUu5• å[0ë†I’-õ)I(R’o‹²ì[à¤+¼K|ñùû);OV©°bîÙ ›šâzœ¾tÂè×5×a.LîÉ^‡9!4\¾e¤ÑÚJ¹š=i%#µ?\uÑç-ï9 •åW@3úÌí»tA©!¾Ùšé¿ô›œ”‰'@%‘ŒzÚV9ý ¨¼tù>Sס>öÓçTÕ×>Kêã$1 Û1I@(þПÚ!M`üžÙ—¼ðÓ9/J“y†Ð¼EÎDm‹ÇVÄVÆVaøW(Z$¾áFîÕìRK¬eÆJ—Ž,Í¥V¦oô'®¨£N&¶6¶ðȵo`ôZe¯Rã:ýƒ²œ*o Zñ¡#†Ž:»Óýƒ›c4”! Šº–ÊߧðůŸTf °(z3bد>ôáóB¯ÿÀÈe}ëWNþ»á'ê[wèK_œZ•Uƒ«óºÐ/ÊçmëÚðd %žï4®\Ô—~ÖÿdÚøá2ü á X^ý²Ve¿íÃ*üŠôZnøÊ¢’(Ê缎vS÷çÅÏ+a/VÕœsÂ0,ý/¬C­+/~¾yñõ9—ƒ endstream endobj 300 0 obj << /Length 3577 /Filter /FlateDecode >> stream xÚµZK“Û6¾Ï¯Ð-TeÄàÁ§S9$^çUµ›¬=[»UqªÌ¡8#®)R!)Û“ÃþöíF7HÅT3°4_ÝÐ77W_|›ËŒB%jss·‰“0Éõ&Md(¹¹Ùo~ ôv'…Á?º¶©Ûªè·;‹`¬úã°ýõæÇ/¾Íô&óD%(BlvJ…*Ièë›CµÝé(ŽEÝRétîOÝ€Õ ¦»£Êú‘°„É$6¤ˆÂ ä vß=#*j1õ9åëâž­–ÞÑh§¾ÛÊ8xSï¹CA—Bè [ÞÕ· 7Ýõűڪ8x»•0Ïþ5Ww=ÉíʆSUÖ A•ÅXw¼È¢ÝS¡ÆúH 0õÍNª$”ZAA†yÌZ2ZÐyp_µU_4õï•ùZÇn_55¾­ÇÕ¶óV`ƒÙŠp»‹Óœõ •w8뢬›z| gñ,ÝH$… W¡y¦IôÇZ—PT¶ )EçV)£h¥ÍvçoT¢e&YÄ*‰I$Ùh:¢]Œã`_ õ}k&µ´§Qp[QkÙO0íÔ³º°ç» ÇÛ®©KªnŠöþL¦oçÇAÝòr³…N’Pg©³àã¹)Íj±`l‡:Ûé’UM3P;ê«îÎm‰ÛÅ=p¿‘:(›b¼ÚÉB•ÛEÓ’|'5†µÅ¶(%Š$Zl_- מ·U?Íåñ¸L¡Hå{ÞEJ Y¨Õ]m”€ËælνÑó0ÉPÊì=À$B•¦+`‚Uk‘/ºcE%Ó@ƒ©Šioï¨~¨§†û*DØ;ªY­Á ;õÕ¾.Ç®žx8L ÔÓ><{wòï$ñÔçÚ#v/ ³¥Jÿ~nF?D«t–µSR„ô`x1éõ‡öjJ䎡}üšéQ B;ú Ž±*[mÑ+Åžæ˜ñìÌöC—Sу¶áìô¾ÉFP–“’pñßwÇîÞ3gè©ää¤æÃË3>àgþÝy`ˆÓ0‘ÙÒMÀëXz¦7ˆç¸8þº…é“-¤ùK…óDþÖWwþƒ*DäŸ-HÞ×Å}ׄ7'¸-y"Öé< þ5О¤èÿxZÍб˜ …¶ö»CUãF¤Yзë…á4 “DýÙÒ'H~L³²Ì% £T1ï %3ŸoŠ}²žÕ„÷2ŸQ†ileØCèçI2·Ç¦Ó ãRÀŒ;îú#˜ ÊÞ´©1ñç@²ƒsæÍoðÀÙ®/¢›–Ó¡&èy·Ÿ¤¬5Z)w/eªÂÆÞçH•^¡lâfæòÄ3mtTvcƒù;umÕŽµ­^±š¥oŽpªâ„¦˜©x…^ë)Fâ¦ØÃ! W~âìRag­¡z1=à(?~z0µý¹/m,¡Z$ÂX¤Ëüyîž1#K‚á`Å-Y‘º4jyëújȲ°r<ãJœHè6ZqÅÔ·"¹}}w¢î8  dˆ+~â9’À ÃL«`†]2_2ÃkÖ „LQ,—hE +«Óh¡àRÑÚ³ Kš;ÔƒÏÜ „’±ÃÍF¿GÕ3+ö: Êñ¹¯(ž› 9ŽqG+ÒNчXlÍ]D)bÁ“ªæîÅLˆ^)° š’fa,Ó•›í8Gp.š±pc¥¶ñMJ¢Ç›\ÙK¥R9T•-”aâKÂ~Æ΀ÉBRë#CyꙢ_ŒtÅ|8ÞXò°lCgïYIæ3ëõÊ'¤D€\™**Éèü5€¶^=»¹úí #±‘¥h#ÞDÈ#¹)W¿ü*6{hüqØg›·¦ëq“ÃfjÔ[³yqõÏ«o0Ù³ôßVp ØUV2‚$NÇ4”ù4íϽaÑxðïݱfª4Fƒ·ÌgUáVP{4©ôSfjWŸÔGêØ5j°éIÊW^):Ÿ¤ à^/IÂ#D¼˜GG—½ã)ªB* ÀÑkeÑ$1RµrWTWvçfOEƒYð4Qßìý°åž}µƒ `à²þÝÀ#V×,ËæM h¼<çMXì+848~ÓÎ}áãxà¦]ÕÄçíPõ¤Ž¾w‘ÔÁ÷ž[‹ ˜\ÙÓ•EÓ=†“,z2`v]s&ê•ÅM棑ž”vÈ(MŒÏ“IW˜&6ê̈,ó­†@&n‚ÞLR2^d‘í9lø¾‚…ï _{$–™aŠ|8Á ¤ƒ*–G”c_[—´ŽÝá¬è‰3¡  ÷+Yì^° |v} Ì]f¨7:@«™,i¨ÒB&¤Ïò©—>gùôOF¶Ôä&Ÿ{cÏ9í¿ï3©yê¡s¹ÌcG¹…?z!ZaâÃhºaç$WD»Œ~yfºéeàý¿øê/7 !ðh]£á-‘½.‰ØMn¨‰¶˜jÊ1åDf{ʨÁk×Útß''"L3µ´8NŠä’/x¥½à•$®φ©" ³§r.1f¾ 1¸‚#Žfãf†-¼Ü9½ˆãHb6òòÅ´9.Œ™aG ·ÜÈHk´x¿5 s-)“Ö‚›5Ý:ŠòP;4tÏ­Ö¯&pWÌM(ü‰M"ǎ¨²Ì‰¡'ؾºiÀ@A îS42‹Kœˆ‘‡ŠƒW綦H;ìkoh¯C¡µÇÙé…_ÉÓ‰ÒÑÞg$”q GÃÓüŸÙ6ȺȳO—³>ú’»}r³t¸«wër9Ísãdqž?™üï2Ì¿ó“*“4Z§‘u>§‘e s“úQ±#%§!^ Kb¼ö)t’,Õì"tÊy¥_ƒZ¼Í/LŠÞÔ`pW·@ÀÃJΡ˔€n¥I5°7…÷Ó¡/î÷jéÈ uÆåÏ„Òˆb „üñXÍ’|«¸½íÉ+L4¿z,èmÝ4ÔxËŸšûr`9Zn~ùÄW¦}w¾?tgsm§§<¢ ö¬‰f õÆGÂóäþÀ"ækÅÊ–ž.94ŽÙ’eQX˜ôÚ=¿­/çµQ;}òj¦6Ÿò’Yï ÝxµP*0ûB)˜zO_žºa ŸàaØ„»óÀÁ¢žnÐäQÑê ™îõÆÊfM×æØ {y=ÛÜ)2¤Çñl|Cz¬Š–.}”èÉgÛUÕÌ—OóUlm‰Ÿz`0´]?ÿìaAñW¢œß_¤à·ð·˜dCÿ-_÷<©<šå9 Ès@X8äYÆ¿3I,ü²4OÉ endstream endobj 312 0 obj << /Length 3285 /Filter /FlateDecode >> stream xÚÕMsë¶ñþ~…ŽöäÁ7ÀtrÈG3mg:Ó&žé!É–h›­$:$¿×C~{w(X–'“žHKìb±ß }yýîÓo*¶bŒTJñÕõíJi¢+±2šªÙêz³úþB~ùãõß>ýÆò”kb«jE=ÐßÛñ/Ý®»  tU‘JsWôŠY¢m^$ðŠIãÊípy%8¿¨¤M ™!2î»ý¶ÝÖã€Ö¨ ìö°_m·÷«Ž>ÅÅðЬÛÛ~pĵÛv]í%W?_2uÑø¹v?6}Wðs7Lsþùxß®ïÃú÷[wûalÇC³s$M? Í««Je$pÉñØ[÷ð97Ö¯ƒ/ß`¥“ I•Ó L”¸J*c&˜f=~½î6 x¼dô¢ ƒ=>«‹]7á¿ÅYØvׇïêýæ ʘ4Dió+I»o†2ÕÌ Çß¾n÷ÍDWØËÍDÞO‡zë_ï›ý‚øya¬Är•ŸÆ¶™¥a‹ȉEýúκv˺ÙKf/î;8öðìêXE˜¨ˆ¡ü)mJ™%+P¢éÑò~'@GÎ"˜A©Ì ~蓼Àâd\° {¾íüìM˜ìnþíÎÉóÊVDØ…äºÃƒ/Ö[X¨ º‚-mTÊÈú¥îJByTqâJFŒŸöS_ þ}8<¥õîaÛ¼”€„´›¦÷¿f~uÛm‡çúØîïÂÃЭÛÚë8ìºM³õ¯í8uÔì›îP÷~™«u·=ì"uÕÌÖ‹•â3¿y.À /°{f‚dN¢Â„ÊÔˆŠÏˆ¸Ž'¡aya="ÇK )r Vò 0«Uúy kEXÑåH+jXñ)¢l†ï“>K”JÐÉ":tg)ºu¢^†Žó:`»³äG»ËU€hžbûkA Р離Åu‹šTb#¸_¥àó xbbX—Œ‘<9RÅʇ%ØsÎA©âA±ÅAõ¥•`ˆ¥¼;{¹uy9UÉt¹Ï <¸â•"šÉ y!¹¾´Þm‚²ÿ@)'óÒ‰íx¿4}3<€= 6èg4ußÖ7Û0²¯ÑöEÅ7ñ¬p™’Õ YõÇK­¦_‰ýë!©'ë?û«×•¨ÒD¸Ûï D°*5Ý>¼8l7Ñ‘E¡DPTä:Ùã•ýâßúÇ'þñUöë붾C5ùö}˜ušBDßz V؃$•”žIMѳ 8/¹Ç…p Îs"T‰pˆvÖ•g¼½Q¹7_tßþå¾¾ä4Ƨ8üÞzÜ~\9€Há›\Ä1(:µîv;çöð‹& é– kó»§è ~Ý| ±d‚‰<6ëÿD_åð°îvŠâÜÞ¨©§Ð%‰á$Rßõc³‰O7½­û¦"]R>ÂXfETlפò`|žOˆ¾î›Û3s¡4SbMèl¶]du9 ²2¶çó ÓX! ‚U]Ï`ÁðušÔí1dÆ_@bÓ7ûuàCÒã>Qô»îÆÙ8üÍ*ËæWåõMÒ™:`÷ögÇz·Ô”xàL<ðé|K•ËI”‘„É…@y»*¨@)”G)Æ!'Åø‚ƒ“¹4AÈ.µ½øÂÃ$»Èt×Ô§zþ ÖèÛ¡<\PÕ›M»Hp·÷‰ߺM¸—\|K$×É £“Ü1Í÷š¬Ì÷‘ôKˆ5dÍ’°ib˜Je —¼K¶TÓëõRþ8…,o‘,=ܱ!ºÊÝñmÙK¥^åÝ·E¿-dÖ5‘"(ç›1>a0 z7CÓ{G˜ ¦3¡)ñ€`a նň^‰ÈcŸÜšÓÇ&,lN›$|ÌÐäYémß”S.Iõì_p> Yá, ¾œ¼À^B6­/›öî~ ôŒÖ§w ¡U¥ìi ­La¬>”ò¶½8Ä,9µÉ8³¹À©R²g(áRœ‘ëÙ$׃¥å̽óZˆâ«‰W1LSÕ)™ädF²kHò*•'××¾ŠeÂÁþ‡+àKV{Á_U3®\€OpI{˜Ü]:õÊ‹jÆ“Ò‡Ã.‹fÜVnê1|ìùqàH1÷V¨1æç>MjE—LÆÕ’çOž:)}."£V#â6p} náJb”B&Þ€m(£:I:FRUâßp8ñoÊù7| ^àë±öC]ëkX:æ1š>/À2ÅN›E¡#³ëuß CJ´ÎJZ¤Sàë|¤OES’X-¥+÷b°Ó,qEYžª[(1R/Š[ûE€õ"øH÷Ü4ˆaŸ†º»©ÀTb¥ÄºdúÀï+-3{¹Hú9ç©•­ÐèýeØZ3KwZÏ\`½bª7€Ï¥$È‘,„øÎq»Ð–PiÂÞýùúÝOïP芭¸e€­ZI!!ëf«õîÝ÷?ÒÕ&!Ö! ¼À£Ý­8#Œg¶]}÷îŸï¾Ä~ƒCLKûÓb ôžKÁ¦YÈâcÛ7$¡•¤æyfpgO3ƒ»8Ó9Î<:‘ ™œÀ±ª°¢÷àQÄ öFúƒ¥ R{£Ç8”¥± :Åí^ê’DZé‡X˜ËÁ˜ƒ c»«!Ï .¬!3‹xð´séÞäê‘:ÖÜ©¦6 ¡¼2f*ýë9ÝÆÁ°øa?«ø4µ¨Å]NÎPÏçZüSö„§~ëþõ{œËE³hº˜,aI̲m2kmlΉ*¥±{15lêØ a€E>ú¹1T—àÕåæXo³L57DB1Û…n¿O«ÌߘzW±4vËqÅ‚ùs¨®„fDØã¾* 5µs\\Š•!eNñ?³B.ϯŸ°ºè¸•9ËêÊ'¬®(šÈ fW gê…²€VŸ2»B€̭¥À· Å^i*ÅISY.0êüž"«2ç©nRd~c¢xÑ Ÿ)B){Yðwºá!ŸÞ!dÛ†ª_å´O‹/ˆOeAÉÅ›ˆÏ´Ôÿ¯øÈ3‰â¿¡ø<ÙÈÈäçOÅÃJb8ð$ISóTÞ×O37šD·u¡}pT¸ÅÙ‚§ÏTø¯ò’=ºà×Vë¯ïK ±¹z>e0sö<L¹’÷.ÐPÊ'¦8²`”Ìâ÷%$¸Å¹Q Î|¬Ëe`Ôö´—£æ†}¨¿#:ÿHÎCÆ~v üȶ¹¯î½ F¢ÛM– 7`0öˆ}f!Ì¢r?ì¦yhö›Œ¸Ö—¢ímlÁcÔá+¿A,J˜‚tÓêÍbØLî§ mÝ!-±¿66C¸¢1‡jÅËGí±B0Â@Š”‹ y{N–?œ–ìðôÂà^?ÏV[,ö„",ô+ÏvÓÝáý…¾Þ¶ÿ´;‰8U^ˆ× ŠG &“dz%ä«‚©¶t}ˆAˆ& áÔñ…Q¼yÖXúüPlÙ£¯Ié*ÖZ±çü³ îšøg=Q.hÑ‘Pi3GBg÷–òU,XF¨Ož@*’ûO"4¹;eѴܦ?“`Z™ã$x\0åÌ1…’°×Á…‘έžˆ.¦µh)¿x™#?“OñßžÅ~uûí²_—j ©V®âûö‘ѹ å¿5CòÌŸ`hõj†ÅaÆü®IÁ ŒŽ<ÓèüÞgdÏ;£× }ns €°F¾‰Í‰kesÄ™*òGµùÕÎè¼’£ü÷æ(c/Z²:úL«Ã^‘LoÅa¹T*¼‡'*Ò†¡­ õ½ÃÖ]+±iƒ ç‹·¡-–ùs‰X<Üî|g¡bÈÜ“ÿ¥˜3RHòÒôJâ¿:O]h2Úp½¼vW8`|nƒÚpa’*ø•ázb~\jErŠ…G]ÈòÒéÜß_·¾tXûf<ô¾ájí|-#ë‹A0Ξ¹¤w%sUøÛ`È€np¯±áʪyvÓö¿h†qûØ°ú¢% 0øÿTƒYAxeÏHí³  &Zdù_?m ˆoaq~¬ÁK5Jf¨?+ˆ-ô“wkæë,IGÔ×@]¹»Ÿ)v/íÿ²ñó“ÄZxBÃØ™×:ä È §÷ß H½Óý:Ü„ÿsFèjTÄ.‹óQUX¥.võG|Ñ®º„ó¥AüuÛw;?_ûwmœ¦CÃçþꎎÅgà´³žð¿2fŃŠ~ôÿ ,P=UZ)Þ¼:ŒÍ0ãu/ë®÷·y7síÄ`'tÑCû±©ûh7}ßMí´Ãp|2T\ÚÝ€†ÖÇž( ¥…ýÙàòÌÞü0ù  endstream endobj 234 0 obj << /Type /ObjStm /N 100 /First 899 /Length 2563 /Filter /FlateDecode >> stream xÚÅ[ïo¹ý®¿‚Û/ gH 8 —4wz@p Ðkƒ|Hlõ`ÜÁ lHÿû¾·¢cɱ´{¥1¸»â’og†3ó†LlÉ[v1W´â’)Úâ$´êJ´æ”}Zu5ÚæZ.¼SÆEt1[¤0^KÝKæOx_Ípkâ[à |Š1SÀ_Êü O³—"þ ï"žªÔEŠ Ø*FŽ´¥Â7\…¯ŠË3§Ôp!èƒirl#¹œ‰0¢BÂweËY]®Æ‘ÍIHx=V|¶rŠæ$B ND )seubÉÓIeg®ŽŒAKÌ%Á•TÓ"ItEúH‚ ù]ø-:‹¸R CŠ+Í xàÖyi§‚ï–ê4sRè@ ?Ú¿£'D¶@ª-TÂwÔ@ѱ"i΂ÊÒ•“ /Ëœ 01ü¤ø Ã-’Fg†9^°Z©g 2MšpaÏš«1ÓJ‚«‰z¶èªdê&¹ZѯŸˆªU00f© v•`3-n‰‹Â'Å5Jf×råÈæZÜ!@×´@Ò5»f0¿TñÓ€»Bdá> A0j³ò úÄ3˜i(AjZPü ÂU¥nhØ¡)­ £DŒƒ«2¶,ø1f|W¦iÇsNX1jæ3jÁð}™Ë|³*MãÂ&Ž«B›ä:ÁøãÂpÛâÙ³Åòí?­Üòùååúf±|óùãÍpÿ‹Ë?Ë×Wç««wëôýòçåß—/ÞEÜ„÷‹å¯«³÷N›§Ñ‹_­xØ¥”â¥%ôzîž=sË7nùÓúíÚ-_º¿\þxW/Ö—>ûøW÷à ü›Iß(´X›XA®`IšI‚½Á¬%ge7%Ï%Š̓P2–8`yeŠÌ%@ ôË5x ¤Ç2¥Ì¥6(Ë¥o\h¼¶„àµÙ^(;hæ4Ý[<†¶0”$Ÿá|2Ÿç:OšVo Ë ¢d6ÜNE3£ ×@+.Kõm“ ø†àž îãmu$ó›oÎ÷Òóm4XM¾¶Ñ•$Gðt9©g¬¨âÚɸË:äv¡Æ4…lš‚ãk“,hŽ±ŽBñé_2ÏXG|!¶‰xæˆí> !‰°Ø܈÷45@:ã#“ëM1ïk¸qŠÌ¾œ‘èy†é–üVBð>_×a”ùåQ¨ h&AÈŠ‘lúmLå«ù:‚Ñƽ iF‰>é(G°×œ|@ü¡•Tæä°Žê8”#lÄ’Fì‰PL!-I Ì89’àIö"4òƈÅ\ã8”2ûÒ‰µúJ(ÐS"sF\Þ—Auº â¥{Ák‚ûÕ-û׿Ý0lÿ‚Ú1ìåç?ÿ|¿·¯ö¾qrßÒ¸¨ÚnßWëË›í«JG]6o½ª¤ÖÚoéµÝþ›Úo@¥R·¿€3åx{“h¾Ã fZ¾¾ZŸ½YAtnùúå+·|»úrãÞïjãõ‡ßW‹å Z]Þ\cá Ÿ@™_¯?_­®r;<úeu~ñáÇõ7h©`©’;‚g"€ü_¸ÂÐGô|ÉY $6­öÖz[{Û6mÚ|©ÿ¦íýsïŸ{ÿÜûKï/±·©·¹·}~éãIOúx²ïý¬vjˆwÔ ¸.R~Ðxä“ÑÒACµÙ®1oÜz|¡é«Å2ºp퉉!G+²I” ,ÛD}<@“wÐ50Ò;<}€,ïâ™?‘4ÐÔTéÑ Ä£îM&â™?ú¸jNmp¸™…©ˆûqÚlGˆÉ 'ja›s@c«ãPæ—ŠÖˆ %²€éYüÒÙ7‡2PV ^À˜£@pû:ê8”ùSI-Í … „XO¬Šz±6Eç‡WX(…¢ ’'….iÉü¾WÁ–‹m˜Â):JÝ·(ز²\«25¾XÃ4_Wà{•"á‚ÀçÂn¾XE&âI³'•>· õÚ`LŸÎýT¾á~šžÈý¤s´Ò9ZéÜ­ôçÚŸk|»·ZRÆ*1!¾´ÂzRñ ¤Ai5ÙþÕòü÷«UH3.[$F‘›¸·@•k§")úÆe¤ÐÕÌ!QDVe,ÔB8ùíbÕæēႯ8àÔ+Yá‰q¤’qYÑâöWr™9‚íäˆ I÷ãøÛÕÅuF 92{ß­[fõß„J§—¶vûn¼Ö)¼»M+WíwSäjªÖ«îÜÔ=¶ëÛ¶¼ãcÝY ߸3«OtgÖÝ•õR“õR“õR“õR“•¹Ýµ¢ˆ¹9r¿`ØüR„\$K§³Uˆt(ßâÈàá`Ÿ§2P*-w@X´2³ɹ óÅQ̳\yjLBªòÌSóÃ,yè© úX˧ƒa¤/åFãN[~º'kÓ\Ù–{Ú)†oûªmôh¢ßø“*Oõ'½m½]»?é; 3ù5¸è,C-Ü3‰T îéìC:‰‰)yÙ¢É=á“ÉȆ ùd> stream xÚµZK“㶾ûWLùMyãA¤S>Ø›8µ);ImƧlŠša™"U$µãÉ!¿=Ýh)hFãr."èF£_7øýÝ_ÿPˆ!X‘¦òæn“f,+ÔÎ㙸¹ÛÝük£Xrû~ýC.é2cyQÜpšÔtãTvU=Ú©ê¦`E&3œ¹µS·"gYNó:µSslëÛ­”*xºÉ¦ßÓpI¶éêr öTšôÔ´-ÝÛUʶ)ÇzG§fz¤‰uY=ÒX?=ÖÃmÂ7ï°_lîo%ßœ&»òc3Znì³ë'Z¡«µ±šöÙ͵$+ H­}?ØÉ}gI€x¶$3Ú6²?²Ûm*Ò͇nW×;`E©W`%%òMIbSÕÃT6NÏG6Q<øÿ¡ßÕ- =¯ïvM÷@íýi0[6Ó!Ï«ä(düwf›È9c«—óÊé4”mlsO·‚oJQÊ7SOÏú¶úëTw;;nD Ã=ÃEùæœdèß( ïhä/}¿;”ÝnÓt3ÒÐÇ÷Á*vZÓMC¿;UõÎõé9ÖÕÔô¶£˜0¼ 1™æ°Í„¤ ´ý©¯Pi ¿"a ŒÂjú'.3«äùMÆ ­rc9L)XŒ%EBóp7  4µÈ‚õ &aý`ê·1ª<™‰ %Zp½¢zN/eiž…ô¾ŠÑËœãL.‰’C§’«¢ärž¿•\%'Vä†ØJŠ)-Bz°\ì„–­Ž¨Š/—ÂÿÁ¬?F]šæ,MÄB}«›ÖDÚçµ8Åt¶ámáãûO<å~ĹZ[O­Ò¾f ­ßVýáØwu7yõ¯‡’TÝhx&Y’ibQ¿§Š§×«xúŠŠ«+T<_Q«ø‚Þï âéË*^¼•Ü%mî\ÃeHî —ÿ楌’SZý&îU\½]ĸ?·ÏÅQ‡Sy±°Ï÷§ak1v™%Zj¸O4GÇÖ¼”-Mr' “Š2gÄl§íO1Ä‘VöoàFD–Ò9pxÃËø¡;ÄVå,I=qò$U}œ"Ëɼܟ†zYO@3ñ“K@ŸqÕÚÊÄ1ÊC‰CÌ”yˆôbœ*0Tá&ŒÈðp:wBáö^êý­L,x™Câú±®šOœËª´ÁšN=ÙV9“b‰63‚4ÎÏ‚‹(þy‹¥fÕŸZëšïítÏŽóÙ¥uîûÞm`Œˆ QÆx®­=ÔãÀÌþ/=>Òã+z¼_ô#ßÙíÓÜ´¾¥‡ Mü–Wiÿ秷âûÎI­‚“HœŠtM ¶b° ‘à`É"½&3H ƒœÒÌ’¥ŠŸ0uÑ1‘ò‚ ¸^>I/{ÜÌ9û[‘¦ ½•HöNÒ³Šgrv6xïÞ~J$˜d€PJê0O¬W8–„/tîŸcÇžpüËþ¨¦ÊÏ1¹ H°'BÓ#Š LŽ¡EN‘¸Ô¦u9;‰1|Žõ±`ØNq©WfƒçK%˜†¤ó E_š$zí}Éy$F”éÊ{¸<Å©Éitúà­Eû¼ÒøóIa§šq)^÷™iÇ|&³q{‘œCþÌ@}²ÌíâáíVpÎ!¾SoO”8|¢¡VI,ˆ-I(RsÜŸ£˜›š ˜¹Ì—Y=ðéq󗺫1Q4BÚÕc54ÇÀc¬ÕÑ£þ<ÖZy.í1AãihPq° Ê× Ô$wÑE¢¶àhrrMØ–:…m¤ÔÓÓ¸½å$H”ÍTcËãcc* 9Ö:7{¢º€º›öÙ† »¯ƒõ½‹‚Óˆ)¼Ñ^¡´Ák¤½ÖÅ‘ŠHY&Ä5[."VIj2f|Îë‡#Cò` ·›s,«_ʇ­¼H7ßÑâ³!©„ÀŽVm9Ž±MIÉòL^µ«ŒÒq塆tÌn¬¤GÛ`\Ä–#Ÿ†©è ‹V`vyÂòT-=‰I»:§è/ÄÅ T\ªåFîŒöDü™fZ{?l1Ê31•ã…–üSø§WB¡«Æx_Žõø/÷…æŽÚ•åËÀtê]SMýó©dzöÉ1”¶7Wñh´pG£)ÂãÈø|¸ïÛ¦¢žvG«·0hNJsk%­®§ÿÛz?ÑÈc‰)³|³«é_k¿88ÔHÁ€Wì¢ñ{ xß>Û ’±\Ë¥&xÏ0ƒÌ@@Ž:ô¬Ú9å0-1­è\Xô4&tbÜ&çÙ\[ŒX$dXo9Y¨0Úʱ­ò,>V¸B(4–ú ˆ‰2î‹ÏŠócy_·Qåü˜ù©xP‘õÀÍå±Ïåpqµ½–! O¼«6œÛ-´´‚i?˜àeŒ,ä@%1G/ûÃmÂÁì‹•¦káÚ)h,LBÍÁ>é+4>£á—CSÞ·µýÕs¬íJ° ò g*)æ—ɺ%é±CX6Ý".ð?M…. ÿ_ZÙÄl]Ý&X¨cÅdSô–›¿w˜~ãœ0ôä¬êº†ø€w¢"—ä0ôeWùʱiïØr·œ^G›Ü.ª½—Æ°¦äæÃ>¾¤ç«+æžD¦Ó 15´/†˜Þeï¥wËÈدÞ{O kNµíz‡€®yX.ä]tì\§±n÷Ë@ë_C€×ÖÛÕÛ~éäÀÁI²Ü×ã¿f€Ñur É8€9®ÀOà7‘1*8™¾Ñ_](‡ÚF½8ÑîÒd©ÎŸ½/<}àûSñ&bΰWù tà ÇþV®9zðs¡½vä³rÕב*.° ™KÔ—³øö”ß^Õø»·1SJŠ" •)¦L½?€U8ô]DÐJ2™&ot³Mê: °ÃNÓíHÌÔjo®B·FŸKî°ð‘‚„ŠëîNFMh·dS¿0çþÙœ(™R›es—£q€ÐûCßöýitÚu~¼°S.òÈÝÈb€ìæâpå ßû¿u)¬§æ€[·ÑAÁÕ…Ú_Ä¥xl±"ԛ̾lcV+8„È Ž(T JØךkå³—õ~ÿµ@ºXáo,œpdW@³ë8º…&꺕¹Tãk.¡OÊ„òHÁ”sóMÂ6§ò½üíçŒgô‰.¢1µÁ5=æélÃ8Ø0þ·ag2»Ö†•ö‚±Uú¸!÷ ­³÷<¹ý(ã,ê—dèí³ÿD€“[›õ´d‰U‚ÅÇ')5µþS½»Ú¡¢È‘%&Aõ¯#.Jâ¢ÆoiÞtdïhé5§öûCÜÞÑÜ×Æt‚mØ›šä<ÕsMRÙ ®ÄO+˜¯€€ä.]@ÙµBþ-M5"0ÀqÂL:b€ÓüMnE©,,SAÏ)4ƒ»fº&J#~(P`ß2àéÝ‹}³ê¼Vè;ûÁ®ÝÛ§ÉFÔ2È Êäà9“3<Ð1pH6–‡ò"3eï:ÉÊ«Áöd0Hæ>fÊ0¿ï ܆æ̽ÑO,G5nú\,ÈìçSø|èú´:û…ù¹‚ù„'7ãuÓ¹;ÉÞgMüBÅeU.¹ÙâÙ®‚Î?Bà©ÈÜÞ[àجùÜj>7ECì.Ñ.O×—œü‚ÓR25go—tàV¢Ö9+.j¬è?43²îè/’&òGCCmѵK0)=¿`Àkrn¿Š ¤ú“‚–K·}•¶å’ÖM£Šµi‚¿S»~ÚŽÓs[C'¡aš³ÐéœRê­Ò6õöS¨I¨¥ûSQÕÜ}›q£óÿX`çr0À&W€Ò§f:áÉ„z“-îŽé±~¸§§]Lµ´eìÂlÞ’»ÙûÎÞæ}á6/MÁGÍPÔ|¼C§ ”îc7_ ò}îÒ—Ž[Ü2:²Ë«`ÃÃÙ>ÏÇ‘/£Í%Ã*3ˆîÃDŒ³ÔÜf÷õ’RIBE/Àr­oÖP{ÖP4–*ç©¢-²ºÿüG„ÝxêW“tOí 4m»3¯sh÷bF­r~ ‚ôyñ1H‘õŠ+ÈÍb—JDÊÑ´ã©¡«Òœ6lžŽ1,7Õ©5¥´œ\°Èüõ*ŽS4ÇÖi<Ñ·gØiºª=íìZ¾Nà¢E–­®E}N[ºk0Ew$ø\¸E·ER9ŸŽ±’O‡–/Ú.’¡ñó*/˜¯c’cÚ+¤q*קŠÏ‚§u­ËVºd •\à B–Jí5#òÝâe7¶Vßƨ ßÆl…Ö€hÒõÐï{—ê~#ÚŽÿ4&Èà_üH†éË×Ägû7¢Ä† ÖJs{}›ã__üùî‹ÿÙæ. endstream endobj 328 0 obj << /Length 2208 /Filter /FlateDecode >> stream xÚÝYKoä6¾ûW>u#n_z ²‡ ìc ‡Éèí¢–:-iìÉbþûY¤DÉlXì!‡FS|T‹_½Èïn¯ÞüP²¤$eƳäö>I3’•"É3FhÆ’Û*y¿$%|»c”ÒÍ÷OêxjôÛíN0ºQÛOé¦éê~¨÷øu?¶û¡îÚí‡ÛŸÞüPˆ€:Mv¬ YtÿÙãÕâ¿Þ²tƒôo°gßµ}]é3~=žë¡nÜ"G¾È3Q™3`cé·]ÛÔ11xNÊ4õÓ&yQúÎq:¶»AŸØªê~?ö½®ð³vao ÏD¦„æv\ZL[þt4ƒWßß^ýye$¥ K¸ Dò4Éa*£e²?^½ÿ@“ J(IÓ"y´SIY!2h6É»«_¯¾3¸Ü›§EK’3‰l*¡Ì‚‰9a夃o"Z*Ȭ#<šÓï4¥æ÷t¬+\ÁCýK@JVú5¿^"‡Ú…Ÿóá+‰LÅL$uÂ3‘“l¢ò(QNTú½j.Qâ”Pš&)„ÓÖÛ¹Ï%á'OI)8Îüm[ØîxVlöƦáAû|å^*†õ 2’æ“o"dÀé´·ù,–T8t:?ÕV1{É Me¨¤!0ØœOªV=n­'uVG VaòMwï'œµ›Òk3ið κª÷Cwî‘ç|£Õþ€CõàZÊ­so4®[`²×§Á׎¥‚È"ƒ³Æl¥3ÆIÐ5ÝFç^÷'½¯ï?ÍÖ{¥á^iFx99YÔˆ^R0o1“:oY±yº‡. ¥„|é nA背àBÅ?'úc$/¬\<“z—ò¬Ãþ5àÄŒíZÔ…ìÿ-b‹Ù–ž¹÷@[ÄHæ}&´TÓwتôI·UÖùÂ~ÅÇ-°QçZÝ5:\P¾,ÒW{‹4@¡F¨e›Ç-`¤› ?[mݸÁÉwç$±¨WƵ›z÷ÉmÊË:ö`ëiº1ÀK‹Í»ñtêzí7T»-úÿž2-ŠA ‘10æFxA8—_sàj³çQÎÐëìƇ~(׹İéiÁ´«˜€þÓ2{Iÿ)Éfð‚j2) Ú—ö%KRHþº}±øÆ‚s5Ÿ“Íã¦56é ÂM:3«2ø0w&2úèð³U#9-¢Î BÞÍ^ËL=f´ýx)È0Ú†_c¯³Á9B‰ äÒ\èFäË ¸fS”šYÇ Ó‰ôÙ±BÒÊïzÐ\äÒ´íÐÊÄ„31é×41Û‚8rÀ¶Æû"ñK”“RæsžÁZ9‹G@)ƒ„„Ö÷hºRÀA®cP=„Ù,ì¿7'Eì &§\ÎÿWÔ€'óªLvNsÂLVnŽ£Õ=´œ×…VÅ¥™Cà]¬ïŽ0†l¨Ì/E?y1úñ´|]ôKs’ç¡œƒ±ndljˆ\í€áÄx¬›['Õ#‚HÀ{±uêÐ=h­g% DÔ¾¡B0:rËCmøÑ,Ón~£îtÓ_ZFkä\­f„˜îßFÔ lÉ™³{[èI¼† ~ ÏŸ­T.ƒóÁoëTl×°PLdl©)«É%™÷샣týÆŒ8ßàŸž ‡k7-Xʧ¥»k›ìZ𛄠¸9ßDˆ‰Å ó»ö~ˆF6ó9–ëC†Ÿç/øQŒèqjË©;pVr¡ûjõ¤ÎέAÐ1‡¶Òã8æÌ"^ÓàcòìËžFQF]…Û1Ù9V¦wl†úÔ¸%>áËà]ÊŽƒ¤¤|Å°,;‹;h!´ò°|„^-à8 Àâ“ÎÅû*pZΠÔgɈä+óò%ŒdiÉÍÉ­>¹^‹ÓíцŸsœ4Ÿ½,¨™z5=ǯóVR[:Aê¾rs°®ïŽŽS]i…}‘ÐVv?6nÞ ÎîÀadiÌ@¸wìY§yèñ „„ ™‹„šc* -ßø±—s‰ÿ¼nÚR,Í–19\ ² ã LðÁH:éå61 »sÕuÌÛØ©ÎíF(}þr%à‘cnBôЯ’Œº­‡Z5n°ëÓ“Ex©aÿ»V_Îhÿ=Á=mÓL¬ú‰Y-Ù0ùò6 ÌEÎdâìòçÉY„* Ïåéå3ñ×ÿé3ž–òKåæW„Ñê” Zmñ¿I.ÛÌ盘å}½•ó¯·ò4nåkgùeîÖN`MrߨÞ%õ–1«åCäµ{$`KØ„`:=‰õC&_)Åâ~`þcwìb™¶›¾ ÝÆ};ÄV–¾†¦¹\áñ<’˜f¬ü%H\ ó,ÀŒ›7ÏPY‹)˜?´Ñ" ŠI¦“Úÿ¡|š|îŒÛú¡Ñ%k îdz¿ÀáƒWXÈx! Bç7— |­€¼ì‹{/H2ò’Œ•?LNàò×çjýæ’7hý9Öƒö/d‘w'óHWŠà•ˆ‰èõa9Wÿú¨„› Ynnw*Kÿj÷_̪Œî endstream endobj 333 0 obj << /Length 3620 /Filter /FlateDecode >> stream xÚZK“ܶ¾ûW¨tšMí"ÀgÛe•“’R®h]9X>`f°3Œ9ä˜äHÞ¤üßÓn€ÅìJ9ì’x°ôóëÆ|sÿÕߔɫJT¹Ê_Ý?¼Êr‘WúU‘K‘äòÕýþÕO›w—fü¾;u‡›»Le›¿ÜÑóáÒîƺk?$Y"„¸¥ÞºFzû+=þñãÛ·0Eþ÷æçû¿¿R‰¨”|u'¥¨²Œèï»qX’>™qw;Ó4HÜþv6í^LÓ˜ò›¯ß¾ÿIÿôÓkØÀëŸvK$KêM=ŒHåÜÛ}½»~I©ÞÒÖr!3Åçôñ®;ºvñAoÏHNòyÛÆ#ö %Ü þÝÆv2Úþ´ 5ç nï­ÙÚf`Âp"¤DóÔ’èÙ £E¯Ý?þ8ô>!»ëšÆœ»Ø~+ÿà©ø£¹wOq°çå'nôw:è]lg(¾Å'f ÊÞŸy)fÏ;wjG5-DRè%ÕßcœÝ5f@Þ'AU'ÖyzÝvmS·¯ñ{Pz=Sz¤-ÅBS÷ÃØ<ÞäÙ¯Òr3-½Ô-Ñ8¡QÇÉ<ÒË–glÍ`÷ôêç˜öFfž×^N[Ûó„z>ÜÈdcœnò‚GÛîìjíË`—Ÿ…‘álwµi¨ñA)Ígœ¶Î ‘•pTwDP®dƯü ”ÞÜé¤Ú˜þF–›ÃådÛQ@ìöÖv2Ò¥ªX‰ž¶š§¸Õ§[C+uæ—™” ²£,Z†©õ@TMkšîÐ]¸9vøÔÌèx~y™¤@4÷D‡Ëvëñ2ÚøúR‡©uË mìX0·”~îÛÞEˆ¦©È‹p(ïþŒÕe8L·_Ï¢þËáS?t=K8Ÿ‚h—¾Æñ"KæºE†ŸNi>$‰rê ÛGz†OýMžÈ­#SQ¥ér•{7UWàÇKߥ§lÉ´HÓŸ–êS¦è¦ùYs…£¦ÝÔ1 ÑÞmkI ƒ…Så ÕÎDUé|H¤Ž‰[TE°»™ éù˜N8ðÜrÛ‚„AqÞ®ï÷ =:?Ú¨]aÏN3OÅJl3i¡¥R3‹ËŒ£‡*A«Ûæ‘ÞöölÛý@œ»)vüÍʱTz³½Q Ø1ÌŒé±*`C nݧ¬R Ä‰Z+1Ñ…³w7àr?ñÎPU®ǞÎãcÌ6•ßÈ~4=¯˜¬×A ÑD,ž”Cì/Kõ2š³Î€jªxPëYÙ!bÞœ˜áY·KÓ8dßǺ=°-ÑãŠË•°E=yg|®\ù\”ÚÂçº3ÁqER®âáûÚE dÄ]¤d¸®Ë…'|ùTGžÕÑ$§Ž<|‚¸YŸ&„ÎYT-Ëkz¢K¡,;îÁA:•qÃæíD|†„˜>¢À}zÓƒ<¥Lç!ÉÎŒ1t>`Uà~õÚ]:Í(%èAoN”~ ¶óÁðÜׯR*‘e_æTzriŽ2s™š.ªÃ“mDƒ—^[áÐ);õ\Úú× ¿7d+îÝ!R?àšiÇW" ¥§(>*Ȥ¸â g| ÛgX'Yyu®…^CÄïÉMXüO‡BGQ¤ q ãi„Ñ`³0Á±(¯Û"8µJÿø§R òŠÒ{^X²Ø p ouåtÁíõÎÑû™–æaÌêjã„ì„¿©G6úB‹¼Ì–|«9†·û³Ö‚ ¦\:0Bð‚¸sHËû8§+`VvqC7EÿˆŠÀêº*˜-¦#`ZW£8:Ôõ¾Ÿ€à“eÑ€W”èÆ N‘ ™ª/„:DX†k8%~ô«¸¾S·· õ°G‚N?¡‡xl÷qF¶ùÆ €}ÝÈeÓñô™{% $sàrºÎ‘AJq%!ù©ã2ᕶ ƒKÅ0…¹bj†wv yäL0grª»Îâ°PwÙî”2 4†€¹HËâÅBc…–s¬6"„\‰T…Ú¦>Õ­‰Vp@¤E^ÆX?U²7õÎl;U›%¨§.ËeÉO9/.êim,Ò=P-îÄ¥:½\ÚJPàÍYŜЕBÓ4‡®¯Çã õ2KX€0pî»æ=ó%u¨òDÖ$ÛÇ%­/ú‚ —žVåsÜšÎJ §´lØÚ%-…W9•hl¿í†¨å%ÀNù|袜4 [©Êݸ'4«ßÿóÇïbtä\óoq-VÎ{áˆjV{ˆFG•T/í,R•/XVåägñêÅ 2î -agèà–XÅy¾º2KÆùvì-ùïK3zg²MV°…ÌUd²™bcȧÊÿ xˆv;Ý„îs_·¬œ‰ò¥.Gƒ÷1TïR.û7EÂÉ{ NØŒ4“?qð?ÔmÈ¡QákÜqÞ¦Q‰ª‰.AOTö2n<+Àæ_¦½|ï–¦¹ˆVþóDâÿ}kvᶷëNr†^UJ¨4__ª:BŠÝížî™ñË®åšuÅEŒJÍfï(†¹³»[7ä¹…žžêZ $`²'þ Y }o+ýtøM=ÐM Ö-Ú·¹‹þT‹Ü@Q²ßEÜQ! ¼ßh_óåòÀÿHgºÁªÂ‹r~` “ä‰åÆY}S3±†b¼¿>ŠXD¢çKÐT-+H"/«9ZÁð™ÂN–ìç0•<õ3¦ƒ³a qIåa…‚œ[r`fcM%¹é®ˆÏîhx‡'-ÏÛçqˆÿ~oíi}±¿²Z*µ<1‘}´~“¬ 7Š 7ï§rÝmM‰ÌQN8\x¢×J7@2Tƒ~˜õ @µüf¡_,mßñ´9õ¿¹**…)zÃHI÷«TŽÔÀØ<¯£ÏfeþjUæ¯8¢Ïw7|Š‡ä aÀ?á]åѪª¦µæØvs°ÉWNÊÇÏš[N¦tŠ'Œþ “Ô\a€?š³³+nvU”˜ Ù0ÝPOcBi¨àëüb–HâzÇl¾n "æÀÌg¡µà,Ðã;h®~ãöÃå·¤Kè‹ÖFR­²¾pM i€ÄJÈvAe4Í• Nýê»û¯þ$p™í endstream endobj 341 0 obj << /Length 3446 /Filter /FlateDecode >> stream xÚ½Z͒㶾û)¦|Ò”W € ±±SåuÅ)§’=xÇ•C6J¢Ft$R¦HLJ<{ºÑ ä@3»k' @ »Ñýõøæî³?|[Ê“-ôÍÝþ&׉6ò¦ÐY’êìænwóÏÕæñv-L¾ºœëm³lÚ{|V«êö_w…÷Óà}­ìû)½yª~àYá.Â$JænVÕßfåê~<ÕíÜ®•R«»Cs¡M.‡n<Mí6¶-¼Ñ7[êûù6ËWõvèzìöÔ¬Ûûá@}õOcu¤î¡£®áÀkÂb›zñî¹ê«Œg«¡î/ÈÇÍÚ¨$åÍ:Ë“çÄÀû4Oí¾yº:w—K³9>Ò-¶k×õ±95m5Ô;žj—Ç¥_QÇ¥æÙ¦kéA% 6ÈxΡ»ùê¡Æ_â™7>5°±=\ ïŸ~ÆÉÕq¬/ôŒôFÎNE"s,o¿Žœ°-eáÏ £5«ži¯ŽG&q—Èd!“B-döÊXd(!û&4í‘ÃÛÑÿxÞDºZ[Å{Ÿ¦ÂJFöDÀ îíQ@£¢?/æ[•®PŠ¢àsÁ]½¿Í€èñ8PÇb1GÐì¬eš:ÝÄêj—ÔŸ‘´íx¬zP˦£³¯‰}Xo?¢.jµz8Ô-µªÝŽ¸‚6|ºp³ãaü“«S·«Èˆ.WpäÛšg¸Q_†æjv¡éû¾;-ft}sšÈûÛõ¨¹­˜’ ÏD¾ys&&½Vj½²Ì¦óS®XãšÖÙCÃ=5êî/Õé|¬_Gt×1I)‚,‰ i9ìöåšþïÛ*òi¤ÇÿÐß÷ô÷ý}3{ú;œ2¾òý+Eõµ*!Ýn)~ë–M ò‚#BKI£ûs[P s"ŸSCHjg~E[\iÛÕ{K‘•O°\_Ÿ±ÿí×üœ¥ŽWÇï°×~×@ìw 0OÏ&˜0ÂaÈ$†% |éælFMF † T°ÌlU»šzÀ駻•:Q™÷(`ȇn7Äç(„·`â—Ïc„¤:ɵ§M 7kvàlÀ™Ö0°{‚dzFO€ÿK–bgö â 6ÝšlwÜtŽeRfÙü$­1—LãÒ\j[Öm‹Ló¶»è´Ž ß <æ…çÛƒ[8"E$dªxÑ1ƒ°ÒÌÍBgœšÕ·hªTaõÀn‚˜}¹¯‡ìS#$ó¤,½ eiìà$Lñç¾wûûÅC &Ñ<Ýg-ËÂjOÄ–"Ì—‰1žùjÓ!«äcÙù>°ÛrL‚þ¸ºµ.gQÏ®kcV³æågö¿Û~Þj® ‚ÉHžýÐ]jlêU7çq îf ¾t¨$ ìÞðTnZêOeçUÏó@lGkCœzb‰B£ÃƒhU¤^GÞ§™ŽðQ&By jÈb,½m4JÊ› û€§kÊ,É¥ì$7zõq¨-ÞÕÌÿ5XSIš‹¬YÔP€¯&]„K„uº„5=¯¯báy’—ùóÈ-ÒDz#8UÿÆÓÄPBÆFf²(WÕ0Ô§ó@CGÿV® 'Î"7±ý#ùÒªr]ñ@ïö5„Tí…^ÿq¼ÌWye=âQe梋ë‹(_Vi–º3gjRó>²w4»ÍÌ3|L–±„• H.hu綘Çôç¾ÛÖ»ÑEݧAkó\€[¾à³M’N"qš áö ì/¯]\ àšs˜ÈÐ J~cŽ{8Ê–éË~XL PT–ˆ<ûˆä„Ešå”¾$*‹’ÆSù}}>VÛ\o©TÏo7S¡e’pÜ¢(?$3ó2ð§X†æiRdÞ~!§NêþÐa²;B-Á:? ¿ ÌÒ¥—/ˆm– ’ø¼}ÆÂÖ¼ Íã æbÓŽ…xDò;žxäSQ©HTÎáAahLñ©¼³j‘e¨³…`p.…Œ¥eW¥ ’BûãØ‚† }U(4©òŠ$23‡Y¶ my@0aó8½!-W8?%í®‹—c3JU 0R=­Ø\«¶HÔ:x(jn*9Ê7b[Œƒlúb£PL»N@&êi„©´nªÅÕ_%Ù~ÌV¯ø¬¾®zjwüO`€Ý\åÀ"W_Wl44#j•e ê¨~²ð|±Þ K% ßµÍÐTGEš'J‹„È)ÍÄ<Šù6^eämÝ;¾æºf)‡äbÆžL¤6¢á‘£CDOd~7QF±û‹™A’+uLâÒ ›˜¥î‡¶\å4•ØDˆàdË2XÓXæ£C/–Ƙ‚(§ôÀô:]@£Ÿ`K‡™‰/G«§Hö嵒夯}³Áfdûþ«–-ürhöîÊÇ)0:Ñ颾±Dô õÈÆKC9ËZ¿Ö}w56‰ò¨SH~[!¶Â|¨&ä3M˜E9äùM0é+Þ7SÁ,“d ÜgŒ¥«4i*¨ïO®O$râ§}Y}9½‹xöQﾎ‰%O€Ž ?þ$Û˜i6¨x¯`Òå°ÉOݾ]®B@¾­%d’ ñ©¶¿”–;rœÇ‚6:qO[d™µ. 5¶* ©â§Q6ÓÅ'’ÿâÿ'ö3øïBd l$©fµÒÎ_v§ó8 LÊâŠß×yXó¹?^©,–Æ㹫uÓª¶>&]yÞŸ@{~Ý^²‹¦…Æ }Ê e,ä€#òáÄìŒsáγ,ч»Å´s_ï¾µ ˜â<œloHÀY‘èb‘àîÇ£»oÚ!'`zC¢}f¬>ɲÿù3‘Ná’‚É7"ŒÂᩬ†àì5¤áoŒç †ÎÜ3Ç‹4ïСëd¶i¼ý§?8Q,¸ÉT¬~ìšv°ÌA_¯]‰•H«å¼ÐÔ–º¢x$”HŠéÊòÓ Ò×B} iYÌÉ\¤+ù" ­•Àr£^º~&*Y`~ 97F Ê]µK¼Ô·å é¾v‘j®ÙïWú¾¾œ»–¿?üÑt ÀœY¨2û@ªùÉà‚“±á·ýíh¸j«ããÐà=Éã­Î¹–FÌ´îÊO„Eör2€#ƒsßíÆ­Ý=|#¤¹œÅ/e6ûvÁÏ@b”.]µÍgðvUÃóÈå 8‡Á„š+þÚ–râHåT8õÐôT‚ݹ¾¿+µ²’ŽéƦá’˯uL+L¹ôjóØÇ#ih“O·5&Is†ï«Ss|Œ_”™E¥S”˜«Ø-~sœxaĹÐ2¥YVGn²ô¸6i$Ë•&µÔbx5’ Æ)9t\êp¿¾©6Gî±/òWca”‹›௠ž~¼½¥S_) Æoì´ømûè ºÆv~‰cÁ®8ÃB*̪ˆæ4%L~Ûõ‰”XNciÞÓ¨‡yÛIæ`9Aò0“PrõÎ~®ÄQ5$ÂeÄAI= G6:À®à6ž¾ óº¾CöMY ä€¾3€Ä/6[ÜÑÈx‰WjEzoQ°Å»+P7-u8Ó…ú–ä=íì©z/„üÛúÍ·y·~mÅUQ6u¶@öêˆuåápzláƒ¾î ¹QÜ«™)a~å*g—ñ4]&."²cS_¹çì«ö¾žnz Àåé#6q­’åö;oÈ2›ƒ*>Û8NfEaÏÇѳu(Ø°ßdrqµ ôeÊùØl›2Ú}šØGª"‡Lnºÿ™GÞóŠo–#C7Ž«6§ù.›G"é\ñ'˜Øé‘‹Ï8¼ÔÏ=•áhVñm76×ÊuÕòÔø)$7ùdÄ׳ *¡ÃÏ|$Ä÷Óg>°þâ3èá¯ýØP»`ÁfZvÄÇÃWx€§ð\-Ͼ¬;{¡÷8 >+—&òmßQV¬¢›'7á.ŸÀõ\†ÿ[s•§^²T ¯ýér\d&ÑZaµÛ^¼SÕÌV?ûóÝgÿ 9¼* endstream endobj 353 0 obj << /Length 1794 /Filter /FlateDecode >> stream xÚ¥XIoÜ6¾çW ‚4¨År—T´šÅH‹öb»½4=h$ÎŒP--ÜC{Q’£ñÎ!>òíï{”ßܽúî:Æ›%’ÊÍÝ~#$’ ÛD’ ,Éæ.ßüTM®J² Á¡ý=ÔÕG,p5Xò?ûsc¾µ?oÔïCÙk‘›+w Ñÿ¶ßýºÁ›”13I¿ÖdQw}ZgªÓr«öÝšÎÖªOAªí-õ£ý±)øæÔª¼Èú¦íFÇ!}l–>PÔ…ñGWÔÇ6Ï2•(a¸ —·ø¥>.£ldL·÷ú¿¦È»mÈú£Ò 4µ Oi›VªW­æ Ó>Lþ¨”å46´?†*€+틦î nÆHÐ5îÀ)þ4Ù?[‚Ðf ôÇ´¶GÙ°ÄΤÝ=Mœ¢ïÝSÚu !§QA"¤ekÀ6 Ôtß´•ñÜnìÛ¦²«ÔþÔªëUnצnhrÌ‚Í*ùYéÿï·öe $ÁrþŒ|Y>eiÍT Y-ËKžÚ&2e‰êuaÌz*…¦îµGi9¨Î6î÷Ø e>JÛÜPH±÷Ý2t:^Æiðù¨t8 ö*íŠ]© xGÁµ¸i-W­PœÊ"ƒ4NÙ°gàce !…ÁÖ x±@4‰Çè[ÕùíŸïÖÑ'>•û¡Î à¬!HŽ6{_䪳.§ö`L”‰G0}°ç)°¹h»åj“$" ]&i<²®&dcÏ©ë\#AbËÍÛ†c¼èa:— "ħ!Å}›õJBÇ’IgàÆçA®ö&F(ÉÕJ²¹DBú^¸¢´&‘çiêÒ©.ªSÓPq˜…¹ªûâ#Æ4ÝeÑ;ïvßÙ4Ëš6·ØRç\‹›1£7Ľô–c‡ŸÚW³[Dj–ÝÃÚ `1 2º|³dJÌG„Ö(/]#pB‘àbY㲨Uª{LHh¨"Õ¥ƒ`4 U0:”=5ƒ¡³kl½]´l;-µ&t²„jÈíÌ‚i»t#V;'Ø?œ ¿ì(ðÛЂónúè!íGvñ/° ¿ÒŒÎ×àÀIò4‘k‘çbB cû¡ÕÓ3Áffê-0gÆ(¬gc÷zxèã¡ž dŠ {YãE"烂qÄI4:û¹ZhÊ|@Y¡,A©¾¬síG9Ï­»{fáè7Y†ã쀯íÉ£‘À)J.S²¼bŒV=,ÜИI½–‡¦-úcuei¸pNv%ÃnVhýV03gC8J8_^lwÛ™3W–ƒnÇ^~εS-ê¬éž…<öuß”¥½âl÷j7NnÊ›*ÂF³_èÜ™^g† "æ§ó$/ºlèl¯Y¦ýèUá Þªl²!òýŠýF q,–o§ŸìË«S=ê”Ê5Î>ïÍÛcS5‡õ§êu«>-^ŽPÁƒvsö`„« ãò5ÛW¤­ÃŠ\}Álür+î–˜ÃU¢Ÿ ´‡—t¹Ý}ZÚ³‡(à ëšúê)3yÚ§ ¡&ˆ“Qž–·}Ú3p¯ÚtÌ‚ùúçßnß_xÞ:èÞ uí±æq˜‚•‡nnzØBpp5aAÐÊà pëM·Ì ¶â‹€KÏãf{Ý»^°¬Œ¸0âø"Èèó@öèãd8AÖ•ù¾p |ÞÂa6[ºË™aŠâø…™i|¿À|ì—aY½.­ÜÊ\‘«ýÊŠ¤\bÁe'k*¸?ÔÛž5‹e»¢6ù…»bÿ8OW£ðâˆ>õ…èí•5xÜYÍÞöÔÞ¾æx=}í,Žôõ;m3Œî{#NŸ*MðÆ2B fðhGÉø°\Î(&sIy¹Úè/5 ÖLz¡8F”c”0Ã"^ À΀+»X3 ‚‰—XÐóRl”"ðž ±“"â|!%çR|’Šã„z©XDÑy[bòPJ/‹Øy)9JQÂœÜõ2JÎKE“‡LÆ‘÷>sŸŠ')œDÜKa–¬æp6Ôu8D@Saè˜dJï’µDÎDÙ( Ço_°ÊGQŠHIŸ˜¯ä‚¨˜¬ Ë©òLFDå(Ê&<öV“ˆð ¢ÑÌaŠ§jb‰ ¢ñä°ŒfP•ìZ†Ÿs¥C8Å&TN(1ÄŠ7ÏQIJŒ¸ØãÀ/TÉôÇ:h 1äÛèO½P)÷~²xÖŒ†x¡JáUF,^¥!^X éUÆ#¯Ò/ô2šÈ·Œ#^¨2žår¸!VT>û•HÆRÄ%Ÿz”sJ¾J1S.™WL)‹×þ¦ÿ&%᳆ …Ö18yõþîÕÿÌhÉy endstream endobj 357 0 obj << /Length 1358 /Filter /FlateDecode >> stream xÚµXßoÛ6~ï_áG­Uòø»À ti‚mX†­ÉÛ²E¦m²”ÉrÒÅþöEJ”SäI'Šü¾»ïŽGI?_¿z{A‰œ™ÄH³ëõLÈD6S’&DÒÙõjö÷üòP4¿T»jsC©ê|“—oKb¾²û&/Ó&¯J|FÙbIqÊ|IÉ5Å;¡Äo@ÈÅ?׿ÍÈlIib„8˜G`"9ôÀ„Sõ,`1æ*zL€{,O–=—\E®<ËcÕkMt6„±gëL ¼ï¸4øí…fƒBA  ÓÊÓœß/¨˜Ûr±äLÏ›muØl;Ûz£ÈK›ÖaÐÖ»½7Óº}®æYUî›:ÅYH>_¡—\ó¸ü.­ÓÅ…þÖ9¿KPÖUýˆn‡Haªwß ëœ ´µÆ0Z÷oe#AŠÄ¨~ŽÍš@V­Ã\2D„ÓÞMöJ@2™p€nZZ®F „cBùq&ùÁ•$a†w¸8³(˜!3϶i¹±¨&`Îom–öÖ?A©ZH—7Úæ£5Š<ÝÛ•¿yÈ›m¿ÊÛ|³uÙpvU¯:s‡—ßy†ž.Ò†–—H”f.„Ö!9ØæÙ¶­Ûeˆà(æÜ9…ux(cu¬’v>E‰ø£éW šÖè3æwn뺪÷Þn£bsW_üÐΦe^nÖ‡Âß·%äŒ|eË&¿!ìÊ/êk¯C+ÅÖ–™õf«KËphî¯l²Z›>W‡Ý.­¿$›r7–W^˜ûàÒÚªÅKðQ{Úo+'ùƒ-+p¿¹hÝÚWc;°D03ØrÄ@œ“å¨SH¯4ü¾¼-컱¦„&J9¯MröÞw£ ‚ëRŸÎ¶®aëJäøŒzÔzÎÒ¢x7Ö¬QP‡yÅšL=ÍOþrQÛ½õŸ¿„ýÚÚ¯¿ëGãóô‡Í´[à¼lÎ@'vå0ºNw¹Ëä „»*ßï«~YÚGXeÙá®…I ÜÍaÿfìx¸·õmåºÀPœ¿_wâ£ÖÁ`‰ oí}ž¶•ïæ²û|uH‹½Ïv8IÙ£xƒ¶dNÿÂAѸ´+Ä£ G¹B¬Ëô³g=B ÅÓH c_ÒO<£ì’Ð8À%Ãî4c‰–&øÓMm‚ƒ¨Î*»^羪}(Tª„)8Žå<œ<>ø«ûQk»vãͯþrŸ‡0íÏÚeúý·¯ß¢°£ènÚ¯®Mfö®ñ¥D©Àа;¡TÐéAÆÖªɯ¯-÷ÆÄd "ˆx*ˆ‚°„&Bqf&yÂ#ˆJN GÄp$ ¨'È)•¡“@T ˜¢#a¸ÕNÌΠc¸> stream xÚْܸíÝ_1µO=7W¤nç*{ã©8•l¥ìIåÁöƒZbws­££Ã³³ùö¨k4™±ºH‘ @âúÍí‹o¤]¥"Ttu{¼ #¥þU¤7n‹«»O^èýÅtÝv¦©¯÷¡ w—¬Í*Ýë–> O.é:tÌ*SÞӼϾh^îšF äõçÛ¿]yW{å‹TJ"ü¡/¼m[GÁt4þüšÆ»³nQÞèãÑäF×=-œ³Î‘rÄó¦îú63µ.haw¨MÌÑÄüRt§÷º3Å•\è¯&«sýŠ>}%ÂH2^&èôÔjÍDš#s–Š¦rt¤i¡×ï~b¬ Dºq•Ÿ‡êààIÖƒ˜:>ãÁo&dÄ€ËQâÈ…Ásø2åŸúü]•ÝÓŠ GPˆÜ|ò<…œDˆþÜ6ÃéL»ýY3eoFYy±P*…‡X¢#‘K†žðâ2«‹ t2± `»m6ðEžÀ9‡¯½–Éî4Tš R¡wèTW[ÜŒDè0 °¾înϺ³ ¸Ú`ãp)ÝôcâçÑ29Úµ?{s "üz-Ã.ï l´<>ÖŸ³žÎd­æ¥†ü½Ðy\†[’]@1ö½©O´~iƒÜåºëu]8jšV¾âͲrÐÝœf æhòó–>ãIx¾"öwݹÊ‚æöšq0¿&r1NwÿÄó÷ÆNcÖCDµÐCÄÇ42ZK÷4”Às{Ð~ÿz“?C {}‹½ÏÏ|üt Ä Ü‰AÎ|ÏÉ÷ÍŸÉβ†­î%Ê^ »¤Üš?<fÞ£ÛÁM4\M]˜|}2?©ÜzÙÇÏz3š>º3Xa4fK¥Uœˆ‡6CÞâ4üðñÏŸØPõT ?öÝÙ™Dñ®YM·±r…N—pWv¢{?JE gW:ÁÃAǘ`2#€C•£ÑÞ~Ù2?_D“Ÿ(LV6'PŸH%h„„z®¬øÒÆqöúù²SNœÅBËìƽ»¶F!='¨ß_ï/Ý„¯Báá3Ý—ÁȈUäZ€?Q°r-D`°dTÐ[Ù>,,ü|ß™²¤™ŒîibÆßtÛKÑEŽ9½wÈ=)É>p’gϲCƒœ vÀ…|5ã¹õ6ÕÔ ^ê™»ã½ã ù¹©š“®u3t´P eo.¥É³¹·¤½ãµôv ~>Ó Äátžú€à@#?y2ÚC"”?Å/J5ì¾Nö‰êéÚ¤þq oŹ…(’ˆì9qn&uS— 9ë¸p]·Í(EAuÌ%rV¥!;éÒ ) fV¸0YÓ²!`N«j ]nac VݘÓÐ2¬d¸²ºý–÷62H ÅrGî‰Èèìõ† b)â(ú J«®Ç“ X©~àBGÀ¡#X‡Ø€lŽv¶œ‘L”H¢'¼8¬0Yz#ëŒQ<Œúá5³Q(R¼´ìA§ëuy9%-R’I[Š`­I¦iŒ @Ó}ïLÚƺîq+þ+Iifh3>§ö8i5Ü»r:+hÊ…µjðx¦' ka>&Ü]O@Ö0`eLoðƒÌÙ_¥ã´°¥*U Á7+ߦ0-xcÈ ¢´0Šv7øF÷4ʲêRº÷L†+C`V¼dÖ±)Kâ9±È·¼À™š)UG[”¡ÑŒaʬ㥵σMëó&håóâQŸÀ ŸM™ž;hwÙ9×#›þl¤lè÷_n.KD ¼¶`¢¤†蠜ÁrqQÒÍQ2ìûŸì›~šîÁUÓö\…] 𠈊AåËUý¶HXJ]Ÿú3‚c 8?†q¿ •Y=僱žÓ÷Sòœ¾Ÿ8]M.ÔB´š jçápuÆ[Ô¥H‚áìgùõ%;]œRÆÝÓR¦°7Âwp‚ì” IØÖ¶(mÚŽàV“øü†0ZˆMå<„KÆðÐ`ûú @¯¾K Ô7+zB æúòrŠayVn(ÅÍË©}° ¥–´l™xËBú£üLèn^ÿýÃÛgVßï,Ë•uŸ8Ö / ÝŠxO]Ã#^ëtîAG ¨¢GÑôî¬ÁèØ"”/ÄÆeª ¦~Š»¶`«¦¦­XmTnÔï‰vü¬òŠä¤œù§L/Áõ^zž·û–‰Ì"5?¦BÊ‘çº4^hK:{ÅP%ÿï3öZ•$o‚%WÒ:¦[*´Aîmóq‘»Âö–¶Ë *ži»¿åDO×Ï{"bÙˆE)}Îó$\v‰8lÙ¶ñ ´änß«æ¾ëé"³½å¹÷Iו]>ýò©iÁEUu¸gÇúBÉÒØÜõ¼4ðWS xȃüÑ‹(͹i  }  @HO<,ˆªËÐk†¢rö'gæqí‡ò\—xk~`ráaŸî=Öé…tS„~ø% ÐG[Y×@•" VΧo¶==§]ÏÛµA²{ OYåÎ*Â0Ù‘q)òCgæMáLÒ1æ,jår}$‰‹­õøJ?̬KŠ\‡kT¨g6‹/e–S÷`@®Ü–³I®/3Œüæ=N+^ î’ϳÒèé$8Ž T5Pô_øšßÑðfñõHbñÈkÎÁÞÌò±‰óTcuŸÉW4ªçåõ†]¦þÏÀVh‚ bÜ&£ȳüY¹]Nš<¤<äýà¼îƒýQ$Ô:ûNníå‚A«ç.ö¾™±ÏcÖVFG=eügƒëxÒ6ì:á,M' ÆÖ†PË?+º@Ò¡¦XM^QÛ¤·¡¬ YB\9ášfÇñFš&]% åÌi8™bnc.v» nß0BûâY/@yJ¤Iº4™VøIB%WâJ¼$XÐÆ¢>‡oÛ1‘ÿ•(à©Í÷ —m ÀŒ2½!·o—>z|ì¿ˬ1ô¥ô—#ŽóB–±LSRj á8FQËⳫ¦¶52¤ŽK$á$ÃRìB¯kÖiìéϱÿOµõlj[§û­§,AÉ´Lª|ûgwÆs qì{ȯ!¢=Pdüþ#ÁOƱHçÆø¿ôfJŠp*1PÄMM¥@è/ÙÉŒ¸}ø´k[Žáß~öRÒÞÀ¯þ_ š{jTǼ K ¶˜·°wi1ªðòØ7F½ªª¬5šqÙŒ– M:‰‰P£vL·è\8œ” ­Ú¯>„Ð4}"ºCQíPÁÃÎMÑ-þŸ°“ûÀ¦·æáŒZ¥[\€OŠ´Ú¥xäm–ÿtƒMÎ6’tÔ§N3’© €¤†wP2Q„=ÈΣe€{/ÞÞ¾øÞ6ûŒ endstream endobj 374 0 obj << /Length 1047 /Filter /FlateDecode >> stream xÚµVÛŽÛ6}÷Wy¢Ãû¥@¤Ûhû¤1Їl´’ÖV–R[Ž‘¿Ïð"Y*¤ v‹Â€IFs†3‡‡¤ MÃJ'R2Ì´HòÃæŸ ¶þ]‚‰[£u0_§þ•ÒÊ?HÅãëß4ùµÛ|ð?‘Ò•N°~Ùm^¿3<±Ø*¦’ÝCÂ…ÀŒðD)Š 8íŠäú»*ÛíçÝàK¦¾Kî¢{§}{ˆ>ÓxŒ–|êÓ6UZ£EGkÍ@Çaé("é„ŽlŽ¬u: ¥±ˆlü˜»¦EWÌcwÞWÃ[QPÙ¥ò…핮ϪÏÚ"oêü1va<•±“V!çlʼ/ªl~8š2Ì—…ÂXÆ~œ¨˜å"ú”|ä¸c0<[‹v[CP|w*}R>cM1.ÏR>œ›¾þâr¤v-Ga1!/ 㫘FU×!|1î÷äZèãP2kÄBŽh·ý±YÚá 3ö÷qK ì"*Ñ·¨Jb Z9[ꥪ›“+¶¡±Ðì´A¡ÝJ*[1¢¼-Š%0 !£èÝoAç¾ïÚ€s©={µA÷eÞʘaX;O/ËûÚ•ç«[‡ã Dº81†>Ú5öpà yvgîˆ$'ø£¥Ê\Ö#. Ý—¬?e±"T§„yR<øU<øLpçF2øvDò–¡tÁ|=,¦RýŸ'€¥A,‹2 w,S÷,ËŒ²•µÿö¾0Ú¸jJAè拚娕r®ªÅ4:–‡ÎÑñºÖ´ Tx’˳ò…˜Çîàf* L®P7Ã6åXÀöšgÕËFhµ];„¤/ÉÍÅt‡øD¢c½¯"ð—¬Èî¨òÉ* ý×Å"_ ´w75;Òþ/¨ö×r!7N17ãá}ÒP Ûµ Z!Fƾÿsi+ ÌùX‘ nÜXô­;‡É\ŠœåŽ6pˆÀ—ó%¶4Ìí±Ù}Ç5Ç TØŠ$u#aá«wõþ|Œ¾ô'‚óã£o¬¿º9ó ÝÜ_ØKïe~{øØã×=ú[ñ¸€…[*± iæé{*häU.Â=_iL(T@¬áží=¨ý; Eù endstream endobj 361 0 obj << /Type /XObject /Subtype /Image /Width 676 /Height 563 /BitsPerComponent 8 /ColorSpace /DeviceRGB /Length 39276 /Filter /FlateDecode >> stream xÚì½\[×Ýÿﶿçù7iÚ¦IšŽŒ&MÛ4ÉÓÆ I 1íÄΞŽ³÷Þ;q@aÀ Û`l°Í‰ ›a6˜!„tïÕ½bØƘ$µ]ÿZ›þ^ÏúŸ«-@ØÆ6øó~}_¼Ä]çH\îû~Ï9ºçοÞñ—ÿ¸ýŽ;þ|Ûmúã òC ±tãøýŸþxËŸoý#1;ñ;±üÊ;ïøë_n»ãŽ[-<úÈ}nã1÷qÿcº‰Ç=Åc–xÀ9ÖÍψŸ˜#ÖÍŒõÖxÈM<13žœ»Äz—xÊc<ÂÇ“ŽxÚ›xêQK<ãM^˜'žpÄóO¼èm¬ñ…õ/y/“x‘Ä“ÞÄ+$^rŽ§æˆW-ñòìxzv¼f‰Wæˆgìñº%^õ&ž}ïÍoZã¹7__x½Ì»{S½V¯óáÅ1¿å×\Âógeþ0_uîþ"sþÍáôG÷îT1vy*’ó–?{½;ÏÉ„Ë?ÿ•,ÿnîÿgýÃZþ‘çùg·],—/¯'öë7+û•Í›Ë óeÓz!õp™q5žy­v½’ϾԻ7‚Y³%2‡qÖñáj(Ï.{|f˜UèA”n­ÊÛÖ…=)Û"ô[ÿtËmþ#ñ;±ü½ã?îø3YxÏÚU@ –GXuÛŸþƒOío#¿®]‰@ b9…E÷·ßö§;n¿®G bYºþ·Üôç[ÿ@^¬¹;Â˸û®ðÕ«CWE†FF†DDˆçðpËÏ`aaAaaägxxPD¸xõªÐ»V‡YŽé}@x–¡zä'±í¼A¼¼jUHdxpH°00ÀO(ðøûŸä5YDV͈ˆ° áDñ¡A‚¡¯Ð¥¿ßä'yM–„…’;‹ô½©@ ˆ±ü-·ÜäëyÑG„††ˆ>Ÿº¯õÓçÆoOoù„ü$¯É¡À‡¬ŒÛ#‚¸ž¤óaAâ ¡ÀåºÇþðƒ·ãäÒmiä'yM–åd-Ùì"ÕýêÐȈU«ÏhßUâ@¿;ïôñ#ŸδK+V“?ß•wòøø‰#ȉ$ð³- ‰ ð¹såJ¡8ä®E=ÎæF ËÈõ¿¿ùwä'ñìADLÜ-=8*{õ_ÉïL'½9ðúéøWÉOòš,!ËÉZ³¸‰åÅáa$‚ÃBƒ¡¡ARɆ-›“RS6%Å'&Ä’Ÿä5YB–‡š·‰³êÞËX½Jèïï»ÒzùôõF® óþó †¬}è¡ûW‰|äÈàÀ'·ìW ªdŠ‚ÂÏeÅu¬ÞóaFiŸ^7ÆéL}Å’ûE"QÄ‹ŠÜö!-Y26PðÉšç6ïe¥Ÿ¬ ‡¬^œóç¬N`±Œ‚Xþ¦›n ?‰gçˆÈHqˆ8`}dÐß䯞Š}áQëgYNÖ>`é© éoŒMJ’+â$qqÑòØ(K×d YNÖ†…’-…–|obUx€0è÷w6t÷Ž‹'§Õw–çÇ¿r¯80"2Ô˃ÌS¹N†¾UϼQqŸ 0,da»‡ …÷ɺùÝu{ß#éÛªUç¦bˆ‹;BŸÞÖ1fþÓ›ôêQJõ^¨ÿÚ¨ªQó’1C?;¸ëå°‡·ôš­y],ˆ\„óç,O`±œ‚Xþw¿»~^× Š|Ÿ>ñ÷/=þá=žâä—Ñ= ô±´Þ‡… >_mø$^!‰‘mI¿”J¾DõùI^“%1²/ãÒ _~D¶$Û{Uíˆ@aÄË»{t–Ë©#Ú$÷ Eá¼ëÅ¡þ¶öR_QP°í:In(|m-w®ôõ¯²Þ„Dˆ…_ËŠ•>~qx€ ìíë¥Ò_î'Š [FŠÅ"?ûAa‘!¡"_Ç8Œ?`hPÈ“_$lÛøÖC‰È9ŠžgPyû^üÚ•> ?¸›7å¼—¹hëÚ•+ýüƒK 5Ù40hÕß»:Gš7³•îãë+´vŽÄû »j:XàoÝl¥_ ¥Oõ ‰ÏÖÓF]þfˆÙÉ_ Xµ¡‰?E»ä÷ ÌuÜó^jEeVì“!Á¡Ï(=(Äåü™ããrÿ9¬ŠðpÏ:ÕC"p%D .×ßÈ»~Udˆ§ˆŒ ‹ä…œøäþoß“0<'l{ÜÇòšD×¾Úg–×d›Ø‡øíÂ<|I᣾ú4ê«O¾Úðñ—_|øÅç|þÙä'yM–åd-Ùæá‡ï ÌQ [ˆÅÂð·•ýf¿7æ|þô«Äâ°Õ¼ðΧ/­‹WE ŸÉj"Û°½»¾|þ°ÆþÁÆ-¯Gúù[/•¶hþìnAPÐÚ7ç6u÷˜Ì ©–ä'‚žÛ\9 2oÖ×[“öñƒAA¡‚ÀgËh²¤sû:¡€§¢]ßÝ™Tžì( Q¾³ußËx ,ëヽ=xD°xö› „Yw ~@’SÞÙßgnBÒ5íܳ©¼íÿ~‡Ë㟠‘.ôø~"u ­zYžÝÔ¯á‹cû‡:ªJb ôøx]a—‰Ÿø|K]·ùÔ¢Ú÷&¯xøsxþÌÃù¿{)ítÚ e<|b·ó}iÕÇÁ¯Wó¯’¹-=ÐqþˆD!žþÄOŒˆ0»ëO`a¨›S]âÅ¿XªA,ãבŸä;Gøw½z÷·o†~ýŠˆÄþu~ko¾¦å‘¿×ýõÞß_Óð¸¯eÙ†lI¶  ýÞ~ûÕ _~LäNâ³Oßûô“w?ùøK×d‰eÙ†l)úÎ] >ÂD‚U6™¯]-_Ýí'……ñƒCÅAbqd¸8DüܶVëŵ^åÚ7>$ù[/žä7@-«zrž “c¾Wo^bhêîîfÕ;ž ôµ»žÕ´«ûÚ{JÞY-8©ò¢-ê^ÑÓÙ#T¯n ‡µ”{ ñQ¡¿Ÿ“ë}<MêlkgVyqpˆ0è™ôóÍ°ÚrÔ_ön8É|pû›ŠÂÂ-;:­fÜab,/Jß U/èøoêLŽúÊÎNóZíPgãei÷~Mì/ðü× ðªÂÁâÇy% |bSƒuK£šÜ“ôÄß'ô÷ðŽDbOŸy £ô‘ÁrÚô×Ç=jwýp'Y¢nßõV˜“ëEnK8ÎÿüŹ9«gŸÀB»SÝ?$úð-ç_-|öéûdû9ªaÛ/öX¿UÇ_Áš„»l@îVb¾èU½¿Ê¥ÿýÊø\’*x#ÄÏI¾‚;ýîþjÅÓïZàï/z4iÀœp}q·ß°é/ 95ú“e¾Q¸C|üAVš[cÃÂü}||…âˆUkûÒ\z½dŸ¯×Ï*ZÈåð¬*/ð÷_ý±¥Ð¸I•|ïþ¤œÏg›>¹Ë_¾ðƒÛÞ”mG×·ìú²Ê<ä¬ñÃpßU_ì5çéyO‘ÏÐßûã …oê*¬ý¢Í,¯‚×B}ï<”¸–ëg "?o*,pýä?hàm®®ùòßò.  =b¾ž?s—Š‘ÓÆÇ×_$´º¾þ­`²ÀÇ×/À/ÐázâßÙ¥‡8ß+zYœëç ðŸ}‡ÜœêáÁóýß!ˆ%¼ëoà]oéaw‘‘!ÿ•{'rò™¿ØcêÍÝëÃÅ×]™µ.tê­0ǪgÿúõÛ‘d{’×ûûÝùùç|ðþ$>üàMr0;Èr²öý÷^ÿì³÷ÉösTÃú >±ŸÝõ‰Äõ!A3־ǧrÃÙωü‚CE~¡oìå7îÈxLàèœ( MkãÅQýr°Ÿ@øĶ–œè@Kü½ƒAΗJR¹ˆ°§K«äÔDô¢È×7—ë\4“ºù¸w½kÑDmgYy_ßÀrYnÖè…´'üCÂ|pó›r ×·ìç·ê+s×s{ôÝþ~B[5^ &·6^?<ØÏ«:û¾˜Ï™ùl€_HèÁõ3ÿš^UØù“·×G¿óù@ßàPþÏ&žãù{®¿Ë=ÿa†8¹Þ?˜œ±AÁή÷÷]zx˜Èù]xYÜÌÏÁÝ ìæT Ÿïÿ@,é°»><,ØS¬Š ù•ÏØËÁ“OÞqdým–h½÷–{ÍöÇC×Þ|MÓÚ[ìË'Ÿºcä¥ >¯ã¿÷îï¾ó*‰÷Þ}ÝÞMo KÇ=YþŽyƒwßyl?G5¬"ô¿ëÓVóµ®ùË»üüI"†Ú/Úf;‡8_Ÿ….º8‹C,„=þÙæÂkSsCÊ£"ß°·ö™/•ï'í 0þ+„.â‡ù‹_Ë3’͘ڬè÷>ø,©Ïõ3Š¶½µ°3«¼ŸOÀ‹æ~ð–âÍS“­‘"{imP@èÂNÞ”Ë:Ãõ៚3Ç ¯Nçjxü°`_¯êägÝlxdz"¿g×{^å±Â.Ÿ¼­>|å}Åü Р9>1Çú gœ!á¡'×ñ– rq½ÏìÒÃ\óz/‹sýü³O`rpw§ºP4ÿ¿X²áäú OaîÖ÷oXçwäÉÛ?þ-knº÷æ«_»ç軫ÚÞz`íMW7Þý;Ë*² Ù2 ÀŸ¸^$ô}ö™õo½ùò›o¼øö[/¿ûÎk$·äø–\ž(ž,'kÉ6dKќհ9öÝ’&Ke}ÖÇëï [ûØ‹½u_ˆH`kíìU½giw-ç/’tá›!¾¾¯ŠBqhHd°ÿÊ•wúˆŽoç~ þ>’¼•ó=¼êÒ×C|È*QhˆÈõJDJ¼7fœ†WùܹRôTáð‚\/ ³¿µá™TÞÏO°ö«vsnõ—ù7±Ò?dõêàq¨ÓçæõÁÉ›rùÀCEÞ¹ÞoÇxSç0±@pot—9-yHþÊÎ.»Èæpœw®w|ò¤ á=Qæ¡•¥®¬¼“T'(T$ôüŽü=ÕÖBn–\]èâz¿»Ò\ÆvzWܬÏaö äöTŠÂæÿ×C K5ì® ô‘üÈ·À ‘%iû¡‡o"ÑûðíM¯­=ò‚y=ñÜÊÖ7îëyäv˪#ëoÿ,âÎвWhppíšÈ7^ñÕWž{íÕçÉ böwÞ~…(žü$¯É²ü•—Ÿ%/î¾+B$š£ö ö|(ªžÕ¤9¼ë¥`’ý?ŸÙnæd]Õ½é‘@þa¾Î­.9©èñÔANǪ÷ww´›¨ÁêwÂüýV}ºGkFiզɋ\šL…A|ö÷ì¶as)ÃmU{ëèùÛð]’ÐPû[ .¼òþ‚Aè9ݶÏAÍò;”<@´¶ðƒ ƒ\?ðçîï9\/ZÀñ½©3¿YÄ{ʾ™h¾8—¡®M¯*ìüÉ Ãß.8à<:ntÏ[a~Þ‘Hèñ3tù¸ø7"9»>0€^´¯s½Ð}éÁNïÂß«âfžþ³O`÷§º Ô‹=±DƒX>^¡ ?‰=IíCü}z¾íð#¿?øÀõ‡Ÿ¸õð3%/,qøé¿^ÿgþÅ£¿¯¾ï6riŒ%* |žy扗_zæÅžz饧-ÒýµçÉOòš,!ËÉÚuëò øâ9«áˆ@Ñšä;K»-ßt¢zªK·|ôD¸(˜Ü¿#^Vä´›W™jò%O† b±ËUÑßÅ~â'ä5]–ÇòèLê:¥â¹H‘0H,ÝõêÖêVsWéÐpCÒ“A®Ý£ä.(H(^½cÿÀ õR¬ïìªI{û.O®w-ZâüÖ‚Ï ò‚±H(\õrÜΦ>µåÓîT¥>+ <ƒƒóoÊy/±·®YÐñ½¬3Ù,àî7â¶UÖÕ4Õ”ïç7©x1ØOàùñ÷¶Â.'•H(ZóFbn›Úüwd{;sß]#òº}GsÔ?`Ö›»¸Þ/0 4$ ÐÙõ‘ÛÒï;Ç«âfœb7'°o°»S=Ø»ÿ;±DÃÉõsDD¸8,$è¡€¿èúý¡o¿÷׳ƒ,'k׈î Z½*ŒÑ=¹þ Ÿ{ö‰çŸ[oùùÂóO¿“ŸÖ%Ï?ùì3ëüˆÂÂ箆K Eþ>ÖGŽóOññø‹Í«Ä¢@¡ýI*>þ"aˆe±ÐüP”•óõ6XèÃoàK®ù!A~~>öÇ‘øø …bó.!bí!æw®ôõ Ï<ˆù°æ'·Üi«Ê>~~3¶ô\ôÌ÷µðÊÛ÷²×|ÂÀ€3>ø¬½œ×ŠÌañ‰ã8Þ7ã Î>=êb‡Ýõ!ä"0gDF††„ˆV‹î,»ëæñû;~ï¯Æî¹vlí/ÉOòš,!ËÉZr…Y½:l5?#^(‰ˆˆ  ¡Hè÷ÈÃ÷=óôº§Ÿzü©'³yM–å¤ôà AXX°88`Þj .¥‰ü"?kàsRz¨‡¶¶`×ÊDø|ÂË°»^,š7VE†’œ](ð}7èO«nЭù5=ùI^“%dy¸-£·ÿ ÎH^÷$gø¯Œ?ðÀšuëzâ‰GÈOòš, V’Ü?Üüäüà ¡7Õ@\2!ŠÖ|º»ªHcyzÌ@ývÙ3a!ÎX@8¹^èM„‡‘?4$00À_$ô!'?Ék²Äütî°aÑ}dDhD˜˜Ø<@è'0Ï_/àç¯÷³ÌŒCîļè^Öq)…@èçcŽþJ_a>XXØ]Orj/ƒx9ŒZeÚ¬°yÃr`5~xˆ9ÄááDñÁ¡!A–cy]:@ ˆ…ÝõA@ Ë/œ\ï@ bù…Ýõü|Y@ –]Ø]@ b¹†ÅõF,Gàz®\¸ËÑõC%ë®ýÕS¥Zë•O\{í“*í¹-U·W²êºË¾ûËnÿ¼Acý^â£s”µ85.v×W¿úë+VÜôÁ^½u‰¡ñ³?%×<[î/ˆë‡òøŵë 5Œëkª+OŸßE¹ßi*~ôjÜ!Xò®ÿá¯vÓë5Ãæúºwoùéµ—÷—±ëûÒVþð¶ØvzæëÅh·€ë,×ÿ/ï<ñ»ëŸ/×ñ¿W½rãõ¾ã÷ÃkÍ®g«$÷ÿé§ß[±â{Wüñ¾èJs*M¨>ûí÷W|÷²kÏ¿·êj»ë‡[¶¼ øõe+V¬øþ¯ßØÕMyt½®)í9áo~@¶üÞå7¯k6=”5똆îÔ°«ÿüúo?ºüç×ÝzÝO­¯¯üëeùö²Üß©&îêYðÀU×®}}è7þúÊ+®‹ø¢zˆ¯Î^}%ÙjÅe—_~ù5AŠ6ŠjÛúÌÊ«ÿ¿ï¬øÎ÷¾å¼T:DnŽ_9¦Ng¾Qrz \D®÷O-~÷–߬+2jUO_wóÅ[‚-®×<~õe·¾‘ßc0ô¼ùçË®z,OͨóùÅþòaéeèÌ~ú†ï®¸ÚâzmÅ+7þà÷/ìhÕQC)ýêŠàM”[ו¾pÝ÷oX¿¹aˆ¦µ]UˆmÝ–åö˜LoÊ?¼MÞÁçòίíe¹=¾}­ûz<ð“—‡Ç7“…Úþðã¿Æ¶Q³òzòÛ5¿|(£“Ü:PªÒ.òb0wÍ¿´]ͯw~ \L®Ï8ÐðÙŸõ`^gþ#¿þÓGûì°¸~¨àÁŸÿtíneSMî½?ûÙ}yÝ^ùóûò͢î$ßï[\?TüøUWÜ•mõÕ"¹õrÑ–>w®*zäÊŸ¬Îpõ¿»²zÜs>×»=¾c­ûzüèŠ{¬Í, ×ãs`YþG/ÀõFf¨¹xwq“yHÚ[}Þ£W]vû;E½Õ[üΗýâá<5­Î{øê_¬ÝÂ{mxߟ[q•¥¿~¨ì¥~|ë‹™ÍC´‘êªÉÎ(éñÜ_ÙMOg4 Ñ´®«º°²—ï¯wSãö˜^õ×Ï:¾c­Ûzzpýpí×ÿ8`S—eœ¿nßæäü¦à ïß´êgWÞŸ?„±y®,×;aw½‘ÑT|uÏ~ò½+¾û“?ܳ¡\m‡ß_üaèõWýú÷·Þ.|ø¹à+íãðõ-¯]ÿãï®Xñï_ysЋ™Ææi6&?í{í÷W¬Xñ~ró= Ë8|we¹;æ¼®w{|§µîêéÁõF}“|Íõ?øΊ?¸SÚ¢.}]xÝåßã‡÷ÿà7¢W¶wR›àzÀÅîz\€ëp=®Àõ¸×àz\€ë¸ôþ¯ô-¸¸ÞÎZÀEåÿºÆ•ÙKæV9\À¥ìú(ÀELKKËlדÿq¸¸®`y»ÞÓB¸¸®`©»ÞyÎ ¸¸®`YºÞùÿ®®_¸ëc2ö µnÁ5Ö1ÛZ ­;bñA€ âúY:§öp=p½—®—¦ÔšÆ·HK$ ù–eM‡ON;È‘WÃ-;ã¼½P) 5ý9ò¥uq§ÎÒÍ c£õ©RX\×Ïþ—‡ë€ëSRR¼v}ÌŽž)Syb´»|v{µoóB ¯¤tùKÌõóÕ9:¡”ûº/©=8ß®ŸÔ;§öp=p½W®Ý©þÛhYBt”—®—¦Vh&NŽÕ*›µ4EÑ Ëjê2m–”¦U 0£SÓ§Ž°èÞÒd—Ħ•´ë ÃY«k)L–9Û²û :_Ÿ½WðÜøäÁ¾’$‰½B©E­:š¦ÓØ¡ƒÆ-½Ãû힎LŽª6‘ƒg–öP–õ,ÕUº%Öë:ólT™Ž í‚ìÁùw½§ÿz¸¸ÞK×˶¶L}ݺUå­ë qÜ©oz ¥–öþ"ÍXgfŒW9²$©Twp@™"³©»t`´¿ >Ú5Åf§L¦QM¥YÆybBŒ­g¡XÍu$Yë#Qd·NLudȼ9r¼jôÔÉCù–:GI“ {FµªD‰÷mÒÍÍG§<}P,Šë=%õöÔ®®÷Æõ±»‡OªpÛ‚?‡ëõT±Âñk¡žVÆ{ãzÙ¶ŽÑ¾].«ârÕœËX²;7ѱS1«F²ôýÆÎ,—Ä:6%35Ö«#Ç«F¦\nH¢b³»ÙÖ ™÷ýKÇOês؃óëú¹ ××{åú]C?\åh%÷ÖõEŽ¡zò|åëåy†ã“&Æ#7>ÞãÒîi÷™Å,èÈñ*†*ŒóÜE?¿ë£Êþ]‹F|pþÇæÍ\\ï륩û&¿íÚ&;/®—¦ÖzvÊÏl˜œl[;Ûáé‹oó™Ïë»\òz9Éë[Ò½Ïëeí_O6¤a(>8ïmøs××Ïëz¢Ð®oïI–œ;×ËwkF÷gX½›œ·}£íðò-TùV¹­…^–˜UV•“ ñÂõQ’Måã݅›d¶#oQuÓ•µMbî#Ç«L'¦F:mý²ä’Þ±¡Gý\u¶^}ð›îLt׃óìú³Q9\\ï4:ïÈ!×ïŽÏñýúèøœ6úèôôQº=‡ù&M)WOœžžÔÖ¤Ù=(I,ìäŽ7iŠ¢´í¥[²q‰b{u¯ocg(ý°n°«:+ÞjUiZµšáœ†ÄÏ-ßZÖMs<¦‘£º± Í»#›o Š·);´Ša9“‰îRmvi"˜³Îduòžñ#-™àzÀEãúïï{ÁABï\ 4Lôd-ëžhâzmÞ™ç?f{ç!½2A ¸pQ¸žX>:::4$ÐK×óÉyEwö¸ÞS_}úžVU2ºê\¸X\O,ÿï+VD„‹½výòFšZÙÏ=}êÇP}xÐ-€ëKßõÄò_mØpת0¸¸WH–¥ë‰åSRRÖÞ ××ã À²t=±eÒT[§Á›c_ij…fâäX­²YKSÍ°¬¦.Óó3ìùj¦O¼p=\×Ãõ—¢ë‰å‰ëß|ýå9\“ÞÈÒ ¶9_¥ »[XªÆUœñJvÊdÕTnási‰<1aFívv{I¢J˵çX&—(²›L'N•mœÕaïnߨ¨¸îÔ7½…‰RËmA‘f¬Ó}3½,­JK7ïC ®‡ëáz¸þRt=±ü§þõÝ·^óèúèÊaª"ɹ»\šVoÔäƹ¸ž›èØ©ˆöxArçkYF ÓîÜ°.Ko›]€ëõT±Âñk¡žVÆÏîíOVö1]y1Å,€ëáz¸®¿D]O,ÿñ#k?x÷M®—¥7S-®£Øc³PµÎ#ãã•”.®ÄÙ¯åyZƒ“¬-úfJâú"Ç ‡<_GÍt½,­BÃ(NBã=€ëáz¸®¿t]O,Ÿ’’òñïÌ•×ë©ÊM.y}jq0Ï5¯?×ËÒ÷3;bÏ&¯ŸËõ±é{tÆŽ\dô®‡ëáz¸þÒv=±|ôK«>ûä}®'®Íhâ˜Æl[}bnGÕ¦Íè¯?×GI•C\Gn‚¥Ç=~W33y|äœäõÑq; Ts¶½W!&£ºr+¾bàz¸®‡ë/A×Ëßôï?ÚðùGs¸ž7·½¦Ÿá8–e9Žê.w‡/M«V3ÜèÔôé£#,Ý[šl÷²u$=?–þäô±ƒüØá–l‡¡c·Vô2&ÇPú¡NUfÑewý\ûFÇç´ÑG§§Òí9ñ¤.Ò”rõÄééIme\!÷Ïcãæ]Ìp#Sgò1:Àõp=\o4Қ把ýƒ´·Ëáú%ïzbù_ýŸ«£7|6§ëÏŠ"ƒ»ñu¸®_ÃÕ%%ª½}\×óË¿»öVYô—ëe)•:}M Ò×Ãõg£ë¨R–U––Ö÷¸®7‰åSRR䲨 åzIBA‡ž1?¤‡5´æ'ÂôÀõpýÙ©~¨µRµ§{ µBµ§Go•=C vì­T•(Uu-í{ËmN÷´®_V®'–'®ß(—^mø¸®?ÛÞøÁær^ò´–(¿¦kؼR7•©j;)†1 4W*KÊ-N÷´®_^®'–'®OŠ…ë€ëq…\®§ÕeDõÆÈÙ+«:tŒyayYƒÚjqÃ=*‹Ó=-‡ë—™ë‰åSxÙ+àzàz\!—¾ë©}¥J³êÙW¶j#Õ¿GUÙ®clýòMå§{Z×/3×ËÕoNI€ë€ëq…\ò®§úöªJ* Jò²bÿÍ'ûåM«Å©þzG^ïv9\¿Ì\O,O\Ÿ¾9®®Çr‰»žá[áUõ´zƒ…áÁŽê’²f M™ëûøtŸÑuíqê¯w¿®_^®'–'®ßžž××ã ¹´]Ï w×*y³;-Ò÷Ö©Ê(†Òtì­,¯¨ª©©kÜ__æ‡ï~9\¿¬\O,O\Ÿ¹®®Çré÷×/yCáCX ×Ë×ïÎÚ¶¸®Éìùv¢>Er÷•ÄgÕöÑ&Çq##Ãûs1€ëáz¸®¿$]O,O\Ÿ·kÇâº^šZÕ×U 9ûÊÒ*öWïPXöŠÍØǪ’a{×Ãõp=\ ºžXž¸¾(oçòn×$×лb!×Ãõp=\鹞Xž¸^Y˜3§ë%ŠÌê^Úh4?ÊÖÐUæ4Ï]”4­j€:Aç+â³÷j–Ÿ<ØW’dM¢­³Õ™›hœ=/m”<½âͱFš¦ ½•YņãS&Mµe²º¹ö•¦Vh&NŽÕ*›µ4EÑ Ëjê2=Íc'MÌïbû â£!×Ãõp=\鹞Xž¸¾¬¤`×Ǥ7²tC–mþú„Ý-,U“:cþzvÊdÕTnásg‰<1!fF§»ûùëUZ®=g£Är?‘Ýd:qj¬Ìûùë¹SßôZž /I(ÒŒufÆÌjÿ¯:òÿú‡š¶!©p=\×Ãõ—¦ë‰åçq}ôFå0U‘äÜ×-M«7jrã\\ÏMtìTxNœÝùZ–Ñ´owÒ³,½mrt®×SŠǯ…zóáJÛ*ûŒýÊ$ô׸®‡ëáz¸ÞëeéÍTK†Ìåò›}€ªu¯¤tùò9.Hî|-ÏÓœdmÑ7Sº×9n8äù:ʽë-·ÍLǬ´¸®‡ëáz¸Þœ×ë©ÊM.y}jq0Ï5¯?×ËÒ÷3;bÏ&¯÷Úõ’äf`7Úñ\ן×KÁ™‚ó縞¸6£‰c³mýõ‰¹mU›6£¿þ \%ITq¹ –÷ø]ÍÌäñ‘s“×Ë·ïiP¥ÇYª-O¯6µn‡ê\××»u=ßã½½¦Ÿá8–e9Žê.w‡/M«V3ÜèÔôé£#,Ý[šl÷²u$=?–þäô±ƒüØá–l‡¡c·Vô2&ÇPú¡NUfÑewý\ûFÇç´ÑG§§Òí9üðziJ¹zâôô¤¶Æ:„?J–œ¿¯Ÿ6˜È~;Ô“(…\××{rýùBQd =·ÃàzÀÒv½,¥R§¯IAþ \XN®—$tèóCzXCk>ZÚ€ë˹ ×àz\€ë€ëázàz¸¸®à’qýúõ?‡ë€ëáz–µë××ÃõÀõ¸×{$&³çÛ‰úÉùÞ—ÿþ~ñà‘¿ý_Sq¬àz¸þ¼ð¿à,Àù³„û륩U}] ’ó½oTÜî.ý¾¬ƒó,9Àõpý"»Â ¢¹þÊåÚ†›¹_ߺ=vÆŒxÀõp=\×_j®âóÏs—YÝKæGٺʜ湋’¦U 0£S'è|E|ö^ Ãrã“ûJ’$¶6tólu¦ÃÇ&gÏK%O¯8@s¬‘¦iCoeV±áø”ISm¬nŽ}¥©š‰“cµÊf-MQ4òšºÌ“êʶÔwç(fÍ~ \×Ãõpý%æúuëæÉëcÒYº!Ë6}Âî–ªI1=;e2j*·ðsÄKä‰ 13:ÝÝÏ_¯Òrí9%–û‰ì&Ó‰SceÞÏ_Ïú¦·Ðò}IB‘f¬3Ó©TIrÅ`±¹é®p=\×Ãõ—¶ë}ôgs¹>z£r˜ªHrî.—¦Õ5¹q.®ç&:v*¢=^ÜùZ–Ñ´owÒ³,½mrt®×SŠǯ…z§ùp£Šú+’%ö-áz×Ãõp=\麞,™Ëõ²ôfª%CæÚž}€ªu¯¤tùò9.Hî|-ÏÓœdm‘2Sº×;\ž¯£®ÝÙwpŒe,pGNŸžÐT¦b=×Ãõp=\iºþÆ®›'¯×S•›\òúÔ:ã`žk^®—¥ïg:vÄžM^ïÑõsm \×Ãõpý%æú?Üró<ýõMÓ˜më¯OÌmã¨Ú´ýõgàú(I¢rˆëÈM°ô¸Çïjf&”ÂõÀõp=\έëoºéÆùÇáo¯ég8ŽeYŽ£ºËÆáKÓªÕ 7:5}úèËC÷–&Û½lIÏ¥?9}ì v¸%ÛáÝØ­½ŒÉÄ1”~¨S•Y4DÙ]?×¾Ññ9môÑéé£t{N<©‹4¥\=qzzR[“æÚÛ M­ègFÈ–G«ÒІàz¸~ »žj­(±¡ª¨kéfàzpî\¾Ph¹9®‡ë+”µ Œ‘¡†û[«•ªú^\–˜ëe)•:}M ²oàz¸~×[~î¨*©hÓš3{Fß×RW¡²¤û­änÀH«Ë+;Zö”+KJJkšû†‡{›jÊø_j[Õ”µA€êj4/T–V7tÑFFÛVUZ׫·­'G)kRÓn‹€ëáz¯‘$tèͣ嬡5?¦®‡ëçq=CëÕmÕÊÒ†~Šò`k¥ªª¥˜fh]ocEi=Éý‰¥KKT{:4Cwו*Ë«ºµÙ¸£ZUe¹G ÕÍåªêVµa ê¶Uy¹ ÐuT•î9`q9­n(+oÖÐ뗘ëoÀ2®®‡¡àz¸~»Þï÷¿]t×Çdö|;Qï<3Þ¢ï+Ï>vtÄhêŽV×l–Á ®‡ëáz¸þRuýu‹îzijU_WA‚ä<î+ÏÑPe Ñp€ëáz¸®‡ëχë/p=€ëáz¸®‡ë½w½D‘YÝËë’”­¡«Ìiž»(iZÕ3:u‚ÎWÄgïÕ0,7>y°¯$Éš‡[g«3>6Ñ8{^Ú(yzÅšc4Mz+³Š ǧLšjëdusì+M­ÐLœ«U6kiŠ¢–ÕÔeÊ]]¯ûö°‰3qœidÄx 4-Zp=\×Ãõp½{×Ǥ7²tC–mþú„Ý-,U“:cþzvÊdÕTn‰å-OL˜aVó׫´\{ÎF‰å~"»ÉtâÔX™÷ó×s§¾é-´\/g ΟErý7^?O^¯§*7¹äõ©uÆÁ<×¼þ \/KßÏtìˆ=›¼Þk×Ëw`›¶bÂ\×ÃõÀõnûë3š8¦1ÛÖ_Ÿ˜ÛÆQµi3úëÏÀõQ’Dåב›`éqßÕÌL97y}|~¿±·l³åFB²qw»I§J”@ ®‡ë€ëݹž7·½¦Ÿá8–e9Žê.w‡/M«V3ÜèÔôé£#,Ý[šl÷²u$=?–þäô±ƒüØá–l‡¡c·Vô2&ÇPú¡NUfÑewý\ûFÇç´ÑG§§Òí9ñ¤.Ò”rõÄééImMš­·!&µ°Yk5™L#¬ºf[¾~àz¸€KÖõ7\±|¿^Qd˜c|®œ¡ë/’géÈR*uúštª×–“ë% zó#묡5?¦®,Û¼×àz\X°ëõïWÂõÀõp=p=®Àõ¸×àz\¸H\Ÿ ¸è¹X\“ÙóíD}Šä|ï%MܵWÍi†e‡›¶cž;×/Üõÿ ¸ˆ™œœ¼X\/M­êë*Hœß}c·í3Ž«+Òåx>€ëázàúe؆“Ù6ª+MÂäv®‡ë€ëçw½D‘YÝKæGٺʜ湋’¦U 0£S'è|E|ö^ Ãrã“ûJlŽµÎVg:|l¢qö¼´QòôŠ4Çiš6ôVfŽO™4ÕÖÉêæØWšZ¡™89V«lÖÒÅ7Ñkê2]šècwö´¤Ëà×ÃõÀõóº>&½‘¥²ló×'ìna©šÔó׳S&Ó¨¦r ?c¼Dž˜3#Év?½J˵çl”Xî'²›L'N•y?=wê›ÞBËô% Eš±ÎLG©’”Z}Wvò®½ÖÄ2úžÊ ´ä¸®®wëúèÊaªÂ¥)\šVoÔäƹ¸ž›èØ©ðlSw¾–e´0íÛn dém“£ p½ž*V8~-Ô;χ+ÝÚHë†u5ïxIü®Víù®‡ë€ëݺ^–ÞLµd¸¶…Çf jGÆÇ+)]þ\£ÜÝùZž§58ÉÚ¢o¦t!®/rÜpÈóu”“ë%IUãG»²ìu’çk’!{×ÃõÀõîòz=U¹É%¯O­3æ¹æõgàzYú~¦cGìÙäõž]OÖ4ܸÕ~»«Ÿ©K…ë\××»ë¯Ïhâ˜Æl[}bnGÕ¦Íè¯?×GI•C\Gn‚¥Ç=~W33y|äåõ䀛ʇj·˜ûdiUÃôž4)Ìàz¸¸ÞÃ8üí5ý DZ,ËqTw¹Ó8|iZµšáF§¦OayèÞÒd»S­#éù±ô'§äÀ·d; »µ¢—1™8†Òuª2‹†(»ëçÚ7:>§>:=}”nω'u‘¦”«'NOOjkÒœ{äÛö¨ mä8C{aLàz¸¸þB¿^Qd ]rs\XF®—¥Têô5)H¿€ëËÉõ’„‚=c~HkhÍO„é€ë˸ ×àz\€ë€ëázàz¸¸®®‡ë€ëázàz\€ëp=`Ù»>&³çÛ‰úÉyÛ7®Àxb’cì˜&O0s>°¸®®?¤©U}] ’ó¶¯lkeuFŒã×mmL댉y€ëázàúeÓ†/ÛÚÄtíˆ\××{œç.³º—6Ͳ5t•9Ís%M«`F§NÐùŠø콆åÆ'ö•$Yópëlu¦ÃÇ&gÏK%O¯8@s¬‘¦iCoeV±áø”ISm¬nŽ}¥©š‰“cµÊf-MQ4òšºLOmôÒ´½LïÎXhÀõp=p½‡ùëÓYº!Ë6}Âî–ªI1=;e2j*·ðF•Èf$Ñæ¯Wi¹öœËýDv“éÄ©±2ïç¯çN}Ó[hy‚¾$¡H3Ö™é6u—$×Ð9è¬p=\\ïÞõÑ•ÃTE’Ä%O®7jrã\\ÏMtìTD{¼ ¹óµ,£…ißî¤gYzÛäè\¯§ŠŽ_ õîçÕ$U† €ëázàz÷®—¥7S-3†µÅf jGÆÇ+)ÝœÃÜÝùZž§5+fè›)]ˆë‹7ò|åÎõÑ ªa]ÉF8Àõp=p½ç¼^OUnrÉëS댃y®yý¸^–¾ŸéØ{6y½®Wjõe‰Ñp€ëázàzÏýõMÓ˜më¯OÌmã¨Ú´ýõgàú(I¢rˆëÈM°ô¸Çïjf&œã¼>.OMUm’@ ®‡ë€ë燿½¦Ÿá8–e9Žê.w‡/M«V3ÜèÔôé£#,Ý[šl÷²u$=?–þäô±ƒüØá–l‡¡c·Vô2&ÇPú¡NUfÑewý\ûFÇç´ÑG§§Òí9ñ¤.Ò”rõÄééImMškoƒl[Se2Tàz¸¸þâø~½¢Èà~|®,×ËR*uúš).ßÀõ€eäzIBA‡ÞüÈz#khÍO„é€ë˸ ×àz\€ë€ëázàz¸¸®®‡ë€ëázàz\€ëp=`Ù»>&³çÛ‰úÉùÜWš˜¿˜ãìkb÷íŒÇƒr\××/žë¥©U}] ’ó·¯4­žÑWoŽ1ÿ"Ïl4êÊ’`{×ÃõÀõ˧ _QD1*û£õ%){˜Ý±€ëázàzOóÜeV÷ÒF£ùQ¶†®2§yîH]5ÀŒN óñÙ{5 ËOì+±%ÑÖÙêL‡M4Ξ—6Jž^q€æX#MÓ†ÞʬbÃñ)“¦Ú:YÝûJS+4'Çj•ÍZš¢h†e5u™Î“êJ’”ýÚËD¼Ò¤¼¶áî¼xLbàz¸¸Þýüõé,Ýe›¿>aw KդΘ¿ž2™F5•[øÜY"OLˆ™Ñéî~þz•–kÏÙ(±ÜOd7™Nœ+ó~þzîÔ7½…–'èKŠ4c™.¥’z>uòÈÁ©üÓ4£¾Àõp=p½ÝõÑ•ÃT…K_·4­Þ¨Ésq=7ѱSá9qvçkYF Ó¾ÝIϲô¶ÉѸ^O+¿êçÃ%òï£ÚóÌw²”’º+G¼Àõp=p½×ËÒ›©– ™Ëå%6ûUë<2>^Iéòås\ÜùZž§58ÉÚ¢o¦t!®/rÜpÈóu”Ò©ƒ¾–:°ÓÑA¯(Ô—&@ö®‡ë€ëÝæõzªr“K^ŸZgÌsÍëÏÀõ²ôýLǎسÉë=º^º¹îÚáh3ç ÐU›0Àõp=p½ÛþúŒ&Ži̶õ×'æ¶qTmÚŒþú3p}”$Q9Äuä&XzÜãw53“ÇGÎM^%M«£ {ÒÍwÑq;šºr¨Àõp=p½çqøÛkúþ±4,ÇQÝåNãð¥iÕj†š>}t„å¡{K“¥Ž^sóHz~,ýÉécù°Ã-ÙCÇn­èeL&Ž¡ôCªÌ¢!Êîú¹öŽÏi£NO¥ÛsøáõÒ”rõÄééImMšÌQvN“––gb5õYx–€ëázàú ÿýzE‘Áy|®,+×ËR*uúš|9¸°œ\/I(èÐ3æ‡ô°†ÖüD˜¸°ŒÛðp=®Àõ¸¸®®‡ë€ëázàz¸¸®®Àõ¸×–륛”=ôĉÑÙ­_,b2{¾¨O9ï¸]œre©ªÿ0ŽÓ·&Ëà*×.Ƽ^ž§¥TçÆõŠBMŽ|ÎM¤©U}] çÖ¹¦\IB±†mßmy4¿,¥LödÆÀV®,k×Ï;AÞ"qaÊ•m.-KwLâ+ÝÒ@wn‡ì\¸\“ZÒa`Y#mÐišó·º¸>6­¤]o4?åÖÄêZf4LKóZ©CGr,gÔõ 2mÖTVšV5À8M7cŠ¼(ûLw¦ÃÇ&gÌi+M­ÐLœ«U6kiŠ¢–ÕÔeÊ]úJ:uÊ6šFØÁ†öyîΪ\ójEfu/m4šìkè*sÌú7o­f4.ä«™úT<Àõ€ ízIB‘ÚÔS´Ébp)Q7}èo&›ë%I¥ºƒÊ«ßcRKFû âísÞÊsu=»l“ØÆíìâ\2Yoòkó×s§¾é-´'®'w>TK†ëèùØìT­m¼þ\µ²gôö@qï\X y½4µÎгӳ4£cbœÓ7 wl“-²ë¥[™®NMç³”{y½žªÜä’קÖó⢼s}lú±#=€ëSýÆ"õÈbëˆ;iRAçøñ8¾_/ßÑBõ—;Ƨɳʪrl_U‹WrGÚ³âl+c·ÖscÍ[®—ïÖŒîψ±Ý$gçmŸ¥Á…»>:±TOפH­#ív7³'ÿ5î’ןy¹1MÓ˜më¯OÌmã¨Ú4©7­ Ñq; Ts¶"Ú~¬êÊ­øŠ=€ëÚõQ–qø”Éı Mw§Q'¦LššÍMIÛ«{ |ë=Cé‡uƒ]ÕYöAA}[Û 7:BöæFFômE.CÞ£$‰…Ü‘#ã&#MQ”¶½t«­ Á:žrúØAŽ9Ü’möht|N}tzú(ݞÔ¦”«'NOOjkÒd¶ñ#_¤it„î.ß©bHÕÕi²³+×ö~kúŽ_ÁqT·ã>g¾ZÅrÿ<6α6¸‘©¿3ùàzÀEàz\€ëp=®Àõ¸×àz\€ë€ëázàz¸¸®®‡ë€ëp=®Àõ¸°D\/ݤì¡'N8ž‡¿èÄdö|;QŸrÞ'‹Y”r%ñYµ}´ÉÄqÜÈÈðþÜDÌàzÀŘ×Ëó´”êܸ^Q¨éÏ™û‰ðÒÔª¾®‚„s+Å T®,­bõ…嘱ûCU2làzÀ²v½7sË.ª\×$?¹†Ø ]¸pá]ÏÏsg`Y#mÐišó·º¸>6­¤]oäç¹cM¬®¥0Ùe–VIb^+uèÈ¡QŽåŒº¾A¦-Ó2•¬4­j€š>}tÄ:ïÝ[ê2 žuÆ9Óác3æ–•¦Vh&NŽÕ*›µ4EÑ Ëjê2å.} %ƒºaEM#ì`ÃNûÔ{gU®yµ"³º—6òï˜1ºÊóùÎ[+—·˜ßÅöÄGCW®\`×KŠÔ¦ž¢MÖÉb‰ºéC3Ù\/I*ÕP¦Xý“Z:0ê,0y®n¢g—mBظ]\çö˜…å×ç¯çN}Ó[˜(µè¹H3Ö™ã¼A´S]¥ïuI Ï¼Ü˜ôF–n°ÍÚ+MØÝÂR5©R¯ke¾%¨:òÿú‡š¶!©p=à»^–ÞÌtfź Z뜲͓mëíÛå"͸\5׸Ej÷Za¤¯&/ssRœôÌÚÒ=º^O+¿êie¼Côq[•-ƒ$û6Ò½vÐøµIÊÞ¨¦*’œûØ¥iõFMnœWµriØVÙgìW&¡¿Àõ€ ìzyžÖP¢pµl¡±º^žg8>ibœ1rãã=ÙN7ѱIÛòÊ÷uÒœ‰é¯Ý&?w®/Šs "È×Qv«Ê6×3¦Î¢´Øhkc„Š¢Ï‰ëÉÕ’áÒI›}€ªµ×Ÿ«V³‡ê‘û¨ŽYi?8÷|e®‡ë€ëœ×KSë =;=K3:61!Æ©1}cÑÐxÇ6Ù"»^º¥‘éÚáäÐYÊ=‹¼^OUnrÉëS댃yqQgàzIr 3°íø‹Ï† D"ÑgŸ}¶üt×ן“þúEê‘ÅÖwÒ¤‚Îñãÿp|¿^¾£…ê/wŒO“%f•Uåؾª¯äŽ´gÅÙVÆn­çÆš·:\/ß­ÝŸc»1HÎÎÛ¾QrÖ®N,ÕÓ5)RëH»ÝÍìÉ»äõg^nLFÇ4fÛúësÛ8ª6MêMkC”|ûžUzœe×hyzÍ°©u;TøüóÏW¬X±råÊ>úh™é®®?‡ãð)“‰cšî(N/¢NL™45›eÖžçíÕ½¾õž¡ôúÁ®ê,û wEÁ ¾­m!{s##ú¶"—!ïQ’ÄÂNîÈ‘q“‘¦(JÛ^ºÕæ>ëhx~<üÉéc9þåpK¶Ù£Ññ9môÑéé£t{?PšR®ž8==©­I³Þ’$+{Œ|‘¦Ñº»|§Š!5VW§Éή\Ûû­ég8~ÇQÝŽûœyk%KÎß×O›øƒåØ¡†œD)<|> =qýÕW_íããóñÇ/'ÝÃõÀõxF.v×›L¦ŸÿüçB¡ð“O>Y6º‡ë€ë½qýoàzpi¸þþç(ŠºâŠ+‚ƒƒ—îázàzo\ÿ?€ëÁ%áúÿüÏÿ$'ÿÐÐÐå—_ö駟.ÝÃõÀõÞ¸þé ׃KÁõÓÓÓ$µ'çooï~ô£U«V-ÝÃõÀõÞ¸>æ‘Ûázp)¸þŸÿü§%µ'tttݯY³f©®®÷ÆõÉÒõp=¸\ÿ÷¿ÿý_ÿú—ý¤±±ñ‡?üá}÷Ý·¤u××{ãú¨d)\.×;vÌ9µ'ÔÖÖ’ìþÁ\ºº‡ë€ë½q}JJ \.×ýõ×'OžtNí J¥’èþÑGýüóÏ—¢îázàzo\þœ®—‚ëÉÄìÔž““óãÿxݺu_|ñÅ’Ó=\\ïëoZù\.×>|ø›o¾±¤öÿýßÿ}ë­·^fã;ßùÎ~ðƒçŸ~Æ p=\Àòsýo®½g®—nRöÐ'ÏÃ_tb2{¾¨O9ïS¿.f¹’„âÁ#û¿¦â8hø<º~llŒüSÐ4••õ_ÿõ_ÉÉÉ×\sÍ]wÝõÀ<øàƒ<òÈË/¿LR{¸®`ù¹~áÏÈ•çi)Õ¹q½¢PÓŸ3÷|sÒÔª¾®‚„sëÜ U®…¸Ý]ú}Y%çYrÀâ»Þd2õ÷÷ÿêW¿úÑ~ôí·ßž:uêŠ+®X³fÍûï¿ÿÉ'Ÿ|úé§K±Ë®®¿è]ïÍܲ‹Á…*—'6s¿¾u{ìŒñÀ⻾µµ•$ò·ÜrËoû[ât’Ú“U7Þxã{ｇïÜÁõ\ò®çç¹3°¬‘6è´Íù[ ]\›VÒ®7òóܱ&V×R˜,si®NÌk¥94ʱœQ×7È´eZ¦’•¦U 0£SÓ§Ž°èÞR—Ið¬3Ι›hœ1·¬4µB3qr¬VÙ¬¥)ŠfXVS—)wék(éÔ (ÚhavÚ§Þ;«rÍ«™Õ½´‘ÇŒÑÐUæ˜ÏwÞZd[ꇻs³f¿‹ïú«®ºêOúÓ“O>ùøã_~ùåÇŽ;qâÉñɯ$£‡ëáz.a×KŠÔ¦ž¢MÖÉb‰ºéC3Ù\/I*ÕP¦Xý“Z:0Ú_moÈÕMôì²)-ng×¹=faùµÇùë¹SßôZf…%•ÔŒufÆ8oí°sr•¾wWlÔ¹(7&½‘¥l³öJv·°TMªÔËZI’+û‹Ípýywýí·ßþôÓOôÑGŸ~úéM7ÝMR{¡™¥;Ñ-\\\/Kof:³b]­uNÙÆæɶuŒöír‘f\®škÜ"µ§º5†‘¾š¼ÌÍIqÒ3kK÷èz=U¬püZ¨§•ñÑÇmU¶ ’ìÛHôÚAã×&Uü9(7z£r˜ªHrîÄ—¦Õ5¹qÞÔ*:¡¨o°"Ybß®?O´],¿ø⋧oØ°Hÿ‡?üapp0IðW¯^ ×Ãõ\Ò®—çi % WË«ëåy†ã“&Æ#7>Þ“íts›´-¯|_ç Í™˜þÚmòsçz'WÊóu”ݪ²ÍõŒ©³(-6ÚÚ¡¢èsâzrçCµdÈ\ûß³Pµ¶ñúsÕ**vgßÁ1Öú9qGNŸžÐT¦JaâŇÈäòÎÇûâ‹/Ö­[çççGNò7ÞxcÉ ¿‡ëç/¯—¦Özvz–ftlbBŒScúÆ¢¡ñŽm²Ev½tK#ӵééÜU¹g—×ë©ÊM.y}jq0/΋;9듯l8/!~ÿØÌ—_~‰±yp=—xýÆ"õÈbëˆ;iRAçøñ8¾_/ßÑBõ—;Ƨɳʪrl_U‹WrGÚ³âl+c·ÖscÍ[®—ïÖŒîψ±Ý$gçmß(9k×G'–êéš©u¤Ýîföä¿Æ]òú3/7&£‰c³mýõ‰¹mU›&‚ë\X²®²ŒÃ§L&Žehz¸£8½ˆ:1eÒÔl¶8[¢Ø^Ýkà[ïJ?¬ìªÎ²zW êÛڹѲ772¢o+rò%I,ìäŽ7iŠ¢´í¥[mMÖÑðüxø“ÓÇrüËá–l³£ãsÚè£ÓÓGéö~ 4¥\=qzzR[“f½%IVöù"M£#twùNCj¬®N“]¹¶÷[ÓÏpü Ž£º÷9óÖÊÑPÑÏŒ- V¥¡ Àõ€‹Àõ¸°„]ðR\\×°Œ]ÿŒsÚ×Ãõ,gו’××Ãõ,c×'oz®®‡ëXή' ®®‡ëXÆ®ÿÀÿz¸¸®`»>ìg¿€ë€ëáz–±ëoÄ÷ë€ëáz–µëñ,àz¸¸Þt“²‡ž8áxþ¢“ÙóíD}Šä|_C¥\yÎð±£#FëTw´ÚöœaàzÀE•×Ëó´”êܸ^Q¨éÏ™{¾9ijU_WA¹uî…*Wž£¡Ê¢a(×.×{3·ìbp¡Ê…ë\¸8]ÏÏsg`Y#mÐišó·º¸>6­¤]Ï7KY«k)Lvi––$æµR‡ŽåXΨëdÚ2-SÉJÓª˜Ñ©éÓGGX to©Ë$xÖçL‡M4Θ[VšZ¡™89V«lÖÒE3,«©Ë”»ô5”t ê† m4°ƒ ;íSïU¹æÕŠÌê^Úhnˆ7ºÊóùÎ[+âzÝ·‡Mœ‰ãL##Æ¥i1€ëÜõ’„"µ©§h“u²X¢núÐßL6×K’Ju”)V¿Ç¤–ŒöÄÛ3Wy®n¢g—mBظ]\çö˜…å×ç¯çN}Ó[˜(µè¹H3Ö™ébNGö,I®Ò÷îŠ:åƤ7²tƒmÖ^iÂî–ªI•zY+Iü¶]¥ÖÛ•†kÙÛ¸p]/Kof:³b]­uNÙÆæɶuŒöír‘f\®škÜ"µ§º5†‘¾š¼ÌÍIqÒ3kK÷èz=U¬püZ¨§•ñÑÇmU¶ ’ìÛHôÚAã×&Uü9(7z£r˜ªHrîÄ—¦Õ5¹q^Õj怴½tOdàzÀ…u½@ÕÚÆëÏU«ÙlTQÚ<9|àzÀâº>åËÏ<¯—¦Özvz¶UtlbBŒScúÆ¢¡ñŽm²Ev½tK#ÓµÃ)_ž¥Ü³ÈëõTå&—¼>µÎ8˜u®—ï:À6m•ÂW®\P×ó~V(¶Ž¸“&tŽÿ‡ãûõò-T¹c|š,1«¬*ÇöUµx%w¤=+ζ2vk=7Ö¼ÕázùnÍèþŒÛArvÞö’³v}tb©ž®I‘ZGÚínfOþkÜ%¯?órc2š8¦1ÛÖ_Ÿ˜ÛÆQµiRoZ¢âóû½e›c­÷=»ÛM:U¢ºp=à»>Ê2Ÿ2™8–¡éáŽâô"êÄ”Ic{ŒD±½º×À·Þ3”~X7ØUeô®(Ô·µ r£#dondDßVä2ä=J’XØÉ92n2ÒEiÛK·Úš¬£áùñð'§äø—Ã-ÙfFÇç´ÑG§§Òí9ü8@iJ¹zâôô¤¶&ÍzK’¬ì1òEšFGèîò*†ÔX]&;»rmï·¦ŸáøGu;îsæ­ÿI6kMc£&“i„U×l‹Ã×ï\¸(\€ëp=®Àõ¸p!\Ÿ’××Ãõ,g×Kž‚ë€ëázІ€ëp=®\„®_øœ¶¸×àz\€ëŸë¥›”=ôÄ Çóð˜Ìžo'êSÎûcã±\i⮽jÆH3,;Ü´óܸpÑåõò<-¥:7®Wjúsæ–4µª¯« áÜ:÷B•KˆÝ¶Ï8®®H—ãQø®\ ®÷fnÙÅàB•“Ù6ª+MÂäv®\\®çç¹3°¬‘6è´Íù[ ]\›VÒ®7òóܱ&V×R˜,s¾IóZ©CGr,gÔõ 2m™–©d¥iUÌèÔôé£#¬º·Ôe<ëŒs¦ÃÇ&gÌ-+M­ÐLœ«U6kiŠâÃ5u™r—¾†’ŽAÝ°¢¦v°a§}ê½³*×¼Z‘YÝKùwÌ ]eŽù|ç­UTìÎþ‘–t àzÀÅäzIB‘ÚÔS´É:Y,Q7ýÿ·w&àm”ç¾7IÉ1!d$4@ š–ÊR–÷Pzá8´ ´å´ÐÓBK íáôRιmé±4Zm˶,ɶly_'vy‘íxÑ.ÍŒ4cÇIlÇ’Ë…pŸöy¸’%Y[’åݱÿ¿ÇOžXÖèûf>{~ó~óÎ÷žùˆõ¹žŸPi9= ‘{õ%L®ê/‰œŸYFzò}a%y]Œ.[8»ø:dýzæÓzKe„GÏjÃ)]–0ð þ9r~R­7_³í Ó[œT³¯j/_ u’ ÉDd½âËm]ʤü£×emë©ÍÀL>€ëËîzAz+­Ëq¦¡uã¾ÜGš’"=Ó’JL†º öÁ¾UVJ‚„˜Û\zH×ÛÈ2©ÿÛR¥‰õ‹^’¦Ñ]Ñ·ƒ²ÛÌFÇy¶"vÚåÅi¬d g žP49 E’ˆzE¤µP­¯w;ž›ßæÄ|>€ëËîz±Êl/—r-[j§½®«ìÇX:3<Ü£ ¸8à‰2UÕÇtFŠaéþÆLñ¹^-ñ'[ÈI« RšhV§VˆxÞɈ ’Z×»®|Hmw^¤É²Œ“¦(kgYºš¼4ÎR<ÎæK³ë{íîÙ{š´Y-Æ®úÜɤwi‰ÑÖÞnd†][3ƒƒ¶v5'å=†/+Õ1££Ã¬ƒ"IÒÜQ™æ›Bðfûóá/_¹pšqÿתUNx”[ØN»råÕQèÎ$äÕú‘Ï®Œ™ÞK’$MÃÝ$;4HuWçUЮëë‚ùµëÛ߆~šqÿ€aÈnÿuÎŒ½òú?ó°Þý|ÃØ;J`z°`®'Â××c\®j×7Ì\\××ÃõÀõ¸×àz\€ëp=®Àõ¸¸®®‡ë€ëázàz¸¸×ÖŠë‰DM5rÉ¿þ¢#Ìêùp¤I¾äb¡]I‰ãÒã/ ÈŽ]¢gªÀ\X†¸^¬2“ ãzi©¡¿0¼íˆäº¾®’ø…uýò´+H«­Ïð—ßd¶ÓmS ä×V™ë#©-»,W»\óçVÞ®,›ëÝuîìN§ƒ²[Ì­Å©¥׋å6‡»Î“uZ´¥IœH•/Sµ‘gFÏ 1NÆaé3ÒíY½ŠºzhüÊgç¨ÞJNØ× ÊJIs›KézY&õ[j£4±~ÑKÒ4Z£+úvPv›Ùè8ÏVÄ.@»¼8•¬Iàsâó&‡¡HQ¯TŸPc7•H¡*×–ÀõbyB8׋Uf{9×I’R;íu½Xe¿8ÆÒ8˜ááeÀÅO”©ª>¦3R K÷7fŠÎõj‰?‰ ØBNZUÒD³:µBÄóNFTÔ‚¸ÞuåCj§¤Ó‰”'ÉF_¾~¸^^3ÄWX-åq0€ëKáú§ÞNœ{\O$±÷ä…–&O$‹L¦Ç©MÙ‚Ev=‘ÚÂÍz›¦ÜyÄõ6²6‘×'qU’˜Y¹^Zj¶UÉx0€ëKáz¹\þ~}œZ?x²Ì›qG$”è†/~â¾^œ£%û«ýùiYnU]¡ïQµX 3Ú‘+ñýP”ÖÄœjMó»^\`:á{ 'JRª²ãøóv=OVi£ä„7Ó® ÕyùóaN\?÷v…ǺEé»_/+jgÈF3+×KTz².‘Q¸°$®¿wß=‘äá“,Ë8iŠ²v–¥«ÉK㬡!Åãl¾4»¾×îž½§I›ÕbìªÏLz—–mííFfhе538hkWsRÞcø²R3::Ì:(’$Í•i¾)o6¼;þò• §÷­Zå„Gy±…íÔ¹+WÎQ…î<@B^­ùìʘ¹Aá½$IÒô8ÜM²CƒTwu^íê±¾^!˜_»¾ýmè§÷†ìö_çÌØ«É;™Çûk“ z×–ÈõX#¸®`u»þÙï<××Ãõ¬b×Kßÿ\\×€9|\€ëp=®Àõ¸×àz\€ë€ëázàz¸¸®®ç@$jz¨‘KþõðaVχ#Mò%_QvqÚ%dÅ'¬Œ{]Öi<–‹…r\X‰q½Xe&+ÆõÒRCaøzsDr]_WIüÂJq™Ú%M´­>ÅSsGœÕâ°T%Àö®¬j×GR[v1X¦v¥j’ö×ÛãËÓ"è Àõ€Ev½è¿^‹¤ÎÝétPv‹y µ8µ”ãz‘¢¼Ãæp×¹s²N‹¶4‰SÔ/Sµ‘gFÏ 1NÆaé3ÒíYBoŒ[7@_ùìÜ ÓÕ[É)‚ç­8Çž½0Ò2¥¶,‘\c¹|ªQÓj¦H’¢NÑ,1ç^Cy§Ñbµ“”ƒt›ý³åójwâÇÒ¬ú^ÊáÞcÚaïªò×ó±WüM¿¹ÙS—HPµ[»U±(bàzÀb»þÙ' _¿>^­g{Ô‰Þb±.uSg>b}®ç'TZNhä^¿ “+†úKüYFzò}a%y]Œ.[8»ø:dýzæÓzKe„GÏjÃ)]–0ð ~‰ò“êl½ù¢˜…hW˜Þ⤚}U{‰ø­“lH&"î•kû³Ÿ^==þÉÿe¶®,Û¾ ½•ÖåŠ8Ikºq_nž ³s¨/Ÿ#MI‘žiI%&CÝû`_ƒ*+%ABÌm.=¤ëmd™Ôÿm©ÒÄúE/IÓh®èÛAÙmf£ã<ëŸ;ŸG»¼8•¬áÜc'MC‘$¢^¹äßGv¨&®òòª«PŠ¸Àõ€Åvýú^¬2ÛË¥\Ë–Úi¯ëÅ*ûÅ1–ÄÁ ÷(.x¢„LUõ1‘bXº¿1S¼p®WKüIÅrÒª‚”&šÕ©"žw2¢‚¤Äõ®+R›Á¹I#Rž$}ùúáz×7’'óü‡FZb´UÆCö®,²ëeïþxîq=‘|ÄÞ“Zš<‘,^0™§6 wf ÙõDj Ý•0uÎUîüâzY›È‰ë“8Œ*IW 1DJ3Ø/qáU—ˆD|×Ùõ’ËÃ߯SëO–y3ÝðÅOüÏ׋s´dµ??M Ë­ª+ô=ª«aF;r%¾ŠÒš˜S­i~׋ C'2„¾ ƒ$¥*;Ž?o×ód•6ªANxçÍ Z—?æÄõsoW˜qœ¡[”¾ûõ²¢v†lT‘Ì6¸çûPöÃé"Ï]†œV‡¥ªp=`Ñ]/’'F’‡O²,ã¤)ÊÚY–®&/³††³ùÒìú^»{öž&mV‹±«>w2éÝ=KÝÞnd†][3ƒƒ¶v5'å=†/+Õ1££Ã¬ƒ"IÒÜQ™&š¼·=‘ ï·¿|åÂiÆý_«V9áQ^la;uîÊ•sTG¡;WëG>»2fnPx/I’4=w“ìÐ Õ]WA»z¬¯Wæ×®oúi÷r8N†!»ý×93öjâÓ ›Ýké0¬ÓД‹µt\X×ÿþÅÇ°F.p=\À*v=Öî‡ë€ëp=®Àõ¸×àz\€ëp=®®‡ë€ëázàz¸¸çR°Fᮇë€ëfõ|8Ò$_¼“£0㘩-[¸ 9>/úþ†…HÔôP#—üU åÚßEiW˜ÙfoËÿ!_’ZÜÐatœ½0ÒšJÀõp=kÁõ„¢Î8vå³sƒîeáš²÷Öeˆ§—a%’ëúºJâË„¼‘nYàSï|ú¼Èûb•™¬X×KK ý…â¥ßßåj×]pðÔPSrÐ_'qFe}yNblv;Ù×Ãõ¬™¸^ZFÚJ|¥_øÒœfÚq4U°´Ñ]NÏ8[-C¥÷Åq}$~ƒåj7&†_ÉœïSŠB¿C ×Ãõ¬Y×OÄØ ¤Eﯷ.{öÂHËôʳDby§Ñbµ“”ƒt›ó¸•ÝÄé5')Æé (ÊÞ[›[f¿8ÎÊÑyåé?ªŠŸ¢z¾4«¾—r8hw‰={WUZÀt¡¨ ‡Æ/QÅÒXåQíd†ÇN÷•'ð#ésø^…Ù–H®1Œ\>Õ¨i5S$IÑN§áHÖ‚¹Ì]mÐÕ°ƒ²[Ì­Å©¥׋å6÷±p8Y§E[šÄ9†|™ª<3zfˆq2KŸ‘nϨə§“ê­ä”"œÏþ†ýyµ~ô#…¸ ö‚)_×ÃõÀõA]ÃO¨% ¦œ&CT™wÇPþ “êl½þó+_Vaf: =…ãùRåqöÒ§§ª¦‡ª‚4íøù¶4î€0½ÅI5ûŠÂñZ'ÙÀ•Õ8ÇYvÈP›ên’/–Å #ès„½ ±¿’æÓzKe„GWjÃ)]ÖB$¸>JÏö¨½%{]ê¦Î|Äú\ÏO¨´œÐȾ«‚Ê¡þ’ØÉ#/.²ŒôäûÆP’×ÅèS"‰¯çº¿!G^íÎ0úŒ‘Òzn|ê/\××ûï³N{%¤ x’4Ö芿\á¨Íltœg+b}gÓ -ݨAzûØP׋ ¬—ÏÔpfðyq+Y“8G@(š†" G%ÌHgž”7+DګeRÿ·¥6J;ÿº ½•ÖåŠ8Ikºq_nž ³s¨/Ÿ#MI‘žñ«ŠHn°ö5¨²R$ÄÜæÒç´¿aF>íÎ8ú‘ŒB\åðe[¡®‡ë€ëç× RšhV§VˆxÞp´‚¤&Ïöb•ÙpBöœ¢é ©å¢|ÓÇgë8§v—ûHm7()O’Û3ª$XŸ#íU÷©ýK\l!Âõî^•K¹-•Ú}Ý«ìÇX:3<Üx?š'JÈTUÓ)†¥û3Å çúPûvôçÓÁ(ðâ«Ol·ëázàúHîׇ9'©-tWN@ŒÌUAú º3ðɧPq=‘|lìîL7²³‘µ‰œÈ.ùˆÃ¨’ÄÌÓõ‘öj®®çIsÛ:;Z+³Io ×»vÝÞ“zOy"Y¼0`2=Nmî 8ž‹ãúð£?ϸ>üèGàzAFÇù±f×ÃõÀõÓóðã Úç±éyøAÏɲJÕ '¼wœ Z—?öGv|™ÆÄtÅ{îªÆæ·Òcƒ=2.Èìúàìá$NVŸ0ã8C·(}wleEí Ù¨ bæëúH{5W×Ǫûâ‹Ë½ÊÙåì¹ü¬§Êãgÿü ;ŸtõÊIS”µ³,]MºŽ¥¡!Eà;õ½v÷콫×V‹±«>w2é]Zb´µ·™¡A÷P ÚÚÕœ”w×õM©ŽfI’æŽÊ4ÑBìïL£?ÇvÃ~$£ÃO:<<ª –™Ç/žh˜=ûñ•‹§Y×ÿlÚ< \XíqýÒ!UÛCd²ñâ5ö‘ž\ÑrœLC÷jöˆ”úOþ>rTqÄ‹«a¶îŒMÏ_ û××_U®Èk-6ßœïô›öòšîæLá’ŸJÃ÷j–ð凇۲D°í²"H?ÜV‘´J®·àzàú•ïz~|I§Í“9î´·ËVÆ xeö ¸pÕÏáàz\€ëp=p=\\××ÃõÀõp=p=\\€ë+Þõr¹®®‡ëXÅ®¿wß= àzaVχ#MòÅ[‰L˜qÌÔ–½ kȩ́ϋ¾¿a!5=ÔÈ¥¡  /Îá_¦ý]”v…™mö¶œÐ%î’+zìî¥ÆÖ^š$€ë«}Ÿ»¾ƒ¦ì½uâ鋹Éu}]%‹¶ê(!od‡ºÉ|ú¼Èûb•™¬X×KK ý…â¥ßßåj7†Hi>54µ¾‚~|™ÁÙQà©% WœÚ,!\Xõ÷ë¹uî¤9Í´ãhêÒ;Âœžq¶ZÆ‹‹áúHjË.ËÕ®»¾B%s¾O$´¤TV¥û_'R›)]¶®¬)×ÇL«_ï­HÆž½0Ò2½æ)‘XÞi´Xí$å`Ææ¼XNˆ&N¯9I1NEQöÞÚÜ2ûÅqÖP?µ´»(OÿÑPÕÔÒp|iV}/åpL,ekïª ¨sC(êè¡ñKT±4VyÔ@;™á±Ó}å üHú¾Wa¶%’k #—O5jZÍIR´Ói8’µ`.s×¹s5ì ìó@kqj)Çõ"Ey‡Í1±ª/ë´h§L=óeª6òÌè™!ÆÉ8,}FºÝ¬zT@%CNIÀyîoèÑŸW»áG?²Qˆ«`/˜òg¬L -ÖÓÁ'àzÀªv} ?¡–(˜rš QßÜCù7Lª³õúϯ|Y…™é(ô”,çK•ÇÙKŸžªšª Ò´ãçÛ¦ ¦·8©f_ÝV"¾@ë$’§Ô¯wŽ³ì¡6ÕÝ$_,‹FÐç{²ž;óé½¥žôùñjÃ)Ý‚L»>JÏö¨½Åb]ê¦Î|Äú\ÏO¨´œÐȾ«‚Ê¡þ’ØÉ#/.²ŒôäûÆP’×ÅpbÕHâë¹îoÈÑŸW»3Œ~£@¤´žoK ;?%PÔ™©ÖqL \¸Ê]ÿøýOÏÒõA^ éž$M£5ºâ/W8j3çÙ _yXA†–îTŽ ½}l(ˆëEÖËgj83ø¼8•¬Iœ# MC‘„£f¤3OÊ›•G"íUH÷ÙÈ2©ÿÛRÛ‚Ôä·Ò:NI_a–nÜ—›'ÈìêËç(IR¤güù Drƒ}°¯A••’ !æ6—>§ý 3úóiwÆÑdâ*‡/Û CöD’¦îRÅ­ô·p=p}DÏÜ%Š-®¤4ѬN­ñ¼áhIMžíÅ*³=à„ì9EÓARËEù¦ÏÖqNí.÷‘Ú nP&Rž$3¶gTI°>GÚ«0îSû–¸ØB.„ëݽ*—r[*µûº%VÙ/Ž±t fx¸'ð~4O”©ª>¦3R K÷7fŠÎõ¡ö7ìèϧÝG?‚QàÅWŸþØj_ ¨18O–%\ázàú¥¸_æœL¤¶Ð]912Wé'èÎÀ'ŸBÅõDò±±»2ÜÈÎFÖ&r"»ä#£J3O×GÚ«¹ºž'Íilëìh­PÌ&½1|\ïÚu{O^è=å‰dñ€Éô8µi¸3àx.ŽëÃþ<ãúð£ëçÇšÁd.J?lqt­øˆ®,–ëùqmŒóØô<ü çdY¥jÞ;έÎËŸû#;¾Lcb:‹â=wUcó[鱋ƒÁdv}pöpçÜ+Ì8ÎÐ-Jß[YQ;C6*ˆ˜ùº>Ò^ÍÕõ±jçß¾øâr¯rvw]~Öž,ófÜ %ºá‹ŸøŸ¯çhÉþj~š@–[UWè{T-VÃŒväJ|?¥51§ZnU‹ C'2„¾ ƒ$¥*{šèf¿¿3Œþ¼Úaôgv=?±þôÝY‚é÷œrZìd«ròÎ0£¾6M×Vµë¹Ï×; Ý_Ÿ)áMÍ–vçK_¾rá4ãþ¯U«”øozö8XÆÉ°CƒTwu^}iœÕûíEi5½4Ë24i3é*²Ô&2èò0‚4í虩éÐ|ivC?͸›d²;Àt1„¢^O3iÞœ$ïú¶Wá¶åŶSç®\9Guºóâyµ~ä³+cæ /.²~þÅçTyüìŸtçá“®^9iŠ²v–¥«I×±44¤|G£¾×îž½wõÚj1vÕçN&½KKŒ¶öv#34èŠÁA[»š“ò)Õ1££Ã¬ƒ"IÒÜQ™&ZˆýiôçØnøÑdbøI‡‡GµA.)eþï…aÆéƒÿ˜.Ãõ€Õ×/Rµ=D&/^céÉ-ÇÉ4t¯fH©ÿäï#GWÁ}àÕŒ0[wƦ‰ç¯†}ë€ë¯*× äµ›oÎwúM{yMwsæÒ/j¾W³„/?<8Ü–%‚m—Aúᶊ¤Ur½×ׯ|×óãK:mžÌq§½­X¶2NÀ+³WÀõ€«~×àz\˜¥ëo…ë€ëázVµë¿¹®®‡ëXÍ®íŽ[àzàz¸€UìzáË÷ÂõÀõp=«ØõIÄOázàz¸€Uìú˜$b\/Ìêùp¤I¾x+‘ 3Ž™Ú²t-ùôyÑ÷7,D¢¦‡¹4tAáÅ9üË´¿‹Ò®0³ÍÞ–jE#~lncŲ à ZOÉøp=àªw½\.x=|MÙ{ë2ÄÓs'’ëúºJmÕQBÞÈûë±/ЇΣϋ¼¿ V™ÉŠ…q½´ÔÐ_(^úý]®vcˆ”æSCSë+x(jNÔçH=-Š2ŽÑöº$>\¸Ê]ÿô/¤³©s'Íi¦GS—¶ò—0§gœ­–ñbÀb¸>’Ú²‹Árµë®¯PÉœïSμX1?©ÈÁõ€«Üõw>ðμê×{+’±g/Œ´L¯yJ$–w-V;I9ØA§±9/–#‰ÓkNRŒÓAQ”½·6·Ì~qœ5ÔO-í.ÊÓ4T5µ4_šUßK9KÙÚ»ªêÜÅŠºzhüU,U5ÐNfxìt_y?’>‡ïU˜m‰äÃÈåSšV3E’ítŽd-˜ËÜuî\ ;(»Å<ÐZœZÊq½HQÞasL¬êË:-ÚÒ$Î1äËTmä™Ñ3CŒ“qXúŒt{–0ð@T2䔜çþ†ýyµ~ô#…¸ ö‚i&‰²â.gI,®\宿uÏ ³s} ?¡–(˜rš QßÜCIu¶^ÿù•/«03…ž’å|©ò8{éÓSUÓCUAšvü|Û”¤Âô'Õì«ÛJÄhdCò”úõÎq–2Ô¦º›ä‹eñÂúa¯BÖsg>ý ·Ô³‚>?^m8¥ËZˆ$×GéÙu¢·X¬KÝÔ™XŸëù •–Ó¹ÀwUP90¨(q‘e¤'ß7†’¼.F˜úI|=×ý 9úójw†Ñ`ˆ”Ösãm!+Óɵ¦ÑOþþÿÎÏ\á•ŠàzàúY#wšëƒ¼Òò¸^ÒD³:µBÄó†£$5y¶«Ìö€²çMI-å›>>[Ç9áºÜGj3¸A™Hy’l ÌØžQ%Áúi¯Â¸Oí?Xâb ¹®w÷ª\Êm©Ôîë–Xe¿8ÆÒ8˜áážÀûÑù*®'’}Ø•)àFv6²6‘Ù%qU’˜yº>Ò^ÍÕõ§ÊãgŸÚíÎÃ']½rÒeí,KW“®cihHøŽF}¯Ý={ïêµÕbìªÏLz—–mííFfhÐ=ƒƒ¶v5'åÝu}SªcFG‡YE’¤¹£2M´û;ÓèϱÝð£É(¸¢õÃãڠ™y‚¤âcýë>TNÆij.”11p=`µÇõK‡Tm‘ÉÆ‹×ØGzr—%¾ Ý«Ù#Rê?ùûÈQÅÊÖǪG˜­;cÓÄóWþÀõÀõW•ëòZ‹Í7ç;ý¦½¼¦»9sés¤Â÷j–ð凇۲D°í²"H?ÜV‘´J®·àzàú•ïz~|I§Í“9î´·¯ùÒ•Ù+àzÀU?‡€ëp=®ÀõÀõp=p=\\××ÃõÀõp=p=®Àõ¸° \/Ìêùp¤I¾x+‘ 3Ž™Ú²t-ùôyÑ÷7,D¢¦‡¹4tAáÅ9üË´¿‹Ò®0³ÍÞ–3ÊFüø2ãèGÿ‡-“Àõ€Õîzîzøš²÷Öeˆ§/æN$×õu•,Úª£„¼‘ö×c_ GŸy#@¬2“ ãzi©¡¿P¼ôû»\íÆ)ͧ†¦ÖWà")è²Ë-·ÖÐë«7®çÖ¹“æ4ÓŽ£©‚%=Ÿ szÆÙj/,†ë#©-»,W»îú •Ìù>eÈÐ^”uÂÖ–-šR/®¬×ÇL«_ï­HÆž½0Ò2½æ)‘XÞi´Xí$å`Ææ¼XNˆ&N¯9I1NEQöÞÚÜ2ûÅqÖP?µ´»(OÿÑPÕÔÒp|iV}/åpL,ekïª ¨sC(êè¡ñKT±4VyÔ@;™á±Ó}å üHú¾Wa¶%’k #—O5jZÍIR´Ói8’µ`.s×¹s5ì ìó@kqj)Çõ"Ey‡Í1±ª/ë´hK“8Ç/Sµ‘gFÏ 1NÆaé3ÒíYÂÀPÉSpžûzôçÕnøÑlâ*Ø ¦üತ6Y» ¥ÓjãÂõ€µãú~B-9P0å4¢¾¹;†òo˜TgëõŸ_ù² 3ÓQè)YΗ*³—>=U5=T¤iÇÏ·M)@*LoqR;º­D|ÖI6$O©_ïgÙ!Cmª»I¾X/Œ Ïö*d=wæÓzK=+èóãÕ†Sº¬…H2p}”žíQ'z‹ÅºÔMùˆõ¹žŸPi9= ‘ |W•Cý%±“G^\déÉ÷¡$¯‹Ñ¦>D_ÏuCŽþ¼Úaô#"¥õÜx[°ª¶ü¤cÙÄm¸®`íº>È+!]À“¤i´FWüå Gmf£ã<[á++ÈÐÒʤ· q½¨ÀzùL gŸ§±’5 s„¢Éa(’pTÂŒtæIy³òH¤½ é>Y&õ[j[z¸‚ôVZÇ)é+ÌÒûró™C}ùiJŠôŒ?¿Hn°ö5¨²R$ÄÜæÒç´¿aF>íÎ8ú‘ŒB\åðe[¡hú쾺ÏX“ÄŸü¸®q}xRšhV§VˆxÞp´‚¤&Ïöb•ÙpBöœZé ©å¢|ÓÇgë8§v—ûHm7()O’Û3ª$XŸ#íU÷ØA\l!Âõî^•K¹-•Ú}Ý«ìÇX:3<Üx?š'JÈTUÓ)†¥û3Å çúPûvôçÓÁ(ðâ«Olž>‰/Êë;}Êé=ŠÌègŸj“ ¸°Æïׇ9'©-tWN@ŒÌUAú º3ðɧPq=‘|lìîL7²³‘µ‰œÈ.ùˆÃ¨’ÄÌÓõ‘öj®®çIsÛ:;Z+³Io ×»vÝÞ“zOy"Y¼0`2=Nmî 8ž‹ãúð£?ϸ>üèGàzAFÇù±fEx‹#®‡ëX›®çÇ´1ÎcÓóðƒž“e•6ªANxï8´:/>ììø2‰é,Š÷ÜUÍo¥Ç.{d\ÙõÁÙÃIœ¬>aÆq†nQúîØÊŠÚ²QAÄÌ×õ‘öj®®U;ÿöÅ—{•³ËÙsùY?x²Ì›qG$”è†/~â¾^œ£%û«ýùiYnU]¡ïQµX 3Ú‘+ñýP”ÖÄœj ¸U-.0 Èú. ’”ªì8þ¼÷w†ÑŸW»3ŒþÌ®ç'ÖŸþ ;k†ë-¸®`m¸žû|½ÓÉÐýõ™ÞÔliw¾ôå+N3îÿZµJßù‘HÒô8XÆÉ°CƒTwu^}iœÕûíEi5½4Ë24i3é*²Ô&2èò0‚4í虩ÏCó¥Ù ý4ãn’aÈîÓÅŠz=ͤys’¼gèsØ^…Û–[ØN»råÕQè΋#äÕú‘Ï®Œ™ü1¼¸ÈúùŸSåñ³~ЇOºzå¤)ÊÚY–®&]ÇÒÐ"ðú^»{öÞÕk«ÅØUŸ;™ô.-1ÚÚÛÌР{(míjNÊ»ëú¦TÇŒŽ³Š$IsGešh!öw¦ÑŸc»áG?’Qˆá'Õ¦…Q=‘\ÓOº>gÔX§À>`ÕÇõK‡Tm‘ÉÆ‹×ØGzrEËq2 Ý«Ù#Rê?ùûÈÑ-5€0[wƦYΕàzÀu½@^k±ùæ|§ÇYòšîæLá’ŸJÃ÷j–ð凇۲D°í²"H?ÜV‘´J®·àzàú•ïz~|I§Í“9î´·ËVÆ xeö ¸pÕÏáàz\€ëp=p=\\××ÃõÀõp=p=\\€ëp=®¬× ³z>i’/ÞJdÂŒc¦¶ì]Kg>}^ôý ‘¨é¡F. ]Pxqÿ2í+Ìl³·å„ZÑH\h½pnÐá-uGé}«Ãõ€Uëzîzøš²÷Öeˆ§/æN$×õu•,Úª£„¼‘ö×c_ GŸy#@¬2“ ãzi©¡¿P¼ôû»\íÆ)ͧ†šB”ªȪ9Ô+€ëWs\Ï­s'Íi¦GS—6ÔæôŒ³Õ²«æü{•¹>’Ú²‹Árµë®¯PÉœïSŠàz¸¸>¢úõÞŠdìÙ #-Ókž‰åF‹ÕNRvÐil΋å„hâôš“ãtPeï­Í-³_g õSK»‹òô M;ýò¥Yõ½”cb²Õaïª ¨sC(êè¡ñKT±4VyÔ@;™á±Ó}å üHú¾Wa¶%’k #—O5jZÍIR´Ói8’µ`.s×¹s5ì ìó@kqj)Çõ"Ey‡Í1±ª/ë´hK“8Ç/Sµ‘gFÏ 1NÆaé3ÒíYÂÀPÉSpžûzôçÕnøÑlâ*Ø ¦|QP×[><Ë2,ðƒƒŽ“• !\X{®á'Ô’SN“!ê›»c(ÿ†Iu¶^ÿù•/«03…ž’å|©ò8{éÓSUÓCUAšvü|Û”¤Âô'Õì«ÛJÄhdCò”úõÎq–2Ô¦º›ä‹eñÂúa¯BÖsg>ý ·Ô³‚>?^m8¥ËZW¸>JÏö¨½Åb]ê¦Î|Äú\ÏO¨´œÐȾ«‚Ê¡þ’ØÉ#/.²ŒôäûÆP’×ÅèS"‰¯çº¿!G^íÎ0úŒ‘Òzn¼-XU[~lf~Fá½X©00ÚL!\Xs®òJHð$i­Ñ¹ÂQ›Ùè8ÏVøÊà 2´tG réícCA\/*°^>SÙÁçÅi¬dMBà¡hrŠ$•0#yRÞ¬<i¯BºÏF–Iýß–Ú¤® ½•ÖqJú ³tã¾ÜGš’"=ãÏo ’ìƒ} ª¬” 1·¹ô9ío˜ÑŸO»3Ž~$£W9|ÙV8cÉABq”êÉÂõÄõ¡\ Hi¢YZ!âyÃÑ ’š<Û‹Uf{À ÙsŠ¦ƒ¤–‹òMŸ­ãœÚ]î#µÜ L¤ý±9æòÂq¤Y%¾ \É9®®Ÿãýú0çd"µ…îÊ ˆ‰¸*H?Aw>ù*®'’}Ø•)àFv6²6‘Ù%qU’˜yº>Ò^ÍÕõììø2‰é,Š÷ÜUÍo¥Ç.{d\ÙõÁÙÃIœ¬>aÆq†nQúîØÊŠÚ²QAÄÌ×õ‘öj®®U;ÿöÅ—{•³ ]~Öž,ófÜ %ºá‹ŸøŸ¯çhÉþj~š@–[UWè{T-VÃŒväJ|?¥51§ZnU‹ C'2„¾ ƒ$¥*;Ž?ïýaôçÕî £?³ëù‰õ§?èΚ~½[Üïè­Jù~Û;XK…Œsu¸>”ÍÃÿ®®ç>_ït2t}¦„75[Ú/}ùÊ…ÓŒû¿V­Òwž%’4=–q2ìÐ Õ]WA_gõþD{QZM/Ͳ MÚLºŠ,µ‰ º<Œ M;zfêóÐ|ivC?͸›d²;Àt1„¢^O3iÞœ$ïú¶Wá¶åŶSç®\9Guºóâyµ~ä³+cæ /.²~þÅçTùžérçá“®^9iŠ²v–¥«I×±4øVzqú^»{öÞÕk«ÅØUŸ;™ô.-1ÚÚÛÌР{(míjNÊ»ëú¦TÇŒŽ³Š$IsGešh!öw¦ÑŸc»áG?’Qˆá'Õ¦ ‚çÒV3{jˆeÙA§¾!ð·ý*¸_*¨‡ë€ëWʹRµ=D&/^céÉ-ÇÉ4t¯fH©ÿäï#GD XF„Ùº36M<5ìË×O÷u˜Áõ¬×ÿö™o­ × äµ›oÎwúM{yMwó2<ü¾W³„/?<8Ü–%‚m—Aúᶊ¤Ur½5=?hP×°Æ]/–',£ëùñ%6Oæ¸ÓÞV,['à•Ù+"q}àx¨×ázÖšëŸz;uîX5®Ÿ íƒz¸€5îz¹\×°š\ïùú"\ÀÚtý½ûîëXM®w1%¨‡ëX㮟‡¯¬lB¹kä×OºþÙï<Êõ€«¬‘ \ÞõÒ÷×€«¸€µ9‡€ëq.®¬n×V6p=`>®¬‚¿h¸××àz\€ëp=®Àõàêäù×={~VaÆ‘€ëp=˜ dwÁ;ß=°m]TTÔºëvßýOo©ú¨Hå[ö£Ýó“ï¬>ìRIc‹»HŒ\€ëAäPÝŠ'¯ÛòÐ µ«¹÷DU:_X™ëI;µÄ®p=`Å»^ô_¯Áõ+ Kõ/÷lx ¹wúOhcÿû_½q}TÔúm÷¼È«5Ð.5—ü`מïýæGݹo÷mÏüú‰Q.6GGGßü„´o“íi?`÷Æk¢®Y¿iÇ7Uir½fÕ¦¾þȾͮM6íûÇ·ó»I½òYî'Û*àÂÀ7‡?Ñçßzå±{îØ·cÛþïþ¥ÞDc(ázÀJpý³O<ׯ4辬ïE¯ÛûÌï$9Í'­~eJ~²{óÁ·‹{ìöž’ß}mó®«ô´K²7Dm~œÔìz#&*wýäæ[eèì.ë´TTvÙæš7ïØr×ë9mÒÔ"?´wÛ“‰:’û A¶ åú¢¢ŸŽmµº.Vß½{ëý¢vÌìÃõ€•àzÌá¯LÛ›Ž¥¾óý‡¿rÓµ®{û}?Mh±¸½ZòÛn|¾ÀàÑ¿lßþ¢Êä’ìM7¾Pdô=„ëͯî‰~èœ#'m´ï½?ٵퟕzoدåŒ~,µû Ó· éú;•z¢~KÍûwû¾p=`y]ÇF¸~EKßÒ¡z÷á-ëþ¹ÙîÈøöæý¿o´y~d;òÎí›JíwOžùËŒ®wP}å}ùáÛ¢¯‰Š>ðÔÛ9ä@Æ#룮ݶs—‡7n¾áaq»û Ó¶ éú]ûߨõtÃÚøÖí;ž›¼þp=`9]/{÷Çpý g ë7ný^¡Ñ׿P8=®÷KÖ3ç¾ç§šp™u¶îŠ??rý¶TÝê—vízijøâ&·2Âõp=àªrýŸär¸~¥ak‰ûõ[¢ÂÆö~«màxÞ;n¾ö›¼®xÚ úÑ®Í÷¾£î%ÉÞ²w¾±yçKžûõ®·6¾}ÛÖÇ»¦çí[Ž¥$°Ó´õDâ3Ûw|¿ØdªúÕí[¾‘Õj¢¤©«A™QÞCr?!ÈVp=\¸ª\/’'Âõ+ ²+ïçï¿uë5î|øõ7xê·mž‰{ÚPóß/Ü}Ãú¨¨u7ÜýÂ_«õÞ<ü×;lÇÅÏݶŵí–oÚÀ wSå[î^ïþÌ-·>öf¶ÎýC›6ãÍ'nÛº.*êšM;<ñF–ëUÎ'èƒm×Ãõ€«Èõ¿yøàÌ×°š]<|\\X g°Z€ëA]XMÀõÀ®_úèi’Yl65Ó~~,ì§-PŸÊ_Ù³÷ÕJ”Û¸þj·|àìJ8ãOYû®×_ ®?em¹ž´SsvýÔmÀõ+KôQ„Õý´ª²n;ï;ôç7ß³1*jíOûËÅN«H;E‹Ý9¯?ºÿw{~=£ƒôº~ï÷ÿðꃻ7¯»fËÿü×:£çÓ¦̵yç+[Mò­Ág,yi×ÎCî5rghwæÚ»®O¡*þó;_ÞµnóžG~ù‡gvOº>ȇOÝ¿SׯP×G.´Ÿ×o‹Ú|ß¿«ºlöíÝ|Ÿ§\lЊ´\×Ç)ª:M$©o<}ãÖ§ROÒžO»öà[Å'I²Oý»ƒ›÷þ¼Â½ìm°‚¹¶c<°õa™Gö†¢wìþI™iæv#©½«W½¼sË}¬ í:åk·¯‹Úíq}ð2»ÜmÀõ«Ïõžš7–ê_ìÛ5Q.6xEÚP0W¾¶÷–»Líú´í×'«ßóò€ò™m{^«4‡(˜koþß÷l}@â.n7÷üM{^u]ÌØn$µw]ïÙqÓ‹ÞîNxh“Çõ!Êìrëö€ëWªë#›Ã¿Þ¿Ô|ðŠ´6Õ_ñþ¿µûÖë·FGoYå­–÷ƒû~Qí½_o©}ãÖ觕†sdë_n½OØNöçxmÔ.Ïýú ×\?ì¹uiC—‹ V‘6»6ñåƒ;wì¿ûëß|ôÐÏíÙîuý®½ßÿŸ=°{óº¨Í·=ó—jƒ7?HÁ\7TgìC®¿29ã>c»ÕÞ¥úËþøÔm»öÝuðÞG_úÅ“;&óðƒ|8\p=àz®\¸p=¸Š Uªåê®+ KÍë·ßùÛ†åS;H<é_hžl}ÿ`TÔæ§3úè9º>Ô'€ëÁ p½íè»6l¿-úúÇåÝ\p=Xp2Ûz÷_Z=ëÓ˜Ô/í¾ub!ú‰åhnyöCü‡ÛoÙu \9ZoØçßzå±{îØ·cÛþïú+áÎäzkýo¾¼ùBù›ý›txWÉ U‚6Äëp=Àõ`n®w—v¶Ø\±w«ø‰m·†.Gë©!ýtl«ÕeÞÆwïÞz¿¨}jy²·B¡¨î·´çɲŽûV¥3W¾vó¦b;Moíßøµÿ>1±Q¨´¡^úÉàz0£ë·ßðÏùÞ¢2Æ¢n SŽÖ]]fÇ!oKÍûw S¯Æ±ø¥›6>«£ÜËòïÿÒ÷ŽÙÔ  ñ:¸ÌÑõœr´¿ WŽ6ô*úá û•ß‹Þ0¡ú‰<û×í»Ñ¦mð×Àõ úܧn8ð^³Ýóò»Ñ{']ãuÿ”Ñï¹ýÝŸýtt˜r´sr=Ý›öÔ樨¨õ7¹Ù¸Îõÿ[~QmY‚6Äëàzû±÷þ!úAÁ —ì­Í‚Ç7GÝ2éúmQ_ºç·Å'I²·ô­{6…+G;×S]²ooÜò?âtê<´U¼{÷ú?QC”  YšÀŠq½\.‡ëWT¿úÇölÛ}Û]ÿñgo>¹# ï÷ÿY9Ú9¸žlÜ·aç+jc ý“ßrã¿èC•  YšÀ qý½ûîë¯PÆÀì]9|¸\àzp=àz,½ë¿ÿi¸°”|æÆc7×ËÅp=`‰}ƒ07‚ÀØaìæàzÌáà øc×|×cìàz€/àz×|×cìàz\àzŒ\€ëá ŒÆ®_…˜4ÿºgÏÏ*PDÀõðÆîê;µÂõ`*Öúßì‹ò±õ¶'~›£#áz×/0´±Qúê£û¯wÿm¼ù/þ©ä$éûÜpâɉŠÑ¤.ã•Û6Ýò‚¬Õ×/ÙØѦ¦Ä7þé®›®uÍæ=÷ý˻ʎÅ:þkÃõ·Âõ+Ôõî“uÙIë@sÆÏïXwÝÓ}p=€ëôÏìè_ܼùë¿L®ï6Yû[òÞ}rûµw¼Q¦§\ooSÚ·éÖCŠv;âú¥»‰¡¹öÀ%-½&S_k™ì7ϼ ë áú¹»þ›[àúêz_Xao~ïÎõûÿýp¿ï’ìÎyýÑý7¸/x7îyøõ ï_€å¸âÞºÅõêúèÏKZ'ÎMVmêëìÛìzqÓ¾|;¿›ÄÁpýDàØ›ñôõ›ëœü› Š~´ãÚ{cNÞ?ÀŽã²nÞtû+º•ów³\ïšG¤íÜÃN?õ¹«îyþ­W»çŽ};¶íÿî_êM´cV§Äµáú×_É®§ 5ÿyßµ›Wô˜ü®/ŽSTušHRß(xúÆ­O¥ºÞhª|}ÿ¦ÛšÒl¢(sW]I{2Ò\óæ[îz=§ÍBšZä‡ön{2QÛ¸Þ­ˆÒC7^û`B7øZÉ¡7~[Ñkqý®»ùɇvl¾ëß”=+êOf-¸>ØÐLüÔWòƒ¢¢ŸŽmµºìÞøîÝ[ïM\$D~J\®¾|/\¿"]u͆M›6\uÝ݇âZ¬Á/>Í•¯í½åÇe&“úå7<«àþÉ”ýd׶Vê½'ZþÁèÇRûpx\ïpè³þqãõÏ9xo}ùK_Ñš&þ×E­ÿê;ÕFzEºµàú`C3 ß©Ïíú;•š<¡|ÍûwO|3‹SâÚp}ñS¸~EºþÚ¯ñ;»ºúŒ¤÷\3ù IõW¼ÿÃoíß¾õú­ÑÑ[ÖEmý^¡q ã‘Íû×ÈM^q½¸>êÚm;wyØyãæ·#°p½/xLì™×oðÆõî‹Q¼q÷†­ý¥Ñ´‚t¿vâzîÐLìÔ71‡¿ÿZ‹ïjíöÏ'Î~‘ž kÂõ1I\¿"]?y¿~òÀû i©ýõþè‡ß¯éwKÛRýo{otý».bwº®W§\Ī_Úµë%ï/p}toúw¦Ü¯×«~´ãÚ¯ýµÕw¿þ$y2ïç_¹ö†'ùGWLJìÚ¹_ÿhì”\<:è©/”ëgqJ\q½\.‡ë¯.×›Ê~ró®ä÷»~HveÚuÝÄ/|åëû7ßùZÆqEYºêKk{Ý9'U¿º}ëÁ7²ZM”ƒ4u5(3Ê{Ö¸Þógvä¿îÛ4‘‡ßc¶Ïÿã“Û¿´ÿßJóð]cY¯ìÿÒöïŠ[¬pý’õè_ؼñîWÄêÖ~“©ÿDyÂoŸ}AÖ\ôÔÜõ³8%® ×?ý )\u¹Þa×&¾|pçŽýwý›zã¹=ÛÝ¿ð®ëÜ–¤×Ú³)**êK7xAêI:µi3Þ|ⶭ뢢®Ù´ãÀod!7Àõ>(}ø§ßþòuîg¸¯ÝùõþSÕ=ýùzÙ™zhï—v>'Ó®€ì×ÎóõGd¯ÿ7®wÍ·=ô£÷U:{ÐS_(×G~J\®¿ówàzÀZtýUÖÍÃØÍÍõ·îy®ÀðÆn»käà øc×|×cìàz€/àz×"€%jr!SLMå¯ìÙûj¥9Ò×gõ!®p=Æ®_ –ÿ"€%2>\àzøc×ÃõK5Ê_®p=€ë1vpý*}H݇®Ò8­ #Õ)¹/ú[q]”Ãa®úùÍQëU¸W1•ýè–}?¯2û\?Ðô¿D?’à~›cñ¡»^*æ—^*—¨øÏï|ySÔºÍ{ùåžÙíÓt¨×ƒVÑ ùf°f]O€¹‚±ÃØÁõ«ÇõA«4+Èh>òwnF9à [þtpÛîèßW¶#ïÜyÓ³yz\OÿËÁè$ä=ÓžŸi¸«B­©W½¼sË}¬ í:åk·¯‹ÚíÑt¨×ƒ–Œ õfpýZv}Ð*Á 2–ýøfW lê‘?¶ïþ÷›_=ð›zS§ø¾èâÝQ¼¿ê}›ðþèoòµ¤ƒîËüî¶[}‹Je²TnÉwÜô¢ÊÿÓÝ mòh:äëÁzØâÍàú5íúUƒÕ¨íÏø§mw½W÷â­ß–µ–ýëþûxõOßxàÇÜË?ûï×S]²G·üËq[â‰è;~ß8umïà¥r¿½åŽ?4yßjR¿¼{Ï„¦C¿¤‡©Áß ®_ ºþîUƒÖ¨%Û÷nýúë‡î<ø^³m çî}ø×ÏíÙó¯š äæѽiOmÿ‡ÿ¥’<ýïM\¼TnÉvÜüJ¹·ÅþŒ'¯»Ùׇx=XC½\¿š˜EPÚõ¡jÔºÞqÛúuënyÕ¥tªSò­Í룶|'»ß1Åõº_ù½›¶Ý±c³»D÷´6ƒ•Ê¥õª—vï|>µËÝʱ¿>xmÔ® M‡z=xC½\¿ªî}¡«4¯Qk*ùᶨÍßÍp¿£é?nŠú¦¸“šæz‡Ã¨:tcÔ5÷y2ý¦¼T.Õ_öǧn۵ﮃ÷>úÒ/žÜ1™‡âõ = ùfp=XXìÍïØúm™ïÑ;p=XUЪŸíÝþlVcp=XmP] l‰ºvß÷¤­v €ë×®sd! Ö¬¼æÀõ€ìRIc‹»Hw ›Ý‹oáÙº~iz®_>\®_•Š÷È·KùìŽ(›£££o~Bê^'HÅX÷Â;{øÞ/¿½ûÚ¨¨­šÖÚœôʽÛ×GEm»ÿ7ª¾‰GéÉö´Ÿ?°{ã5Q׬ߴ㛿ª4w}в¶S· Ò+p=˜c =%‚Z1ÖåúmQ×=ønY¯ÝÜ,||ë†Ýw=/hГ†šwnùÊ;GløÍ·ÊÐÙ]æh©¨ì²‡rýô²¶A¶E\p=Xׇ¨i[òƒ›¶ÿ‹§d,}Rþ­{åY]—lùßwm}³}Ç÷‹MÁš ZÖ6ȶSz®sv½Ãv\üÜm[®‰ŠÚò-BK­ëM•o=º?z}TTÔú-·>öf¶.Ä~°²¶Á¶Ú+p=àzÀõ€ëèú}vÂõÀ*vý­p=°ª]ÿå »àz`»~ÿ†Ýp=°Š]ûÆ›]ÿâ _ø¾ð…¯ÕúuÇÆ=z½¾£££©©©ªªª¤¤$;îÍ'îÜqý–m_ý)?Q¾2I’üþÉí×Þúƒ÷"y³ô½_ýöO’DyRü_^9¸eëƒoI“æÒhÜp+~7ðµ„_û6ì¼uÃÎ/oصÃnWDïýM{ÿ?[A— endstream endobj 362 0 obj << /Type /XObject /Subtype /Image /Width 674 /Height 561 /BitsPerComponent 8 /ColorSpace /DeviceRGB /Length 50600 /Filter /FlateDecode >> stream xÚì½tוÿ¯$¿Ýµ“]'ÙulÇ›]'û‹“ì¦ý‹ X%Ûr—{ï½ÅŽKb‡À 콉`ÀN‰) , !D›`†%‘e[’õóÏ–ÿg÷ÿß Á&JÔýž{xÀ™Á¼ûæ Þgî›7sÿø‡ßþþ÷¿Žà…!{àþ»|Úƒ¾íîðaù³v·=ÀZb÷>À^jºì>öÈR{l©íeÙ£,{ܯÝÏØc{"{ü§=Œ=Ყžx0({’±§=öP {Š±g|ØÃ>ìiÆž]ÁñØ3<¬=úܳ>´½€ì9dc/"{ÞÛ`/9í…åöÄr{Ùi/°'í§½Œ=õꢽ¼²½æ²§_{eõ†¾åøz0-zõ cATÄQå—YæÿX9æK¾ÍW‹lG‡y5zp§ŠÃ˜³+øS·ÌÙÜyŽ~¬c~JΟ›ï_Ⲭó‡¼ÂÝÝ'8»ˆ û“Åþ'˜Îj±g ¦ôî6]©ŸnvIo¼´¯f÷äË»zßDpÀb9DçaÆØ„òϲ‡–š…~@铪 m}Qز@ÿõýò·¿ùOÄ÷ÿóûß ïس l{˜ ô¿ý/ôaÏí‰```````ÛÉœ ʃmKÊÿê—?¿ý¶„ í¶[ãwïŽÝ•›˜“À_Ñâ㣑ÅÅEÅÅE¢¿ññQ ñüÝ»boÝçÜgð€oôˆ³+"ò®]1‰ñÑ1ÑÜȈ0.'”‚þ¢ÏhIb\ZµÄâ¢Å#¸ÇFDEr"¸¡Üðáa· ¿è3Z‰®œ¸Æ000000°UY0”gŸÁㄤ>~WÿOÏÈÞ¸ÿ>ú‹>£%\NZ›˜À_´DyÂÇEñ£¸œð?´÷ÝwÞH–bÅE¹è/úŒ– åh-ÚìýîØÄ„˜]»×ôÝ]üÈ°[n CGα+Ëv£ŸBXèÎ[…„ñЉÅ s/‰IŒ‹¹eçN.?æÖM=Ösƒm#Ê#Â0„`Dm~4ïþ„ÈiÑK_g½y!ãµ i¯|•òú‹>£%h9Zë@6â;?>Yt\lTd$'66 ~œŸ—‘“–™‘’ž&AÑg´-ul“ç}¶{?&2<þèÏhK´}P'Dr^¨Òšœ©Ç„wsyñ åù±‘áîÑÑP^T´»‡D—¡îèÿ–(üŽÞåº*ˆIàs9¡Î;CÂ8üøNÜ=®N2ŒYÆK@[&òù¼ð°Åpâcby¡žá~³Ãب˜ÇþšVœúú}üˆ¸ÄE/±58¿ø-fíÎÕïÜG¥¼¿å(ÚµvçΰðÈ(Ç96ŒÚµÆýçs¢c3wé!¡¡\×nàf:šîÚlgX¤³ÍáÇÿÄ(ï¡r=:mt-¯Å„ /…s#8»>>Ìœ¢ÃÒ»8x‘w¼ÓÚV&y,&:ÖIð%¥GŰΟ‡Ë÷qØ•àç^vªÇ$@¶ý)¿+1ÆŸ%&ÄÄðyÒûbν÷§¯ò‘YŸæ<âüŒlø‘Pã“çg´ä>fûø¸è¨(Îý{ïBa{Òß>HúÛûûø½þúî_ÿòÎ_>|ýEŸÑ´­EÛìÝ{Wt$'€nãó¹ño(Ædï­þË÷ìâóãvßÿì›<¿;šÏß•ɉF\Ö71‰¶±¶V~ôP 7"!ùsPÞ621ì£íUÝËŒJLàÇGEÜúrJÛг[Û¸¡7ÿ•Ä°pW'鶾o çDEíy7¯æðˆ–v,ÄÕYED=×6aœtl66Ú™ûÞ½QQ±‘œÈ§š ´D³ïanx„¿¢Ùµ[‹óè‹Q‘Üøç¤ýæ[¶£ÍeïÝì΢ùË+ɉs}=:*úau‹f|Ì1l2i:\‘)Îl8ÊÔ×ÜÛ’òTy¸Úý‡ñ8AøÉÛõ‚´üð¸ž)Î>>9ÔÞ$Ùé÷€í0Ë¢£yüGþ’`ÄqjჳæGDøiÿÇ<žiwáuÚL=}wúˆ÷iû{ ѯt0Ÿ'²îåFDù,=Òsþðx1þšØ·Hyï˜ëãTçÄÄÇñÓ»\ Qõ®,""|ø¥Û>}-öô‹=øÇ;qCÏC¡ÎUh´%Ú>.6šÇ {ã—>þè=„ud~ðöï¿õþ{o: }FKœ«Ð6hK74°ŒÅñ8»Þ=ìèµÔ»-ŒÃ‹‹c¦ùÅò#¢øüÄx~ /úéâ~W·fwõoƒ©÷Eqyá®nun„͹J[ýt$/íóínÇzâðÈȈ]Wútdè"åíúAÝØ ¶éÍÝ<Žg'”ŽD’î ç=Q®™ÂGMZ»³Ü£épÃü(á§häóbÕÖæ$t×_:bsùã膿.—ŒÏ±\Ξ¿8°U÷rlè-œûÒ,£üò ÆaNûÈ¿ÓÃp\×ùÑ=<ì–¿üåwþô*²wßy ¡¹¡åhíŸÞ~åÃÿ„¶à†ë¹<~Ø"åÓåc¢–¬|V΄oæò§yaѱ¼°ØW2=È ‰ôŽ8ä0Èèx!:ŒÃ‹|¤ø¨3:ª®“¾}o'Ê»“D9Šˆ‹ðêT¹(ŽN@ˆç%¾’×bbMèC¾…ø¦<»hµu:ùlZ6Ka"÷‘ˆð˜¸UïÜQ)/cW9,l×ß·˜·…‡qÝn¼.j‚Þ|tXP>G‡F>WK9öùTDXë‡òK[3(‡½ü¢?–Šg"C£c™fã¨Q¸ÿYWwÌÁŒ‹Œð¢|x4:c£¢½)º¼ôø8žw-‚,néqðuû8Õ£#ãWúÝ]Ö†(íÏv% ˜*äØ Ñóý~îÑß:­ÿÎ_Þù‹ö=»çæïùåâòùÇ?õ|ËÇóëß~ëÕ·Þ| ÙÛo½²x;ÞiÎôhù›Ž Þzó´}7\à ¿õƒ~G/×÷Ñ­aá(ꉊ‹Šáó"##c»k¢c¼{f. ”odð⢹œ¸‡>̫׺–{²à…ƽ~ÈÑI¦Þºë¨8æÁ@ò¸ÑÑqQáü—å6´ÙU&xû3ì+Q~IÑîªÅ­Íù°ˆçê÷»Õy©9Y.Ë=¿'*"võ;G•bðØ%”ÿÀ- hz»üþã¢Cƒò9*̵™¹ô)^XÇ?åý¯òë0ëÈ»ýaœå3 £b£1Ž_ÿ#¹KÎøØÈŽ壘—DE±(²¼ô8v,dqìãÎY~£û:Õ¹ü¨•z```—­9(åÏ7îÃ{›{ì·'ú2õí?¿óæë{_¾ãÔ[»^¿gÏϯï½í;W¡mЖáˆòúúk/¼öêso¼þÂ[o¾Œbvg\ïŒßÜÑr´mƒ¶ätÃmhß· ;oDv—½÷èqQ‘Qq{|îϯßÃã¸Ç6G•o;GY[˜î‘¨-&4ÔoÈåÇÆ$F‡ïÜyKooÊ ³ó£)w¡€­…¹“«S½‚V…óbcxì>< •x§ø³Ÿžww…ܲ“÷x½yU”çÆ-V-†»çÃÂ8{þ6è¸QÛñѽQ!¨;ÃcvàÇz· wŽ*Å:à±¼à(¶ŠýÇp‚ñ9ŽÏáÜ)vDMºÇ±bxaèå=GĽ#iÈ1yRõî휷 w¢by\ÿ5 ÷çÿ²3]&±)É¢<'ÜWé¬Ù›Á·ì8,?£|žê\^DÜÊ?=00°ËÕåãb#ýY"3·-òãÄ?¢PýÄÞŸ#Ýû»Ã/ï™{6 }ž}zgÿ«wiïÿsÕÜ£¿û0á–Øô­Øè(îžÛ_}å¹—^|úå—žAÓß|ãEwô}FKÐò_x }¸íÖ~/€‹y_R±lÓ\ù|4Šø¢Ÿ)tOdr­ɼ?’y¯÷Ø&+å=”c LvÝ‘‘¡AGÐdèx3.:;óʼn¦g"PâV¿snû€Çxßæ@yNÔ*öŒÏÌf o+Æ–64Skr#»5ÃrØûÈ#sãߨ;ê=ÿmzÿëqa~jÄãú=摬ÃÅT6†Çó¦|dó¶çPïûò\ߥG{Õ"<¨â–žáËO`ߧ:'""6ˆŸØejˆòˆËþ …ó±á!Ú½¿=yÿ/ŽßóÓ“üúä“Dœvò‰?œ|ô7̇~Ñq×oQ§ž˜‹,.–Ïå„<ùä#/<ÿäsÏ>þüóO8qÿÊËÏ ¿è3Z‚–£µ?¼—Ë ×óºá±˜ÈpÞíÏJ+T#çók“¸¶C•ÿçGâyÑèš…›ð‚¬zбŠžè¬>Çáðù¬þ0œÅ…0þ#ÒÎaç vL´î€Böt"Åçðn}© £ßqKtÒÜ“ñXû6hd$ºþ‰âò”™0¸:a‹f¸3÷[ÃýQž]4'Æ»jÑkpžÃçq¹»^H®8<¦s ³F™ó?"2r ;g*åý-~°”YÕþƒômqÛ«ÉÅm:wÖµa6žj}.:Œãÿ€„ë0ë¤âqy·¿š^3 s´£}TSóÖí¼p®Ïð?bYeù,ʇEFÄÆDDzSžÃóYúm|ïýUÜ’ƒïãöuªG÷»»LÍAùˆ–Ï‹‰º/â–û~qâÞŸÍÜù“冖£µ·ón‰‹Ú½+=êù£"¹O?õÈ3O?êüûì3!²£¿®%Ï<öÔ“FF„!ÄÅEvƒeÑ\^xˆëUáÌ{BBÂ8áÑ|Ç*>/’»øN”p7Æù>×ñz“GOÍ a6E½}DLTXXÈâ‹EB¸\¾ã+1|Žûåã·ì ‹â/݉c·Žw°Üâvå–°°ˆ%[ú/zi½Vïüâ·}@Gƒ±æ/û–÷Z^„ã…*¡<ž?7‚Ý>ÇDóöì滾|§à Cۉܽ<7Ð Þa–9üY< ÂÃùjÀg—çØCh$ÏS–ëT¥/ÙOpÅ-©àòØß©¶ Q>ýüZbblL o7ï–æ[ož¹ûßgîü×cwÜxlÏÑ_ô-AËÑZÔ·ìÞ·›É^‹,!!&*ŠËã†Ý¿÷®'Ÿxø‰Çzü±†>£%h9*=:ŠÍŽXÑ °+Éx¼°Ä{˜8”˜Ô®ñê.é]\^ŽX†8Ëæ­h»cQœÎ儾õ_­»~fºý'ñè/úŒ– åñî(ÞiÌk6У8¾3!žÏ=·?üð}£%\ÎNïÇ;ÞxÅ Æ °+Æ8\ÞíTµOêï™èÞ'z2ŽÃ…ó læ <7‹BA}lLddD8‚Øþ¢Ïh‰ã­ÚqKÌ úÄ„Ø„8>âx7ŒãÈ/Ïaòˇ9sÙ +>ƒxN>€]IÆᆅ,¾ÿ~gh7Ž ØêÌ1fÎ Ò‘ã˜7Î9S\Å­hNú»Xã0~|<‚{tlL”sŸQA—¶*C”Šä€m?sP> lûóÐmD8Øö3Dy000000°íj6@ÛQ@y€ò €ò €ò .OÊO6=|ã¿>®2ú]¯xäÆS7¶TÓAá®›®þö·®þÝ_zô‹ŸßN @Y›ã @—.åÍ/ýdÇŽ?ç ÅµÄÚûá¯Ð’žj ˆ[BùÉÚ{~tã£õz’ý–ËRj‡qß_ ¼ÖoQ\× ºŒ)ÿ½ŸüóÏ_é4;X¼õËÞxÍ·| S~,wç÷~+$–~ÞŒ±  <.kÊ_õ‡7ùß?}¦ÅÄüßþâüô7þw£ƒò¤¡]x÷ýð;;v|çÿy— Í>Ê÷âþýªß¾úFÎ3oïº~‘òfuþ³œŸ\½cÇŽ«~ùjåî—ò¦Ã¹Osÿí»hËï\sóžä>«ÍOYËöiɉ»þïÑ¿÷×üËM¿¾é‡®Ï×þñæÚŲ|ìßË_~ÖÝsÝ{^y˜÷Ÿÿñ“kpSÂ_;&QñºòÝ×¢­v\}Í5×Ü%Àñ‚'w^ÿßÚñ­ï\uíÿy^5‰.‹zÒ_yA|Àä¸Dòú @ Ð%AùðœÆ·~ùo7NÚŒÊ'nºùÕÆüh'åõu]õ¯_­ÕZ­Úº×~sõuÊu¤N~ÿ¾û‡wU¸USþÄϾ½ãz'å­/þÇwñli¿ Ÿì;ï_©Á}R~RõìMWýìѼžI‚0·×µEœõY–Ï}’£Ù·|ï·Ò!&~÷þ¼X–Ïý/®õígÝ=ßßqM|JŸ]"t½ó«ú£d_Ë£ÿnøñ}EtÑ€Oô*UÃ胡æöº*jŸŽYïý@ KƒòEG{>üÍ¿Þ+×ÔÞÿ“ÿúó¡£¥NÊOÖÝû/?ÜS¥wnª¯¹óŸÿù.ùHݽ×þË]rƒc9’z•“ò“]÷ƒ[Ë]„ÃÕÂ__ÃËóEùɆû¯ýþîò 6ù}•¥õ¹Ï•(ïsÿžµ¾ý¬»çÚkï«ŸtŽ4´>wÓõÎØ”7*¿ñšÐ7Kµp^.#a hÛi5”Ÿ@´ûÃ#ŸŽþñï?îÃu.ÊO…_}Óë]®yy–oþìêÐ|uQøwÿãínË"±¯¿ÑAù‰"ÎwvüÝ~tS?úáÕß“è}PmyõM¯¹w»¸ÐgY¾ö‰¯@yŸû÷^ëcŸºº{®»é¹6çX»¹ë•Ÿ]{{aåmÄXÓÇ÷‡ýôšoí¸ææØWK‡pÀèr¡ü7 h{iU”·áC²ïìøŽc ZçËßQ½4–¿çÚnrF½¶ñ¢èïÝàŒåö^wÝ^W4¼Œ­Kbù¡hÚG,¿´,­Ï}Ë/ß¿×ZŸ~ú§ü 7>ªX:¯À2¢ü çp‡kHº,(ÇÚ–¿è (o#'û«;&•ºïË˸îêß½Ù0Šã£oþþêí•ë|ïõ?Ú“Ï<“f>ôqÈßí¸Îy_~²ùùŸýÓ¯Ÿ+é›$løäpgyQ“Öÿ}ù«þDÑáI‚0 wÔ·2÷å}”EúÜgP÷å—íß³Ö§Ÿ~(oîzõ§ÿ‘9ìœÃo:”—U{xÂJ’æ#™»þùÚ»k'aö(.Ê{i‘ò6Rßú·;~õýïìØñíïÿêŽ[tŽ9öãïÆþôºŸüâ׿ãî}:úÚÅ9öuÑ‹Q?ý§oïØñ­«®½9ê¹?³ï.{³ž½ñª;vü¯ïß|‡Ì9ÇÞWY¾ö¹"å}îßk­/?ýPÞf9,½ý§ßýÖŽß½SëT¯poºæ;ÌÔýïþïÅ}f߀ò èÒ¥<ʃ@  <ʃ@  <ʃ@  <ʃ@  <ʃ@  )¿]&Zí‹°€ò Pu{@ Ð%/'å;ÙZ¾$0Äò ЕIù$t K­V/§<úuåA  E›j¥¿kܪrò M¤üò@Þ;œʃ@@y?”ç4 Yíva5'újóëY”—ä6 Z˜Ah›¶›ÔõY¬Ahaº¼?1wbš²S6Ó˜(;cåÜö rzáÂW§¦ìN£ª,Ìû«iuT,}òÌlo¶$ÐnÕÏž?Ö¥è38Nv»þ@‰”ug¡iÈ`2[qÂFOÙ =)î |]å:VËJ:F ›cØÝfn. ‚õ QÞôéIš¢)Šžš²UåŠS ¤¼¿ß;PÊû¡¼0­AGk2ìÆ´‰ŸÑnÊ 3T¦ãŠlÙÅ9ª‰éñº”ÅhUZcšÕV&»þK®¦4ûÄ«‹©Åû†ðCËi›\G}ùÉh}:æsƒþ˜¦„ÅLOÄ,Ìj·ŒVJ’6¢\qa¯è)s]4`iUj;Þ™ƒé•0¥¸²(s]ˆ(õ”º8ÚÊû äÃy <”÷AyQa©)“°¦¥iܳïDÅCÓc•,\&×è¨Þ|l1¼í´NuÊKò2’±µœû¥¼o”yþ­·Šâ“ jŠ¸m„Õb4ØNÓÊ” (Wª0ã­Þ7ë±Ün›¾&9(¯–ÞúÏ=HhËó  ¢|à×ÛåA  ¼ÊKåFk“ŒÍ×z+颼Tn=;O“Þ²Q33Úr¯Ë$£XÞrHc (šï*–nå’=cáµ&|‘§¢¼n’Ö4äJ®%NlåÑ5®.bÏ‹—”Å»Üsñyµ\©JÜ(—©@6û.€€ò P~Õ±<–sÀª­ðÏ)$=Mì5tžÚ093T,ÚdÊcù½äp©WŒ¼ ¶ëˆå-x[&+–Ï9`3È““Ö@yiåQûá Hʃ@ -»/ŸÚ ›:ÚèšS‡eÔifÎ~áy^^ZªÆÇ[<3ÐDéeÍíÕîÐRÔÜ`Y²{¥¤ ›:ÖWࡼ´J?}¤Hì¾$È*—ïK®›ò‚t•…èÌÆ\séªúìç¿žaÅòk/W\t˜"{ËÝ÷åÓk(¼+ f„!)¥vÜ6Úœ'q]ñT Ò&eº@ʃ@ -£|’sŽ=NÓ”$óPca~nÖ»ßè"”íëµ2cõ$n1› Ãe‹ÚeuËÀ€šžBߦ¦¦, ¬éìIÂôz 577CÛǃª÷°k¦;3×ýü…3Ç)æ£Y]î ¨ ¥z€8uáÂ)b°š™é‡e·èf¿º0oìÌu]Œd)´6¦HzzŠi©P’Èc]G®h}åºëÛ9NRÌ ŠÂG#}lš¦é)»®³8ªåA ÐSåA ÐåJùè(.PÊCßmKÊÇÆDåA  <ô ж¤|B<(å¡o¶%åoÝ”€òÐ7‚@Û’ò{nKʃ@@yèA mIù»î¼ (å¡o¶%åï½ç <”‡¾Ú–”`ïÝ@y(}#´-)ÿðƒ÷åA  <ô ж¤üc<,å±L…–˜=çyý¦K\¢ýt¶;û¢¿î}ËÅÒ+êHAÚíæÃû '(6“òO>þðjby©Üˆ+7†ò²zýxu`Ìa9ícÃuiKÛ­*IR|È6£k-”Â+ìA@ùí©o@ëœ?›AùgŸzl‹(LØÍÐV•›$.˜6©2 (¿­)aÍl‚ƒ°”ṧRžÉIgµÛm„Õdœè«Í¯gQ^’Û4h±19éì´Ý¤®ÏywBÂty?~bîÄ4e§l¦19PâLøŠå¶OÓ ¾:5ewŠU±Ö¹²ÃÑ'ÏÌö.É‹å´êgÏëRô g†¾õJ¤¬; MC“ÙŠ6zÊnè©XL“·®r«e%£„©1i³7{²î®èU’¤b|J](6€ò@yPþ¢QþåŸõOyaZƒŽÖ6dºRº"h'>£Ý”f¨LÇ'Ù.p‰sTÓãu)‹£ÑÒÓ¬¶Ò¶5¹b˜Òì¯.¦ö›_žúò“ÑútÌ æý1M‰Ø{ψ¸0«Ý2Z)IÚˆrÅ…½v¢Ç[K«RÛñÎ,8¯„Ù]–áò¬ÊƒztADZ´mE0nÊåA@ùM¦ük¯¼à—ò¢Â>RS&a :kܳïDÅCÓc•,\&×è¨Þ|l1¼í´NuÊKò2’±µœû¥¼o”yþ­·Šâ“ jŠ¸m„Õb4ØNÓÊ” (Wª0ã­¬w,·Û¦¯IÊ+¬ —0M’º†î”Ê~;ŒÞƒ€ò@yP~“)ÿÖë/û¥¼Tn´6ÉØ|­·’.ÊKåÖ³ó4é-53£-÷º,H2Šå-‡4‚¢Éñ®béÆQ¾!Ù3Y Ö„/òT”×MÒš†\‰À5¡Ä‰ ¡<ºæÁÕEìwIùQ¼Ë=?WÌÈGûÌ©á2©g¤Ã`ëÌ̃€ò@yP~)ÿÎ[¯­1–ÇrXµþq)¤§‰½†ÎS&g†ŠE›Ly,¿—.õ¾_ÛuÅò¼-“Ëç°äÉIÁPÍÃæÞ‚ÅúK*ÇÉ9@yP(Êo"åß{çÍ÷åStSG]sê°Œ:ÍÌÙ/<ÏËKKÕøx‹gš(½¬¹½ÚýZŠ‚š,Kv¯”tSÇú<”K’V駉ݗYåò}©ÂuS^®²Ù˜k.]UŸýü×3¬X~í劋Sdo¹û¾|zÍ…wåbIÁQ>I˜Ùl˜ìÊw”,Êm7ûs1 (”å7‘ò¾ÿ'ÿ”OrαÇiš²“aj,lÀÏ-ÐúÎ<'­…²}£Vf¬žÄ-f“a¸£lqB»¬Î`0PÓSèÛÔÔ”e 5=I˜^¯¡ææfhã¸qPUà6pÍtg溟¿pæ8Å|4«Ë¤T§.\8E V33ý°ìÝìWæ¹®‹‘,…ÖÆIOO#-Jy¬ëÈ­¯\w};ÇIŠYAQøˆç gE¯\ä/Þ¯cæþS”u°>Êå/²&›òoü×B•1Øå@ùËžòÿåÏ)€ò Ë‡òdßGݱãåø"# ”Ê# >þ(å¡oÜ”·hß¹ùÅþékÿÑ1Båò6›HðPÊC߸-(oî(û÷«XyÙMW ’]˜'&úß‹{÷ªÏ_}£è™·S¯wÓÜßr ü¶¢¼T””€òÐ7nÊ[UO¼}ÕΖ¡ÉáWnzé7Çä—ßÿæwÿ WMVÍ¡'~öüŽë4÷·(¿½(Ÿ*Å€ò PúÆm@yÃá½ÿòRHŠž°á]¯þùÝ,?dµÙ&ûî½ö»äç]û‘ áUNšû[”ßf”ÏH‘åA  <ô—=åÉñòôkþ^˜¢a¦Ý™»*núöŸ_íÂm¿ûÞÛݸëþ{Cîõ7:hîo9P~›Q>;C”€òÐ7^î”'' b_Þ±ã¹ïüÃKW]õÒUÿð<úü㧓}÷\ûöÃMΘÝ6^$ýÞ ®XÞ÷r ü6£|^vPÊCßx™Sžn ÿ‡—ùi#C½†±qå;ùεù ³|ï[?ÚÓ5Œ¢vóÑC^Üqƒæ¤Ÿå@ùmFù¼, <”‡¾ñò¦<Ñ/Júû!¦{s¿=⻯ßYe&Ƽûçë~òÁ¯'Úû´ôÚÅ9ö~–å·å÷æåA  <ô—ý}ùËŸMp6ƒòå%ÁRËTh‰Ùsž÷ØoºÄ%ÚOg»³/zJ—M(7¹Îvnžò¤ï£çÏ‘+åÌ€ò@y yf¶wIX,§U?{þX—¢ÏHà8AÚíú%RÖ…¦!ƒÉÌä~£§ì†žŠÅ4yë*×±ZVÒ1Jؘ“6ëp³'ëîŠ^±ª{­¦@@y <(¿¹”WÔWû§¼0­AGk2])]´‰ŸÑnÊ 3T¦ãŠlÙÅ9ª‰éñºÁbÜ_cšÕVºÓ¶&W Sš}âÕÅÔ~óËS_~2ZŸŽ9ÁÜ ?¦)aÆ—³Ú-£•’¤(W\Øk'zܹu±´*µïÌÁ‚öÊãS'1Q 7åA@y <(¿Ù”onªóKyQa©)“°¦¥iܳïDÅCÓc•,V%×è¨Þ|l1¼í´NuÊKò2’±µœû¥¼o”yþ­·Šâ“ j“¦°ZŒÛiZ™²å Rf¼5CÈŠÉ»múšä ¼ò‚|F«u²Nåò  üVR^*7Z›Ø4J®·’.ÊKåÖ³ó4é-53£-÷º,H2Šå-‡4‚¢Éñ®béÆQ¾!Ù3Y Ö„/òT”×MÒš†\‰À5¡Ä‰ ¡<ºæÁÕK&ÌIÊâ]î¹ø¼ò¾ZHSšMM©À(P(Ê_±<–sÀª­ðK$=Mì5tžÚ093T,ÚdÊcù½ìymË`»ŽXÞ‚·e²bùœ6ƒ<9iU”—Õ-Íé`(¿=)Ö*8.6å2릎6ºæÔauš™³_xž——–ªññÏ 4QzYs{µû´57X–ì^))覎õx(/­ÒOq?\&d•Ë÷¥ ×MyAºÊBtfc®¹tU}öó_Ï°bùµ—+.:L‘½åîûòé5Þ•‹%­ŠòÉrÞž)D€ò h‹)ŸäœcÓ4e' Â<ÔXØ€Ÿ[ õyNZ eû:F­ÌX=‰[Ì&ÃpGÙâ„vYÁ20` ¦§Ð·©©)Ë@k:{’0½^CÍÍÍÐ6Çqã ªÀ=làšéÎÌu?áÌqŠùhV—;*H© N]¸pŠ¬ffúaÙ-ºÙ¯.Ì;s]#Y ­)’žž"FZ*”$òXב+Z_¹îúvŽ“³‚¢ðÏΊ^-ŽûoËȃ€ò èR <ʃ@  <ʃ@  <ʃ@  <ʃ@  <”ʃ@  <”ʃ@@y <”@@y”@@y”@@y”@‘òX¦BKÌžó¼Ç~Ó%.Ñ~:Û}Ñß»9åbéµGÌóz\Ún8T‘¯¹å·¾­Cpþ\±¼TnÄ•CyY½~¼:pn8,§}l¸.mcq¸Eåb¹Ý¤¥#Ï™%GZÒk35gçA@ùíGy8kf„íEù`2Àn†¶¨\YNzrã ³÷“U(”å·’òLN:«Ýn#¬&ãD_m~=‹ò’ܦA‹ÉIg§í&u}+›0]ÞŸ˜;1MÙ)›iÌ@”ˆ]qmû9½pá«SSv§ˆQ+a+;}òÌlï’ °XN«~öü±.EŸ‘Àq‚´ÛõJ¤¬; MC“ÙŠ6zÊnèñŒ¯«\ÇjYIÇ(acjLÚ¬ÃÍž¬»+z%ÌPŒ{œik± ù€yDžIæA@y <(¿u”¦5èhmC¦+¥+‚6qâ3ÚMya†Êt|B‘í"»8G51=^çA—´Æ4«­t§mM®¦4ûÄ«‹©ýæ—§¾üd´>s‚¹ALS"öÞÀƒOaV»e´R’´åŠ {íD;·.–V¥¶ã9XÐ^¡ïŸüòüÜñ…/þíõ=(”å·€ò¢Â>RS&aMKÓ,¸g߉Š‡¦Ç*Y¸L®ÑQ½ùØbxÛië”—äe$ck9÷Ky Þ(óü[o!)Ä'(ÔqÛ«Åh°¦=#åë(Wª0ã­¬{éXn·M_“”Wûcø Üq ÊnÒÃÕ2ˆåA@y <(¿e”—ÊÖ&›¯õVÒEy©Üzvž&½e£ff´å^—IF±¼åÆ@P49ÞU,Ý8Ê7${& ÔšðEžŠòºIZÓ+¸ ”8±!”G×<¸ºˆuK"IR~ïrÏÅäU’0» ?Zá94²:ƒE•˜åò  ü¥Ëc9¬Ú ÿ¸HÒÓÄ^Cç© “3CÅ¢M¦<–ßK—z ”³a»¾XÞ‚·e²bùœ6ƒ<9ˆk$,¯‡ðöKZ=A´gÂ${P(ÊoÝ}ùÔÝÔÑFל:,£N3sö ÏóòÒR5>Þâ™&J/kn¯v?€–¢ æË’Ý+%ÝÔ±¾å¥Uúé#Eb÷%AV¹|_ªpÝ”¤«,Dg6æ%¯ê³Ÿÿz†˯½\qÑaŠì-wß—O¯ ð®\,˜ftÿaÝ_(qÞS(í³™Zò  ÑÓÜzÄ@ø)(”@@y ü†ÇòL(?Ø¥lîgj|ü3âÞÜâTs³RµÄ‚ëz[ÚÜ`ÇǨ:4f7£;•ûFããû•î13Vßæ\nîPuXÐúƒÍ­Ž·‹ù.‚ÊåA P(¿)÷åMCŠ¶þI&ÐîmiéÕ쉕)Ï ÷cyÇ]ó†Ñn*À1´ï³ˆåò (”ßÊÛHëØA•êИ•$ ê6U§ã¦9 ó­Æñq½ã¾üJ”G`?Ü¢ìdþq=38Ыs Óm—²e»²Íéû,(”@@y üæPž™ ¯nUt ™xÍãýÝm*3¾¥½[=åz÷t¨\sì‡ î[íóûÑÂöAÓ"Ê}”ʃ@  ”åòÛ›òXn»aþÂW§¦˜×¹Û™i£íEÒåÉR±œö±áº´Í¢–ÝEÏônp§»Ÿ7¹¾AH*7âÊ¡¼¬^?^-½øõݪr™ä€Ç¦»s|žNÒ"UGSifʾ¼(Ê寀X^Öˆ[êÜÉZ„²ÒÒv0_tq#ºRíÝ’™Ø7‡òÁäáÝ mU¹II‚4uz¬\â Q1P”Ê_”wÄÕ¸I‘æF®+‡}òÌlïòü°XfÓÁd¶â„ž²z*ØYؤ…­G Ên#Â:ÚVÖh=»@ë½RÇ9%©Ð}6Ýœ¶òBYIÇ(ól'“Ï:Ü\à5Ä€å¶OÓ çˆZYJùA=i§fæ5eƒñ9°W¾‹å´êgÏëRô 'H»] dÃ(ÆdDÛ«É8ÑW›_Ï¢¼$·iР›¶›ÔõY¬c(L—÷ã'æNLSvÊf3%bïµ8Zc·£*VÚÀõÔ×믫ÜÀ­\+¤*é3“• X¯ÖJy¬´«h°O™»š ŒcyTu«¶ÂM’ô4±×ÐyjÃäÌ×ñÜÊnýuÆò[?Ê‹ŠOÏ÷äb@y <”÷žcŸZÕOÙ-Ÿcï³7NWYˆÎlÌug¹ªÏ~þëO4'LWLRC5iλ§)•}äüÙ)_€‹Š‡?9¹?‹5oO\t˜"{ËÝwfÓk(¼+KZ/åƒõj­”Oi°ÿ÷7ßœ-_ݬÞâ™&J/kn¯v?€–¢ æË’Ý+%ÝÔ±>¯[ÒÒ*ýô‘"±û’ «\¾/U¸îú®Ðúë*w…Ö_™òÂ̎㟌”ˆ`öPº²)Ï~^Þn§ÈñŽâdÁÒ™ÐÌ\èóΧ˜fu¹»‡Å²ZMÙ)zzŠi©P’çhg½¤ u”¤iŠÄ-“eIÃ$îóE/¢õ܉¥Ï7 eû:ÇIŠ)’¢ð/Æ%a¹:’òšÂÍšÀ½‚Ͻ ô]AJõqêÂ…SÄ`53ó ËnÑÍ~uaÞØé‰Û¥5毿ùšhJ[ýSÌ{ye' Â<ÔXØ€£c©ï̹FǨ•«G^›M†áŽ²Å í²:ƒe`À@MO1M15eh`MgGW6õjnn†¶8ŽU’¨ïJ­¿Ær·~0­$ÌÚ?3§ö5÷N˜Vë(˜>ùù…³ÇiôÉ¢®Hʃ@ íË_<ɬ~æª ÒÖYm™d+ºQÿ^­^’rÝÿ3{0÷2ˆ·³Äû4',Š4áv¨ P]&”e·™,îÞå7ç³[GzŠÅ½ ìÕ*%ÌÞ?5Ó_"În©D…ûû•YÛäJ (.eÊ Óê†,ÎYávkmú¥Ñõ^š^@@yt؃@  <ʃ@  <ʃ@@y <”ʃ@@y”@@y”@@y”@—$åK@ Ð%¯‹Ayq‰öÓÙîìÍ{›˜¸èÐdÿ¾ }+Îz|Þôú–©Ð³ç¦}¾xsÿÕwSÊ÷[ûKý§£ËQj­Ì«(Ê2PŸ%ºâ)ÿ º„5??¿ʳßco# ëh{‘tùKرœö±áºM{g(–ÝEÏltòõø¼Éõ BR¹Wn åeõúñjéůïV•›„åõ›^šÁ)aZ£Þ>XåÌ ÊnÖÛÕ%b <Ú®”OZš“NVÚCÚæ_ÜG\ª] [ÒI Í |0`7C[U.“AE+÷΋òTÍ…žåX~¡Ù'ʃ@ +„òIËòË»²‡Ñ'ÏÌö.ÏLŠe6 Lf+NØè)»¡§"…–I [”ÝF„u´­¬ÑzvÖw,M½.©Ð}6ݼ4›PVÒ1JØlŽÑZ‡›½rÒ%a¹íäôÂ9¢V–R~POÚ©™ùãcMÂ`|ìU€ïb9­úÙóǺ}FÇ Òn×(Ù0Š19éPÁ6Âj2NôÕæ׳(/Ém´Øïä¥í&õ’faº¼?1wbš²S6Ó˜p¨®å•u•¾oõõßúë*7pë× ©JúÌdåŠdµ:ÒwД@Û”òIÂŒ6|¢jIé'ÿ87y¾˜Õnõô¬Ât¥‘¬v¦ÊÊÓç¾<Ö¼<<¨N÷/I*.ìµ=îìªXZ•ÚŽwæ,É/o_ éi}[>S¤Pšž&Âç ½ò›oúò“Ñzç›ï…i úcš ðE»ÒÑÚ†LWJWmâÄg´›ò •éø„"[ä¾PML×¥,yiiV[énÃäŠaŠŸS¯µ¾~[]å®ÐúA´–×wj¡¿ à˜”(·ÝHô•J“’€ò è ¢¼%~) H.P¨ (æB!¨Åh°¦•î$®¢"59è QáÀü´ÊKªÌçO´²Æë© 3Þšá=.€åvÛô5É,ˆP³C2Áª¬W~©gÁežë-’µVTØGjX‰wÅ%š÷ì;QñÐôX% FÉ5:Ê3Ëé´NuÊKò2’±µœ¯©¾Z=å®ØúÁ´Bªjæ¼¥Ú0e)ÆÈay꥞ž(å·.–åu“´¦!W"p… JœXìç¥r£Õ«+vvΤiã’ÊÉÏO¶³:uD=\]ÄÄ$åGñ.ïÙØ+BÄ—ÏÁz€z žƒ%­5áAyÆ«&»¤z«Û-©Üzvž&½e£ff´Þ÷’ŒbyË! hr¼«Xºq”÷W߀­¿žrWlý ZAÖrüs£¿!{Qn«Þ~´1ã2HF”€ò›{_>@oŒå÷’Ã¥^q1…GÈ!ïç™üÅòXΡùO‡‹Eìh΂·e²¢¹œ6ƒ<9i”Ö«µR^ +íêìSæ®fcàXUݪ­ð_S$=Mì5tžÚ093äu<7‡ò[±|àÖ‚ò¢¢ÁÓó=¹¾0.)Üo² Õ\òQ{ãt•…èÌÆ\w–«úìç¿žñDsÂtÅ$5T“æ¼{šRÙGΟòõ¸¨xø““û³X½®¸è0Eö–»ïÌ¦× PxW.–´^ÊëÕZ)ŸÒ`ÿïo¾9?Z¾º»½ˆÌº©£®9uXFfæìžç奥j|¼Å3M”^ÖÜ^í~-EAÍ –%»WJ º©c}^·¤¥Uúé#Eb÷%AV¹|ß2Ä­¾¾+´þºÊ]¡õW¦¼0³ãø'#%¢åw˜J{­x_ùâ}qQG[(¶)åÙÏËÛí9ÞQœ,X:š™ }þ™ãóѬ.OöÜÜÔÚhÊNÑÓSÄHK…’<·@ë<“è%­£$MS$n™Ô(K&qŸ/z¨çN,ê,”íë')¦HŠÂG¼—„åvèHÊk 7k÷ >ô*Ðw)ÕÄ© NƒÕÌÌ7,»E7ûÕ…yc§'n—Ö˜¿þæk¢)mõO2sìqä•$óPcaŽŽ¥¾3Oä>£Vf¬ym6†;Ê'´Ëê –5=Å4ÅÔ”e 5]ÙÔk¨¹¹ÚFà8nTH6¢¾+µþË ÜúÁ´B’0kÿ̜ڽ“ë©ÿwf†²»EM-|NÖJò hûÆòO²«Ÿ¹j‚4…uV[&ÙŠnÔ¿W«—¤\÷ÅÿÌ̽ î÷ng‰÷iNXiÂíP <”¿L(/Ên3YÜ#¼ËoÎg·Žô_üדöj•fé/‘g·T¢Âýýʬmr¥”€ò—2å…iuCç¬p»µ¿6ýÒèz/M¯@  <ºŒGìA PåA PåA  ¦K*'??ÙÎêjõpu;“”Å»¼gc¯_>ëUê5x–´Ö„o寚dì’ê­n·¤rëÙyšô–š™ÑzßwH2Šå-‡4‚¢Éñ®béÆQÞ_}¶þzÊ]±õƒhAZËñÏ+ßqg®¯†JÄ@yt¥Þ—Ðcù½äp©WɆ@árÈûy&±<–shþÓáb;š³àm™¬h.ç€Í ONZ'åƒõj­”ÈJ»ú‡û”¹«™À8–GU·j+ü×T IO{ §6LÎ yÏÍ¡|àÖ_g,¸õƒ ¼¨hðô|OîŠOm³:Ée—´@y´])/L­ê§ì‡–ϱ÷Ù§«,Dg6溳\Õg?ÿõŒ'š¦+&©¡š4çÝÓ”Ê>rþ씯GÀEÅßœÜÏ~žI\t˜"{ËÝwfÓk(¼+KZ/åƒõj­”Oi°ÿ÷7ßœ-_ݬÞâ™&J/kn¯v?€–¢ æË’Ý+%ÝÔ±>¯[ÒÒ*ýô‘"±û’ «\¾/U¸îú®Ðúë*w…Ö_™òÂ̎㟌”,¿Ò’îÛߣ,LvîX -ì4Óýû`Ž=Ú®”g?/o·SäxGq²`éLhf.ôù gŽSÌG³ºÜÝÃbY ­¦ì==EŒ´T(És ´Î3‰^RÐ:JÒ4Eâ–I²¤a÷ù¢QzîÄÒç›…²}ã$ÅIQøˆã’°ÜIyMáfMà^Áç€^ú® ¥z€8uáÂ)b°š™ù†e·èf¿º0oìôÄíÒó×ß|M4¥­~Ú63ÇG^ÙI‚056àèXê;óDî£Ñ1jeÆê‘×f“a¸£lqB»¬Î`0PÓSLSLMYXÓÙÑ•M½†š››¡mŽãÆAUd#ê»R믱ÜÀ­L+ }ÿÌœÚçÜ;QVí¡q‚f•²OöT§cII@y´}cù‹'YƒÕÏ\5AšÂ:«-Û’˜Ê¿W«—¤\÷ÅÿÌ̽´Á±í%Þ§9aQ¤ ·C]€ò Pþ2¡¼(»Ídqð.¿9ŸÝ:ÒS|ñgAöj•fé/‘g·T¢Âýýʬmr¥”€ò—2å…iuCç¬p»µ¿ö½4½€ò è2±@@y”@@y”€ò@y(”€ò@y(€ò (€ò (€ò è’¥¼¸Dûélwöæ½ML\th²߆¾g=>oz} ËTh‰ÙsÓ>_¼9‡‹ê»)劋û­ý¥+¼›H˜Öh˜ûìÿÒÉ@y´})Ï~½$¬£íEÒå/aÇrÚdžë6í¡Xv=ãÉ—¾A;]‡Ï›\ß $•qåÆP^V¯¯–^üúnU¹IX^ϱé¥yØJ®¶*k²zg½¹´)Lo”€ò+夓•ö¶ƒù¢‹Ú“‰Kµ tKº ´”&ìfh«Êeò"¨¨Ócå~ÃyIÉKÿ>É’Üv—<å×q <”_!¿¼+{}òÌlïò̤XfÓÁd¶â„ž²z*RXa™´°õ(AÙmAXGÛÊ­gh}ÇÒÔë’ ÝgÓÍKÓ¸ e%£„Íæx­u¸Ù+']–Û>AN/œ#je)åõ¤š™?>Ö”! ÆçÀ^ø.–ÓªŸ=¬KÑg$pœ ívý’ £““l#¬&ãD_m~=‹ò’ܦA‹ÍñN^ÚnR×g±Ž¡0]ÞŸ˜;1MÙ)›iÌ@”ˆ½”WÖAVú¾uÖ×믫ÜÀ­\+¤*é3“•¾1/Êï6TË–e°ʃ@ +òIÂŒ6|¢jIé'ÿ87y¾˜Õnõô¬Ât¥‘¬v¦ÊÊÓç¾<Ö¼<<¨N÷/I*.ìµ=îìªXZ•ÚŽwæ,É/o_ éi}[>S¤Pšž&Âç ½ò›oúò“Ñzç›ï…i úcš’˜L€v¥£µ ™®”®ÚĉÏh7å…*Óñ E¶È}= š˜¯KY<òÒÓ¬¶Ò݆ÉÔÆ{ŠC01õZëë·õ×Uî ­D+`y}§ú}åžfµÆ7 €ò@yèJ¤¼%~) H.P¨ (æB!¨Åh°¦•î$®¢"59è QáÀü´ÊKªÌçO´²Æë© 3Þšá=.€åvÛô5É,ˆP³C2Áª¬W~©gÁežë-’µVTØGjX‰wÅ%š÷ì;QñÐôX% —É5:Ê3Ëé´NuÊKò2’±µœ¯©¾Z=å®ØúÁ´Bªjæ¼¥Z²|,¿aÌК%\ÜP(A,n’Ö4äJ®T‰‹ý¼Tn´zuÅÎN•ô1m\R9ùùÉvV§Ž¨‡«‹Ø˜¤ü(Þå={Eˆøò9X¯PÏ‹ ÒZ¾”g¼j’±Kª·ºÝ’Ê­gçiÒ[6jfFë}ßY É(–·ÒŠ&Ç»Š¥Gyõ Øúë)wÅÖ¢i-Ç?7.²—TŒ?fwEjfõm9P]±÷åôÆX~/9\ê³!Px„ò~žÉ_,åšÿt¸XÄŽæ,x[&+šË9`3È““ÖIù`½Z+å²Ò®þ¡Á>eîj&0ŽåQÕ­Ú ÿ5HÒÓÄ^Cç© “3C^Çss(¸õ×Ëný (/*<=ß“˜ßËåA +òÂÔª~Ê~hù{Ÿ½qºÊBtfc®;ËU}öó_Ïx¢9aºb’ªIsÞ=M©ì#çÏNùz\T<üÉÉýY¬y{â¢ÃÙ[î¾3›^3@á]¹XÒz)¬Wk¥|Jƒý¿¿ùæühùêfå!2릎6ºæÔauš™³_xž——–ªññÏ 4QzYs{µû´57X–ì^))覎õyÝ’–V駉ݗYåò}©Âu×w…Ö_W¹+´þÊ”fvÿd¤d…+- =IS4EÑSS¶£ª\1P]I”Of´áUK:H?ùÇ™¸ÉóŬv˨§g¦+Ô`µ3¥¸PV~˜>÷å±æåᩨ@½pºIšPqa¯èqgWÅÒªÔv¼3gI~yûMOëÛò™"…Òô4q>é•ß|ëÔ—ŸŒÖ;ß|/LkÐÓ”l%Юt´¶!Ó•ÒA›8ñí¦¼0Ce:>¡È¹¯TÓãu)‹G^ZcšÕVºÛ0¹b˜ÒxOq&¦^k}ý¶þºÊ]¡õƒh,¯ïÔB¿¯Ü³Â”âÊ¢TÌu™¢ÔSêb1P]A”÷±Ä/É µÅ\(µ ¶Ó´ÒÄUT¤&½a#*˜ŸöAyI•ùü‰VÖx½ UaÆ[3¼Ç°Ün›¾&™jv¨B&XA‚õÊ/õ,x£Ìóo½eC²ÖŠ ûH +ñ®¸D³àž}'*š«dá2¹FGyæ1`9Ö©±NyI^F2¶¶‘ó5Õ7@믧Ü[?˜VHUÍœ·T¯˜Ë=HhËÄ@y±ür ˆòºIZÓ+¸BP%N,öóR¹ÑêÕ;;gÒÇ´qIåäç'ÛY:¢®.bb’ò£x—÷lì!âËç`½ @½ÏÁ’Öšð <ãU“Œ]R½Õí–Tn=;O“Þ²Q33ZïûÎIF±¼åÆ@P49ÞU,Ý8Êû«oÀÖ_O¹+¶~­ Hk9þ¹±rå$À©JÜ(—åA Еz_>@oŒå÷’Ã¥^q…GÈ!ïç™üÅòXΡùO‡‹Eìh΂·e²¢¹œ6ƒ<9i”Ö«µR^ +íêìSæ®fcàXUݪ­ð_S$=Mì5tžÚ093äu<7‡ò[±|àÖ‚ò¢¢ÁÓó=¹+>µ!­ûù¯g<Ñœ0]1I Õ¤9Tö‘óg§|=.*þääþ,Ö¼=qÑáÿŸ½w‹idëü^g¤‰òšEyÈy:R4‰æᜓ™/ù”œQF‰’É$£‰”)9R¤ÌhF:Q"EI¤œÁU¾cæbÀ€Á`À`s¿4ÍÅí+¸ ØÆ·ºÛ0†¾Múë¦÷þ¶Ô)c®2v¹°Íe÷÷ÿ=lÁfW­µjy¯Ÿ×ªÿZšØ4sofõc[éB[µ¼ÔZÕky­úý÷ï×Aó懬™C©·SŘ:´mŸ¹ü\Ú/¯ta{s¥4¹~hvq”Û€¦µÓ§Þ!5÷G¥q>tò^I«,áô›>÷• ÃlhEnoÞo¨Ü½_ÛòHûÒÑûÓýoZÚñ=28Û­ä>í^&æÐ#-`y~LË ÷ËSMì-õ«eå‘ÐùXè뛋#:ÿcÜeæFX´Ã š¢™t ß™vWçL¨D¯4Î †¡ ,qàw˜lXŃ^äF×éqùþfD3°¼GÐù"iÛá9®íZ 4/„[À]£Î¢µ»V¦õàg77g¸w4ù†æBÙ¯7¹èriÞ®‹ûþ Ÿ®c§V>ÆckE8÷MõÚ0öY†¹3[ا±Læ×êÙZÇc‘í¥¡»€vÍD$áñDèt*ß©TÂc„³³ßl&ýôéi†!q âÞ£²í­Õûu–+ÞûRz¡éXÍœºŒòÊÏyÒeÓ Ã¤¨Ð2ÿÓ–àœË?[²J¬šLgOfCÊçF«×êá(Í¡Ï¿d_w¡-À3¢ð'ì:äGh XÀò¿ËË ±·Â{ÿå¼a~gã¶4‰×ê †ÕTÆmR‚gŸyïªÛÑñƒ|ÓËXþ%[ÑMø…¨p*é׿Œ¡÷eÖ ÀòüŠWìË–,XÀò`y˃å,–°<`yÀò€åË–àÅZ^a |È®ï41Eßú{ ©§â4RçGo¯(h»=€g¯Ò~œÇÿLí}”rýî¤{Pül"T?ò:D8AQñ­ÈIÀjyá9ö$'ƒ‹}ªû'{£‹»Ûvf(jXa2¥|éMºiu~äöJ@ebŽæX^3ÞU=}{Ÿ«Ü´{ã0]ž‡²Ì„æ{U/û{°<ÍÏI§Ü È×=ò'ɃsfNÿ+s•–—’ö1x®róyfèw»fe•õO:6Óö+9å,`ùÇÍ/_ÌÆœ\d7ïg&EÛ§}‘X<‰á$“¢"ÃZÁà©ê‹Ó‰ãx2¸04•¼ToÓ~~ˆƒ”9u½í­Úû •[£÷%ôÚí<;wWÈ=‹VÛ掑×aöë‘,ô½ìu{°<€å›mù ÿ¦ªdj£Ýaç\ì4ï—ÄUÞç"¼|ÙÈ{=¹tË+-ñëãyÁz½¬ÕÇ拪h×S $Bg}ÃÙƒ "µVU­—À¦4¥_'MÉZ+ïu~Aâ]…ÉÎEßÉû}éÝ.Õc!ºÇ€v.'S»ËVSw›­oå¼®öŠô~#åÖì})½Ð:“¹NŒÞŸÌ£ÆM, FZÖz˜«O8Sšßb+ühìš©Tg©µ±ž­ô°Tã1¬–Ï×jZ#,i2ÉUKeM^æ‚Ig2þ{g™²­ß:·îà4Cì­ô«šgùjííýFÊ­ÙûzA¦›;ú­°d´-fζ‡T¥u¹Ü€åø}//2£=›Äö o^,”@ïÂÇßÏTm.v®ç>l÷Ë…³¹¶Ð.˜Íu¾"#VuKƒ–—Z«z-/Ó ®¸}^§£ë!o~ÅçòlÓ“áê-•)õ:oé¼ÕvññžçãX^¼÷œË‹÷¾ËËû¼ïr]h¥g½ß,-å+GöˆW`yþ0,´ZÜ4µ~?ƾâh¬ŸIàË´øfÙ⤮¿eJ³9Do? }cºÂÛS툓È]¦*m—÷o¿?YΧ}[4±iæÞÌêÇ<4¶Ò…¶4jy©µª×òZõûï߯ƒæ‡Eå±f¥ÞNcêж æòsi¿¼jÐ…íÍ•"Ðäú¡ÙÅQnšÖNŸz‡ÔÜ•Æ5úÐÉ{%­²„ÓoúÜW‚³u i¸½5z¿¡rkô~mË#íKGïwL¿i!í³‘ƒ•žÛzÉ»ãøjÚ–à‡´¼p¿Šÿq®C~„¶€å,ÿ+±¼Ü°Kp+¼÷_Îæw6úO>ˆŠ×ê †ÕTÆmR‚gŸyïªÛÑñƒ|ÓËXþ%[ÑMø…¨p*é׿Œ¡÷eÖ ÀòüŠWìË–,XÀò`y˃å,–°<`yÀò€åË–àÅZ^a |È®ï41Eßú{ ©§â4RçGo¯(h»=€g¯Ò~œÇÿLí}”rýî¤{°òÙDê ò*G—’û1¹+b\–ർð{’À“ÁÅ>ÕýCØÑÎÅÝí‰G;35¬0™R¾ô&Ý´:?r{% ²F1Gs,¯™ 爵¾½ÏUn Ú½q˜.Ï‹P@n\Xê+}™”÷{wŸ,À;—æ¤Ó nä랧÷ƒsfN/kÃòR2À>ÏUn>/ ýn×\ó¨a¹qK˜=,Àmù–{ùå‹ÙØ“‹ìæý̤hû´/‹çó{1)*²1¬LËT½óoqš"qO†¦’—çLx©<õºr8ô1=[žÆ ј–‚8IÞD›Üžåå¤kA»÷‰ôù>®Ñš_‡ ŠÎäŽv§Û)u¯•Èµhç|8{}¸bwFq à Š ¿25ÍbùœtlÁ$žŒE÷ã=“Ë+»¦½ òöL^†Š¹&;ÏÑ[ÝØñéqš¦h2¶!<&ÿAñ² Ò÷5ØÞê½ßP¹â½/­ZÌÅÁH Í£]¯‰àðËN;–°|“-ß‚´-`û–²±¯Jþñü¼©taÇb"XY½#J{G )Åy‹¹úr8ÛZaBå:ç.Kªèݤð .»*ª³¸(l¹³,¿ToÓ~~ˆƒ”9u½í­Úû •[£÷%ôÚí<;wEפŽe|TÕ–àÊòþMU ÈÔF»+ÂιØ)h"!ß1.‰«¼ÏExù²‘÷zré –WZâ×Çó‚õzY«=ŽÍ·!‚y×S $Bg}ÃÙƒ "µVU­—À¦4¥_'MÉZ+ïu~Aâ]…ÉÎEßÉû}éÝÔc!ºÇ€v.'S»ËVSw›­oå¼®öŠô~#åÖì})½Ð:“¹NŒŠÍÓ‘¶ùäÁ„bìa˜ËW±€¼{`ü¶.¥¬8u`øÝ8¯²F“Sš2eÂÆ•#ŸNƒ:k=ÌU¥4¿ÅVøÑØ5%R©ÎRk%b=[éa©ÆcX3,Ÿ¯Õ´Ð7êÉ$W-•5y™c>$Éøïeʶ~ëܺ?‚Ó ±·Ò¯jžå«µW´÷)·fïKè™nîèSTlÉ^¦sÄcÓ­-`yEàú½¼ÈhŒöl c—„è}Cøøû™ªÍåÑÎõ܇í~¹p6—ÀÚ³¹ÎWdĪniÐòRkU¯åešÁ·Ïëtt=$€Q|.Ï6=®ÞR™R¯Sð–Î[mïy>ŽåÅ{¿Á¹¼xïK°¼¼Ïû.·Ñ…Šíˆ&f_~Ä'XÀòM±oµ¸ijý~Œ}ÅÑX?“À— hñͲÅI]Ë”fsˆÞ~@ûÆt…·§Ú'‘»LUÚ.ïß~²Ú!ˆÛSômÑĦ™{3«óÐØJÚÒ¨å¥Öª^ËkmÔï¿¿šöº—5s(õvªS‡¶Mø3—ŸKûåUƒ.lo®&×Í.ŽrдvúÔ;¤æþ¨4®Ñ‡NÞ+i•%œ~Ãm“);ÌÖV¤áöÖèý†Ê­Ñûµ-´/½ß1‰|ÓR[CØb;Ò–ËÀlyá~yŠ¢‰½¥~µ¬<: }}sqD猻ÌÜ‹vØ$CS4“Ná;sÃâêœ •‚è•Æù Á04%ü“í«xЋÜè:=.ßߌh–÷:_$Mc;<ǵ ]K!‚æ…p ¸kÔY´Vb×Ê´£üìææ ÷Žæ#ßPÃ\(ûõ&].ÍÛUcñoß¿áÓº‡Ïó1ö[+ŠÀñ¸oª×†±Ï2¼Ü-çžÆR0™_«gkE¶—†îÚ5‘„Ç¡Ó©|W¤R MÎÎ~³™ôÓ§§†Ä1 ‹zgŒÊf´·Vï×Y®xïKé…¤c5sꋽ“÷oí-t¼|Ƀå,ÿë9áVcKV‰U“éìÉl`èYö4U¯ÕÃQšCŸɾîB[€gD1à?NØuÈа<€å%–—b n…÷þËyÃüÎFÿÓŸO"^«‚VS·I ž}V佫nGÇòM ,`ù—lyD7áK¢Â©¤{\ÿ2†Þ—Y+Ëð+^±,X°<`y˃å,–°0³ŸÞ›(ÉI5ËF¸>ToÓ~~ˆƒ”9u½í­Úû •[£÷%ôÚí<;wWÎ=ËÞýäËõéÑùçß1Ë/<‰XÀòͶ|…SÕ2µÑs.v šˆFÈwÌÝj¨¼ÏExù²‘÷zré –WZâ×Çó‚õzY«=ŽÍ Þ—¢]kdxL-õ kd2ˆÔZUµ^›Ò”~L4%k­¼×Iø‰w&ÿ9}'ï÷¥wGºT…èRÚ¹œLí.[MÝmj´¾•óºÚ+Òû”[³÷¥ôBëLæ:1ª¬ô}jóZo¿!È Ó|{Tsy`._Éòî5‚ñÛº”²âÔáwã¼ÊMò†âÂàLTWŽ|:Y ê¬õ0WŸp"¦4¿ÅVøÑØ5%R©ÎRk%b=[éa©ÆcX3,Ÿ¯Õ´FXÒd’«–Êš¼Ì1’Îdü÷Î2e[¿unÝÁi†Ø[éW5ÏòÕÚ+Úû”[³÷%ô‚L7wô)zÉ1¬`o‡KÿZ3IÌèd`yþPßË‹ŒÆhÏ&±=È› %Ðû†ðñ÷3U›Ë£ë¹Ûýrál.-´ fs¯ÈˆUÝÒ å¥Öª^ËË4ƒ+nŸ×éèzH£ø\žmz20\½¥2¥^§à-·Ú2>Þó|Ë‹÷~ƒsyñÞ—`yyŸ÷]n£ ­~ók­ÝÇÛ°<–GZ-nšZ¿c_q4ÖÏ$ðeZ\ µ8©ëo^ü²Þ~@ûÆt…·§Ú'‘»LUÚ.ïß~²Ú!j}[4±iæÞÌêÇ<4¶Ò…¶4jy©µª×òZõûï߯ƒæ‡Eå±f¥ÞNcêж æòsi¿¼jÐ…íÍ•"Ðäú¡ÙÅQnšÖNŸz‡ÔÜ•Æ5úÐÉ{%­²„ÓoúÜW‚³u i¸½5z¿¡rkô~mË#íKGïwL¾i¡]¯ðäj¯²ðÆaÐIÆæÚa'?ªå…ûå)Š&ö–úÕ²òHè|,ôõÍÅÿ1î2s#,Úa MÑL:…ïÌ ;ˆ«s&T ¢WçƒÃЖ8ð;L¶¬âA/r£ëô¸|3¢XÞ#ò‡—P4íðÇŽÔK!‚æ…p ¸kÔY´Vb×Ê´£üìææ ÷Žæ#ßPÃ\(ûõ&].ÍÛUcñoß¿áÓu¬çcì1¶VãqßT¯ cŸex¹[Î=¥`2¿VÏÖ:‹l/ Ý´çW=žNå»"•Jxl‚pvö›Í¤Ÿ>=Í0$ŽaXÔ;cT6£½µz¿ÎrÅ{_J/´ «™SWåØ;¶ìÑ­hþTš¡ÂkCp*Œ¢ðcÏåŸ-Y%VM¦³'³¡gÙ·\½VGi}þ%ûºë…Çmÿè(üÇ »ùÚ–°ü¯ÄòrÃB,Á­ðÞ_H5Ìïlô+ž|¯ÕA «©ŒÛ¤Ï>+òÞU·£ãù¦–°üK¶<¢›ð% QáTÒ=®Cïˬ€åø¯Ø–,X°<€åÁò–ËX,`yÀò€åË–@ 2EÀò¼Ë#­cž8APdz›Ý?Þ¶sEßú{ ©gõ(LÙ5òÔ×6ÚnàÙ«tÅCŠçñ?S{¥\E¿;é¬rb¢î_öFÈ“‹¬³ËÐ,Ë÷¬á—?ÿt¹ÇË^®²„.úùÕ#=«Z•\'-šÉðÞhż-2­ÅwôõæœÞ_èD[ÐÎ…}úüæË‘Ï¢}9Y@Qà “Ùìiî—´sqw{¢¾ÓW¹¶9¨¬QÌÑËWÿlÏUn>[à ýn×,r²¼,@Ó-m Yß’›·¼ÐžMrw|h»µ|>÷}þ‰( Š¨avO]§îçv©`y´kqŸà¥+KWÛòˆÆ´ÄIòö Úäöl1+™L;æK]|$Âx:}x’Ù_rlŘÔáÉapJÔ¸–SUïü[œ¦HÇ“Á…¡©äå9^*O¯}LÏ–'—»s±ÉWø¸Fk~&(:“;Ún+V«˜‡9¹¨øvC´V"×¢óáìõáŠÝÅ1 '(*üÊÔ4‹å3屓x2ÝwŽ÷L ,¯ìšö&ÈÛ“‚*æšìÜÎÕ@ìZV=Ž(í-$:G4æ-æêËáìýé©Üè:ç.K^*zg®ÉÔ9äà =ùÁQéu ÏJb­ªf§¿¼NÎãGt¶ð¡ßÔŒ`öV!&`k/&še¥d8Ë#m3±£}»AÎ}˜ÙOïM”Þ¸¨ÆbÙÀ×ÁêámÚÏq2§®·½¼ÔÉ‹‰ À«õ—[£÷%ôÚí<;wå`yžÖòr™Î/ ÌÅóK÷ì„þi,O}ÊÑ“»"¸ÿVÖjcómˆ`m` © c¡;Q(MnÜJ8{o!º_' u®qmŸ‹ðòe#ïõäÒ,¯´Ä¯çëõâw¾k2õ kd2ˆÔZUµ^›Ò”~L4%—®¼×Iøé€&ÿ9}Ç~×JïŽ:W=¢K’B;—“©Ýe«©»MÖ·r^W{ej£ÝagÜ$žLD#ä;Æ¡mB¹5{_J/´Îd®£J°<OlyÖjC;§În‡í'³|Õ¹<ëÌÕ'œò(Ío±•Û¸gqË‹_«²F“¼¡¸P ¢BظräàÓÉ¢`P¿³T‰TzVRk%b=Þ“¬ÜGu½„ONk„%M&¹j©¬ÉËCð!éL&Àï,S¶õ[çÖýœfˆ½•~Uó,_­½òî5‚ñÛº”²â„Ûbùš½/¡dº¹£OÑ°<OoùDÛ7bº]2¾gù‰§¶<;oJ` í‚ySç+2b•6—»¶÷ áãïgª6—G;×s¶1ˆâwnÄòRkU¯åešÁ·ÏëttÉ›6—g›ž Wo©L©×)xKç­¶ƒŒ÷<Çòìg—Øä­ŠÜ{ ÌåÅ{_‚åå}Þw¹.,ÀÓ[ž7xñ-¯0.jJiœÛ;û]ZºåU–púMŸ‚ö;ÌÖVDÚ{ù¾-šØ4sï@õc[éâÞË‹Y^üÚDo? }cºÂÛS툓È]¦*m—÷o¿?Yíl§½s–—Z«z-¯µQ¿ÿþý:h~XTkæPêíT1¦m›ðg.?—öË«]ØÞ\)M®š]å6 iíô©wHÍýQi\£¼Ošèg£îöÊô3 |Ù€ã ,Nêú[F0—¯¿Ü½_ÛòHûÒÑû“¢ïx*ËúRW7—ÇtÜ9¨*ŽÜÎ8}tys•ò©Ðöé`*“b˜TŠÜ_›\ >å¢Ë…9a1"9“|}sqD猻Ìj¾½&ýôéi†!q âÞcaj(ÓZ<øÙMÙ~ù¯9Ücá^i#šå=‚Îß”¦±9~ŒýÕ× ê [c¹Ñuúv«YËož|¼«s•kï„3$†&°Äßa²`zÉßü¸|³Èó;šÂ-à®ù¬Dj%v­L;zû$ÏpïmãQÃ\(ûõæ® qpñoß¿áÓº‡ï ÌÇØcl­(Ç㾩^vu΄—»åÜÓX &ókõl­ã±ÈöÒÐ]@»f"’ðx"t:Å^M§R MÎ^õ³Ñ`{Ñ{€ÌɤSøÎÜ°ƒ`kâm¡¨¯\ñÞ—Ò -HÇjæÔU)öÑßÌœ|º¹J¹Š~wÒ=X=]§#Ì•@Ó Ïd‡,@3,ß³†_þüÓå/O¸ÊºüéçKüUôüeÕ²Õh&Ã{£3¤È´ßÑײsì¿ù,ZÙ‹qÏUn>/ ýn×\a:/ïž™í-ý{´g÷—Àò–o<óìõ-¹y›+ íÙ$wLJŠ™góYÖèóODaPD ³»xêì:%)ó,Úµ¸Oð´•¥h«myDcZ â$y{äkr{–Ÿ“îâ#ÆÓéÓÌþ’c+ƤOƒSz¤Æµœªzçßâ4Eâ8ž . M%/Ï™ðRyêuåpècz¶<›Ø‹M¾ÂÇ5Zóë0AÑ™ÜÑîtÂMØn3ž1'ßnˆÖJäZ´s>œ½>\±;£8†áE…_™šf±|N:¶`OÆ¢ûÎñžIå•]ÓÞy{&/CÅ\¼…æÂ7F«;>=NÓMÆv#—Á¸ög£‘ö¢íÓ¾H,žÄp’IQ‘á»4y •+ÞûÒz¡ÕÁ\ŒÔÌ( ܤ,`ù&ä—Wu¯A‹Z¦ ’ìoF^~ù2ùVÈšý(syEï&…opyLQÅEaË\~y_.ãÌ¡êу‹¬k@•Ï»¶ƒs5»–U#J{G )Åy‹¹úr8{z*7ºÎß¹ËÒ„ŠÞ™k2uÎ0éðBO~0GTzB³’X«ªùÖé/…“ï-|è75#˜€½Uˆ ØÚ‹)]YiãÇÎòHÛLìhßnsßföÓ{¥7.ª±X60Âu°zx›öóC¤Ì©ëmoé›Ò±˜ ¼Z¹5z_B/ Ýγs·Qt™LÞµŹDÐ`yË7Åòr™Î/ ÌÅóK÷ì„þi,O}ÊÑ“»"¸ÿVÖjcómˆ`m` © –w' ¥É[ gïí"D÷ëD¡Î5®ís^¾l佞\º‚å•–øõñ¼`½^üÎwM¦³¾aìA‘Z«ªÖK`SšÒ¯“‰¦d­•÷: ¿ ñ®Âä?ç¢ïØïZéÝAçªÇBt)Ží\N¦v—­¦î65ZßÊy]í•©vW„q“x2 åÖì})½Ð:“¹NŒVŸÌ£ö]bÛÚŠ@Œ=MµÞչʵw™ CXâÀï0Ù°Š½äo~\¾¿YäÎù Í áp×|V"µ»V¦½}’g¸÷¶ñ¨a.”ýzs×G…8¸ø·ïßðiÝÃwæcì1¶VãqßT¯ »:gÂËÝrîi,“ùµz¶ÖñXd{iè. ]3Ix<:b¯¦S©„Ç&g¯úÙh°½h‡=@æ‹dÒ)|gnØA°5ñ¶PÔW®xïKé…¤c5sêª{§ž¤w‘¡):u·±,`y8á¶ÁSRlÉ*±j2=™ )_V­ŽÒúüKöu ýœ(üÇ ûCÏÚËX,ßH¹a!–àVxï¿œ7Ìïlô+^X­ºΰšÊ¸MJðìó~ÒzWÝŽŽw)XÀò`ù{O7áK¢Â©¤{\B­€ XÀò`y˃å,–°d× ÈS_Û8h»=€g¯Ò~œÇÿLí}”rýî¤{°ÚÙDˆvhegš¦S©ø›1=–  –ïYÃ/þér—'\e ]þôó%þªGzþ²jÙj4“á½ÑŠRdZ‹ïèkÙ9ö_Ž|­ìÅ85¬0™R¾ô&Ý´sqw{B‡<õµÍAebŽæX¾úgãqÛû\嶠݇éò¼Eä]óo–5…•}ëDr±˜' ,`ùFsÒáAŸ-Nýlœð OÿCr”6;óì˘ÈΙ9½¬x ËKÉû¦gËÓ¸‰Ý¹Øä+|\£5¿ÉíN·«UÌxÆœ\T|»!Z+‘kÑÎùpöúpÅîŒâ†~ejšÅò9éØ‚I<‹î;Ç{&–WvM{ä홼 sMvž!¢·º±ãÓã4MÑdl7BpŒk6i/Ú>í‹ÄâI '™Ù¾K“×P¹â½/­ZÌÅÁH Í£úñmjo¢¸¨–°|òË«º×‰ E-SIö7#/¿|™|+dÍ~”¹¼¢w“Â7¸<¦¨Î⢰åN.¿¼/—qæÇPõèÁEÖ5 Êç]ÛÁ¹ˆ]˪Ç¥½£…”âˆÆ¼Å\}9œ½?=•]çïÜeiBEïÌ5™:g˜tx¡'?˜#*½N!áYI¬UÕ|ëô—÷ÁÉÂÉ÷ˆÎ>ô›šLÀÞ*ÄlíÅ”®¬´ñã gy¤m&v´o7ȹï3ûé;9å?*c±l`„ë`õð6íç‡8H™S×Û^oZ¼˜ ¼Z¹5z_B/ Ýγs·Q^õMÁÂÁéç_~:ÞêWBôM´¼\¦³Gã sñüÒ=;¡ËSŸr4ÁÁ䮸œÚ-²V{›oCkkdxL]°¼;Q(MnÜJ8{o!º_' u®qmŸ‹ðòe#ïõäÒ,¯´Ä¯çëõâw¾k2õ kd2ˆÔZUµ^›Ò”~L4%k­¼×Iø‰w&ÿ9}Ç~×JïŽ:W=¢Kq hçr2µ»l5u·©ÑúVÎëj¯Lm´»"쌛ē‰h„|Ç8´M(·fïKé…Ö™ÌubTt2húvÉ={¼— y–g­6´súáìvØ~2ËW˳~Á\}Â)Òü[¹{·¼øµ*k4ÉŠ Õ *„+G>, uñ;K•H¥g%µV"Öã=ÉÊ}T×Køä´FXÒd’«–Êš¼Ì1’Îdü÷Î2e[¿unÝÁi†Ø[éW5ÏòÕÚ+ï^#¿­K)+.@80¼)–¯ÙûzA¦›;úQJù~å+,€å,ßË· Ú¾Óí’ñ=ËO<µåÙyS[hÌ›:_‘«´¹¼Øµ½o?Sµ¹<Ú¹žû°-ˆA¿s#–—Z«z-/Ó ®¸}^§£KÞ´¹<Ûôd`¸zKeJ½NÁ[:oµd|¼çù8–g?»Äö oUäÞÃh`./Þû,/ïó¾ËmtÕܵt,û%X€&Zž7xñ-¯0.jJiœÛ;û]ZºåU–púMŸ‚ö;ÌÖVDÚ{ù¾-šØ4sï@õc[éâÞË‹Y^üÚDo? }cºÂÛS툓È]¦*m—÷o¿?Yíl§½s–—Z«z-¯µQ¿ÿþý:h~XTkæPêíT1¦m›ðg.?—öË«]ØÞ\)M®š]å6 iíô©wHÍýQi\£¼Ošèg£îöÊô3 |Ù€ã ,Nêú[F0—¯¿Ü½_ÛòHûÒÑûÓýoZªÕ G¯ºpc™ªw9θ Æ€Æ-?êK]Ý\Óqç ª8r;ãôÑåÍUÊ7Z¤BÛ§ƒ©LŠaR)rmrú”‹.æ„ňä|LòõÍÅÿ1î2«ùöšôÓ§§†Ä1 ‹zgŒ…±K¦µxð³›²ýò_s¸Ç½ÒF4Ë{¿)Mc;süû«¯Ô¶Ær£ëôíV³–ß<ùxWç*×Þ g>H0 M`‰¿Ãd;À*ô’¿ùqùþf‘;çw$4/„[À]óY‰ÔJìZ™vôöIžáÞÛÆ£†¹PöëÍ]âàâß¾çuߘ±ÇØZQŽÇ}S½6ìêœ /w˹§±Læ×êÙZÇc‘í¥¡»€vÍD$áñDètŠ½šN¥› œ½êg£Áö¢ö™/’I§ð¹aÁÖ8ÄÛBQ_¹â½/¥ØújæÔU1öNÞ1¾¾‡3ùGEÑÔÁƨ…è;š1—4¶d•X5™ÎžÌ†”/«VGi}þ%ûº …Î~Nþã„ý¡gí€å,–o$ŽÜ°Kp+¼÷_Îæw6ú/¬VÝgXMeÜ&%xöy?i½«nGǃ»,`y°üƒ½§›ð% QáTÒ=®G¡VÀ ,`y°<€åÁò–ËX,`yÀò€åË–,X°<`y˃å,ÿX–GZÇJ¹Š~wÒ=Xãl"D79ýøÿ1Sj°<Í°|Ï~ùóO—{¼<á*Kèò§Ÿ/ñW=Òó—UËV£™ ïVÌ"ÓZ|G_Ëαÿrä³he/Æñ¨a…É”ò¥7馋»Û:䩯m*ks4ÇòÕ?ÛÞç*·íÞ8L—çE¢¶l'Ö‡¦“¼¬7`yË7š“úðø¬¾0¦!úÙ8ážþ‡ä(mvæÙ—1‘ œ3szY ð–—’ö1x®róyfèw»æªÓy¥éMÂ= þ–°|ã™g‡¬oÉÍÛ\YhÏ&¹;>TÌ<›Ï²FŸ" ƒ"j˜ÝÅSg×)I™gÑ®Å}‚— ­,E[mË#ÓR'ÉÛ#_“Û³üœt‰0žNždö—[1&uxrœÒ#5®åTÕ;ÿ§)Çñdpah*yy΄—ÊS¯+‡CÓ³åiÜÄî\lò>®Ñš_‡ ŠÎäŽv§Ûn5ö6ãsrQñí†h­D®E;çÃÙëû3ŠcNPTø•©iËç¤c &ñd,ºïï™X^Ù5íM·gò2TÌ5Ù!x†ˆÞêÆŽOÓ4E“±ÝÁe0®ýÙh¤½hû´/‹'1œdRTdcø.M^CåŠ÷¾´^hu0#•5/ïY‹ïŒjÊÿ¿ËX¾ ùåUÝëDТ–©F‚$û›‘—_¾L¾²f?Ê\^Ñ»Iá\STgqQØr'—_Þ—Ë8óc¨zôà"ëPåó®íà\ Ä®eÕãˆÒÞÑBJqDcÞb®¾ÎÞŸžÊ®ówî²4¡¢wæšL3L:¼Ð“Ì•^§ð¬$Öªj¾uúËûàdáä{Dg úMÍ&`ob¶öbJWVÚøñG†³<Ò6;Ú·äÜ÷™ýôÞDé‹j,– Œp¬Þ¦ýü)sêzÛ[úæ…t,&‚¯Ö_nÞ—Ð h·óìÜ])÷,Ò1Ù›º}I– é–—Ëtöh|a`.ž_ºg'ôOcyêSŽ&8˜ÜÁý·²V{›oCkkdxL]°¼;Q(MnÜJ8{o!º_' u®qmŸ‹ðòe#ïõäÒ,¯´Ä¯çëõâw¾k2õ kd2ˆÔZUµ^›Ò”~L4%k­¼×Iø‰w&ÿ9}Ç~×JïŽ:W=¢Kq hçr2µ»l5u·©ÑúVÎëj¯Lm´»"쌛ē‰h„|Ç8´M(·fïKé…Ö™ÌubTy-߶™ï@*|ûËX¾–g­6´súáìvØ~2ËW˳~Á\}Â)Òü[¹{·¼øµ*k4ÉŠ Õ *„+G>, uñ;K•H¥g%µV"Öã=ÉÊ}T×Køä´FXÒd’«–Êš¼Ì1’Îdü÷Î2e[¿unÝÁi†Ø[éW5ÏòÕÚ+ï^#¿­K)+.@80¼)–¯ÙûzA¦›;ú½¿d¯Þ=:¤ŠO‘>ýú5^(.€å,ßË· Ú¾Óí’ñ=ËO<µåÙyS[hÌ›:_‘«´¹¼Øµ½o?Sµ¹<Ú¹žû°-ˆA¿s#–—Z«z-/Ó ®¸}^§£KÞ´¹<Ûôd`¸zKeJ½NÁ[:oµd|¼çù8–g?»Äö oUäÞÃh`./Þû,/ïó¾Ëmt‰ïÚ€¹<byÞàÅ·¼Âd¸¨)¥qnïìwié–WYÂé7} nØï0[Ziïåû¶hbÓ̽Õyhl¥‹{//fyñk[½ý€öé oOµ#N"w™ª´\Þ¿ýþdµC°JôÎ X^j­êµ¼ÖFýþû÷ë ùaQy¬™C©·SŘ:´mŸ¹ü\Ú/¯ta{s¥4¹~hvq”Û€¦µÓ§Þ!5÷G¥q>tò>i¢ŸºÛ+ÓÏ$ðeZŒ+°8©ëoÁ\¾þrkô~mË#íKGïwL5¾iåhžåG}©«›Ëc:îTGngœ>º¼¹JùF Thût0•I1L*Eî¯M.PŸrÑåÂœ°‘œI¾¾¹8¢ó?Æ]f5ß^“~úô4Ã8†aQ05”i-üì¦l¿ü×î±p¯´ÍÀòAçoJÓØÎ?Æþêëõ†­±Üè:ý@»‡Õ¬å7O>Þչʵw™ CXâÀï0Ù°Š½äo~\¾¿YäÎù Í áp×|V"µ»V¦½}’g¸÷¶ñ¨a.”ýzs×G…8¸ø·ïßðiÝÃwæcì1¶VãqßT¯ »:gÂËÝrîi,“ùµz¶ÖñXd{iè. ]3Ix<:b¯¦S©„Ç&g¯úÙh°½h‡=@æ‹dÒ)|gnØA°5ñ¶PÔW®xïKé…¤c5sê2ÊÅ6êÏï)ö>§‘Å.X±à;œpÛ„SRlÉ*±j2=™ )_V­ŽÒúüKöu ýœ(üÇ ûCÏÚËX,ßH¹a!–àVxïÏ­ ó;ýŠV«‡î‡3¬¦2n“<û¼Ÿ´ÞU·£ãÁ] –°=NÓMÆv#—Á¸ög£‘ö¢íÓ¾H,žÄp’IQ‘á»4y •+ÞûÒz¡ÕÁ\Œ(+Z>öá„¡šÎç~|;Ó¥ËÐ4ËUÝëDТ–©F‚$û›‘—_¾L¾²f?Ê\^Ñ»Iá\STgqQØr'—_Þ—Ë8óc¨zôà"ëPåó®íà\ Ä®eÕãˆÒÞÑBJqDcÞb®¾ÎÞŸžÊ®ówî²4¡¢wæšL3L:¼Ð“Ì•^§ð¬$Öªj¾uúËûàdáä{Dg úMÍ&`ob¶öbJWVÚøñG†³<Ò6;Ú·äÜ÷™ýôÞDé‹j,– Œp¬Þ¦ýü)sêzÛ[úæ…t,&‚¯Ö_nÞ—Ð h·óìÜ])÷,¢íékE‹_SaÚUL’–°|,/—éìÑøÂÀ\<¿tÏNèŸÆòÔ§Mp0¹+‚ûoe­ö86߆ÖÖÈð˜º`yw¢PšÜ¸•pöÞ.Bt¿Nê\ãÚ>áåËFÞëÉ¥+X^i‰_Ï ÖëÅï|×d:ëÖÈd©µªj½6¥)ý:™hJÖZy¯“ð ï*Lþs.úŽý®•Þt®z,D—âÐÎådjwÙjênS£õ­œ×Õ^™ÚhwEØ7‰'ÑùŽqh›PnÍÞ—Ò ­3™ëÄhÍô€h×k<0¤ËÐ,˳VÚ9ýpv;l?™å«ÎåY¿`®>á”Gi~‹­ÜÆ=‹[^üZ•5šä Å…jÂÆ•#ŸNƒºø¥J¤Ò³’Z+ëñždå>ªë%|rZ#,i2ÉUKeM^æ‚Ig2þ{g™²­ß:·îà4Cì­ô«šgùjí•w¯ŒßÖ¥” ÞË×ì} ½ ÓÍ}ŠŽÔNÜêÀ¢VX€æY¾Ñö˜n—ŒïY~â©-ÏΛØB»`ÞÔùŠŒX¥ÍåÅ®í}Cøøû™ªÍåÑÎõ܇mA ¢ø±¼ÔZÕky™fpÅíó:]ò¦ÍåÙ¦'ÃÕ[*Sêu ÞÒy«í ãã=ÏDZ<ûÙ%¶y«"÷FsyñÞ—`yyŸ÷]n£«æ® ÕÈ[jˈ‚åh¢åyƒßò sᢦ”ƹ½³ß¥¥[^e §ßô)¸a¿ÃlhE¤½—ïÛ¢‰M3÷T?桱•.î½¼˜åůmAôöÚ7¦+¼=ÕŽ8‰ÜeªÒpyÿöû“ÕÁv*Ñ;7`y©µª×òZõûï߯ƒæ‡Eå±f¥ÞNcêж æòsi¿¼jÐ…íÍ•"Ðäú¡ÙÅQnšÖNŸz‡ÔÜ•Æ5úÐÉû¤‰~6ên¯L?“À— h1®À⤮¿esùúË­Ñûµ-´/½ß1Ýÿ¦¥ß#ƒ³ÝÊâS·x™˜£K –°|#–õ¥®n.é¸sPU¹qúèòæ*å-R¡íÓÁT&Åäc÷×&¨O¹èraNXŒHÎÇ$_ß\Ñùã.³šo¯I?}zšaHð¨wÆXÉdZ‹?»)Û/ÿ5‡{,Ü+mD3°¼GÐù›Ò4¶3DZ¿úzA½ak,7ºN?Ðîa5kùÍ“wu®rípæƒÃЖ8ð;L¶¬âA/ù›—ïo¹s~GAóB¸Ü5Ÿ•H­Ä®•iGoŸäî½mJ¹Š~wÒ=(~6ªy"Hœ ¨øÖ䤠aË÷¬á—?ÿt¹ÇË®²„.úùÕ#=Yµl5šÉðÞhÅ )2­Åwôµìû/G>‹VöbV˜L)_z“nÚ¹¸»=¡CžúÚæ ²F1Gs,_ý³ñ¸í}®r[ÐîÃty^Êþu2šïä\ËX¾áœtxЇÇg‹)°ýlœð OÿCr”6;óì˘ÈΙ9½¬x ËKÉûË}þ‰( Š¨avO]§$ežE»÷ ^‚¶²mµ-hLKAœ$o|MnÏòsÒ]|$Âx:}x’Ù_rlŘÔáÉapJÔ¸–SUïü[œ¦HÇ“Á…¡©äå9^*O½®}LÏ–§q»s±ÉWø¸Fk~&(:“;ÚæFîbÆ3æä¢âÛ ÑZ‰\‹v·³×‡+vgÇ°üboø•©iËç¤c &ñd,ºïï™X^Ù5íM·gò2TÌ5Ù!x†ˆÞêÆŽOÓ4E“±ÝÁe0®ýÙh¤½hû´/‹'1œdRTdcø.M^CåŠ÷¾´^hu0#•4¯ÞK¹zåð^€Ç°¼QÕ½N-j™j$H²¿yùåËä[!kö£Ìå½›¾Áå1Eu…-wrùå}¹Œ3?†ªG.²®ü L¥yçj v-«G”öŽRŠ#ósõåpöþôTnt¿s—¥ ½3×dêœaÒá…žü`Ž¨ô:…„g%±VUó­Ó_Þ' 'ß#:[øÐojF0{«°µSº²ÒÆ?2œå‘¶™ØÑ¾Ý ç¾Ìì§÷&Jo\Tc±l`„ë`õð6íç‡8H™S×ÛÞÒ7/¤c1xµþrkô¾„^@»gçî ¹gÃJbÛÜ1ò:Ì~]"…>îûXÀòM°¼\¦³Gã sñüÒ=;¡ËSŸr4ÁÁä®î¿•µÚãؼ`ùíZ#Ãcê‚å݉BirãVÂy;B»_' u®qmŸ‹ðòe#ïõäÒ,¯´Ä¯çëõâw¾k2õ kd2ˆÔZUµ^›Ò”~L4%k­¼×Iø‰w&ÿ9}Ç~×JïŽ:W=¢Kq hçr2µ»l5u·©ÑúVÎëj¯Lm´»"쌛ē‰h„|Ç8´M(·fïKé…Ö™Ìubôþd5nâ±"´”·;¢qSÜê=XÀòÍ°õ©¹?*kô¡“÷IýlÔÝ^™~&/Ðb\ÅI]Ëæòõ—[£÷k[i_:z¿cªøM iŸ¬ôÜÖKÞµÇW wËX¾ËúRW7—ÇtÜ9¨*ŽÜÎ8}tys•ò©Ðöé`*“b˜TŠÜ_›\ >å¢Ë…9a1"9“|}sqD猻Ìj¾½&ýôéi†!q âÞcaj(ÓZ<øÙMÙ~ù¯9Ücá^i#šå=‚Îß”¦±9~ŒýÕ× ê [c¹Ñuúv«YËož|¼«s•kï„3$†&°Äßa²`zÉßü¸|³Èó;šÂ-à®ù¬Dj%v­L;zû$ÏpïmãQÃ\(ûõæ® ë¿ñoß¿áÓº‡ï ÌÇØcl­(Ç㾩^vu΄—»åÜÓX &ókõl­ã±ÈöÒÐ]@»f"’ðx"t:Å^M§R MÎ^õ³Ñ`{Ñ{€ÌɤSøÎÜ°ƒ`kâm¡¨¯\ñÞ—Ò -HÇjæÔe¬ºœ¢ê_ åwÐtÒ;Ù†Bôßá„Û&œ’bKV‰U“éìÉl`Hù²jõp”æÐç_²¯»PèìçD1à?NØzÖXÀò`ùFâÈ ±·Â{ÿå¼a~g£_ñÂjõÐýp†ÕTÆmR‚gŸ÷“Ö»êvt<¸KÁò–Ë?Ø{º _¢N%Ýãzj¼PÀò–ËX,`y°<€åÁò–,X°<`yÀò€åË–°<>«/Œiˆ~6Nøƒ„§ÿ!9J›yöeLäçÌœ^Ö<†å¥d€} ž«Ü|^„úÝ®¹æQÃrãV){.XÀògž²¾% É­ÑžMrw|¨˜y6Ÿe>ÿT\=NÓMÆv#—Á¸ög£‘ö¢íÓ¾H,žÏîƤ¨ÈÆð]š¼†Êï}i½Ðê`.FjhízM‡•ð^€¦YÞ¨ê^'‚µL5$Ùߌ¼üòeò­5ûQæòŠÞM ßàò˜¢:‹‹Â–;¹üò¾\Æ™CÕ£Y×€*Ÿwmçj v-«G”öŽRŠ#ósõåp¶µÂ„ÊuþÎ]–&TôÎ\“©s†I‡zòã4¢Òëž•ÄZUÍ·Nyœ,œ|èláC¿©Áì­BLÀÖ^LéÊJ?þÈp–GÚfbGûvƒœû>0³ŸÞ›(½qQŲ®ƒÕÃÛ´Ÿâ eN]o{Kß¼ŽÅDPàÕúË­Ñûzívž»¢ËdHÇ2¾_z¥–°|,/—éìÑøÂÀ\<¿tÏNèŸÆòÔ'^È/â¨EÖjcómˆ`†³F†ÇÔË»…ÒäÆ­„³÷v¢ûu¢Pç×ö¹/_6ò^O.]ÁòJKüúx^°^/~ç»&ÓYß°Fö ƒH­UUë%°)Mé×ÉDS²ÖÊ{„_xWaòŸsÑwìw­ôsÕc!ºÇ€v.'S»ËVSw›­oå¼®öÊÔF»+ÂθI<™ˆFÈwŒCÛ„rkö¾”^hÉ\'FÅ&óHÛ|ò`B1ö4Õò¬Õ†vN?œÝÛOfùªsyÖ/˜«O8åQšßb+·qÏâ–¿Ve&yCq¡D…°qåÈÁ§“EÁ .~g©©ô¬¤ÖJÄz¼'Y¹êz ŸœÖKšLrÕRY“—9†àCÒ™L€ÿÞY¦lë·Î­û#8Í{+ýªæY¾Z{åÝkã·u)eņ7Åò5{_B/ÈtsGŸ¢bKö2#›æuå¢Ë…9a1"9“|}sqD猻Ìj¾½&ýôéi†!q âÞcaj(ÓZ<øÙMÙ~ù¯9Ücá^i#šå=‚Îß”¦±9~ŒýÕ× ê [c¹Ñuúv«YËož|¼«s•kï„3$†&°Äßa²`zÉßü¸|³Èó;šÂ-à®ù¬Dj%v­L;zû$ÏpïmãQÃ\(ûõæ® qpñoß¿áÓº‡ï ÌÇØcl­(Ç㾩^vu΄—»åÜÓX &ókõl­ã±ÈöÒÐ]@»f"’ðx"t:Å^M§R MÎ^õ³Ñ`{Ñ{€ÌɤSøÎÜ°ƒ`kâm¡¨¯\ñÞ—Ò -HÇjæÔ%{'ïßÚ[~±ËXN¸m-Y%VM¦³'³!å˪ÕÃQšCŸɾîB¡³ŸÅ€ÿ8aèY;`y˃å‰#7,ÄÜ ïý—ó†ù~Å «ÕC÷ÃVS·I ž}ÞOZïªÛÑñà.ËX,ÿ`ïé&|‰BT8•tëQ¨ðBËX,`y°<€åÁò–ËX°<¿v˳ÿôù|ðOø'üóeþ,`ùÆ-À „ýß,`ùF,)¼dÀò–‡¹<À\,`ùæÏå=ËmÿÓ_üæüGÿþ¿ó'ÿbâÍ7ÊnüËôW›M½åà?ûwÿä¯í¶ ³Êä¿ÿÍŸü[¯eÙûTÿ™Úû(ån™ÿê·eÚ¨öì\óºÿå/þìOÿìÏÿüOÿü¿øW=+˜ËÐœ¹üý­¿ýoÿ7ÝëwÃÍëÎÿêþößú£ßüoãÒ­½5ð—¿ýW,ÿJó_þÃzů+ùwÿ¿óÇöÿoÛßÛÿóÿìÿοùwÿ²ónx{‹¥ÓÿÇŸþÿ|¼¹_^ÞØþ¯ÿöÐ.zžúÚæðºí?ÿ­¬9–¯þÙxÜö>W¹¾7ÿò?ü{msUüÛä_ÿæ?û?'¶òOÖ»>ô/þþ?ü›9Ìåh|.ïþ«ßüwùÿÅßÌÆ4ÏüßüÅoþÇú›ÿÙì|È<¥²å½+èoÿ¡îu5chÿÓߪÖî~]Uüö/Z_¿ ‰ü¦éŸýñŸþ¿ ð–ÿl<ÏUnþÿ­ÅýgÿÞ?é­´4ôJõŸüF¶Â=W·ýÿÿuçÌåhÆ\þ¯~û¿öµýÓ¿ÿÏÇò’~cýÿÛ;ó &Òô·0(€DqT„@ E¼a<ðBÅÄaÆQYÅ T<(‚‚‚ "g *®$TMÕü±U¿ªÝ­©Ý­Ú­Ýª½~ èñû)Š‚N¿Ý~»ßO?Ýo¿Oœ`åâ8‘Úò’ûß…ó's‚4ò­yxd¥Ð×ÃaNRIÝÀ–'㡈 ÷ñ,ß ÂÈc5¦[^Zzë»UÁ‚À  aPàÜõG²Ë¥š›ç7úºL. öñ™íéqàøÎ0¿9³=g¯:]\;@Ym;_ž}(Rè( ­:(>%rv÷[và~Ÿ‹”ÊŒpWŸ#}:cKÖ|åÉN”g%Wv/ ð½¦ÎZqª¨V×ȧD‰ø3\¨žnß*#e%ù‡–MwðþöDìÂ`‘(8( `éÞ[eÃtS½®*ïÔ¦¹A pn؈˜”œ3ú–¯«¼2:T@î A?`Á¶³›Vû8u«hæ´™>þ|Á‚K„›oV™xl¼Ë÷­)<¹iIØü¹"¡€ï+XòeïÝïwZ¯ñÚ7©êž÷s]œ^Ñ?×%­XµDu¨HŠR7Ï_—V‚X0<±¼hgvyn¼pÕ¥2iYú*Áλåww‰zbù>ò-û!LtÂË¿c,_ug— 8^¬m kž\Ú&%äiÖPµqªWì­òºú² ‹]¦o»A¶¦WÖ ãµ7ו­—_è}ñ©Z½ÒgWwú9ÙxyÚÏŠ’»ÛÜÝ·f~'£KÖ}åw>ßgÙÁÜJr™µeŸTÕ ¼¯LÜ*šý\vŽoÃ[u¶¸F#ª3KgGi}ú®Áç™pþÚ3…šJŠÓ¶ gNäë,_[t4lÖ'V×i¯ŽEø,Oé}âRž6}Íe­éêÊ2¢øQ7ªSñûÖI{v\mÁ÷¡‘—+ê‡c½úµ_W')¾°5 äÛÞÚ7¡$¹±ý*ÝÒ³ã›5ÙÞŽ¿?_BÕÇþ„Q`y`yºXþ®DúäÄÂùo$ÎWݺ—ä¼'Ë 8SùA:ød¬WúôÄ|Ñ¡"ý8º&o·`YZ™Æò[çiŒ.ÉÞ9/VÝhJî}=O³ÍÆËJnoŠÖïOW}g‹‡…å+.Íw˜yÈà~½ñ%ë¾rúÆŒRé  bêVÑZo^ÈéÒÞφŸ(†ÉÙ1AÄú‘gÕ­ “}Žj,_¹ÉgEºÁî²óáü]=ýjò÷Ïõ]±?ífnÑsIÝîœéûJ˲Ol_,‹B.Lâ×;`‡¼^ÃÚ'å»qcÔí8þÒJM¯…ºz9„^¬ ¸žJ^ZL®²®ú᩵ÂõŸõ‰åXXž.–Wݬ¯¯ýŒ7eEz™ê¶ýoËK²cEÛ² {T^Y-úVÝïÙ¸å—-K]8÷Ô³>›t´¿å+//âx~o tãK6U"TûÊÔ­¢·žþž¤®£!<„ï·U¤ºæiûØ—§‰¸½i*ý½¼Öè?w–Ve¥&ÆG- æû —'d•× —åi¿¯$wwCr^…ºæêjŠ‹„ÃbyÃÚ¯ªªRAÁ“¿Ì”H¥&Ö‚ôIâ,ÎÂþ·ìk|+ZÑ3¹®4eɼ£š‡Eú±ü€ÀòÀò4±¼úwÖå[?ª Z>äÝ,Ÿ2ËKŸ&…†|d5ï,I}nJ,o¬¬${‡pãubùš¼ø©¼õ}/ù],oêV ÕòÒÒë [7FÇ¿_=ÈX>Ê –¯Ô‹åkòöŠÖfÐwV—V>~R)ÕÛyÉ‹¼6õîOãÇÆ¿¯$'N¥¿'ËRÂB’úZ>eH±¼~íWTTçŽÇñìÙ3))z,/¹=iê—ý»ÙKrã ¶ºübDð÷…Ò¾±¼ƒãŽ=°ü±¼~›£oùÊ«‘~[nª ºŠ»‡WxØù$™jùúòKK}vÜ®Ò5ûWÓnü(5í¹üíþÂ]šîHªGÂç·ø÷<5nyãe륓ó7¦=Q·´µ%—c…S¹¾±|}uæz7Ͻ/M]ò;XÞÔ­¢åëJ’c Â!òJù "|éÓ3ᾫOhv«¤(%Ê‹kßó¾|]ùõí!Ëïêz ÕIŠÅG¾¿ø¤¶çÑÅ´hq™îÃÊìoügÇêiF!_éã£ó‚÷?Ðx´öÉ¥Xƒµ—~,ÿëÕ¯ýŠŠgäôÓO?M`³£¢¢JKK¥¥g²¼ôÑÏÝÖݪ¦z“no°ho¶úmzéók±a‰j ŸË¹'oüSX€<–wï<Ó~ìµ2Í»º×cçûî<ÞÉwÓEMGªšG'#}½æðý|}#¾9û»ÎÔ…û5]¿¥OΩz$«ú$;Œw™Å'ÿÌß~åyoÓV[|6ÊÚ4/~`°H²0úè]Í° Ò’K[‚=Æ÷y_žå¼åÒ3m“+}v#aEj˜ÿµ‡õûØ;1];®•×I²·ö)kÆó:ɽ]3&öl3MY­p*îŠ òãû‰Bm8~+y‘è…åëU Ÿ¹;ß0ð2²dÕ Aþz]¸ƒõ;p´¯Œm•±²Ò’‹›U{rŠ0ú‚êËK$†Og÷ÐÕ‘æyù|kÂJxrðo€«úØoñùþapðüM§³ÏˆœÜùË´$÷ÆU"¡0(P( ¶$ꀸ§÷]é¹%¡›·,ñ÷™Cî-þœ9ó¶œéÓŸæØxÇï+)HZèG®ÒÏgNðºÃÇ…N“ùáz¯P m½}jŸ<Èsç¿ÿý¯\.·³³Ûµh½ÀÝÚz²‘Z¨¯-ØçõÙ6ê¾wäÉÅ U£âøó–~£ÝýŸËÓò°<°<Ʊ§%%yn0õSlé“s§¯WŽ¬­‹ÿWÃí-¦ï¦ Íûýë_ä‰&“ÉlllâââÊÊʤRÚ+©ÊfÎKì•VË÷7µ‘`y`ù~{Ƀƒ æéîðR|zh]|VÕÛªARûpŸ¯×Ö›•Pó°[þÿøΓçZcc#“Éܽ{wyyy]åÅYuöÞ­Ç W¥”}ì)yXX±¼^ž²1T$ ÌÝúCq ¶ ÅòûÛß4á< 9‘}BB½è‡BËëŸÚtÓay`yÄò¼£åÿò—¿üóŸÿì9+++IÑ$EO~ª9ç×E7ö]Ÿ@––§´| `ð”––’gyöõçIîÞ½kkk›œœ\^^.‘HÞq]t–Ç·ÀòÆ-x~ÿûßÿñÔ„óÿùÏf̘1NÇ5õ"Œp ,?XˆLkk+)ߟþ™< Å­[·þýï_¼xqòäÉGŽIIIIKKKOO/..noo  ËËeùŸ~ú©©©iâĉ,ëOúÓßÿþw;;»¤¤¤ºººæææ–––¶¶6ò–Àò|p–¯­­uvvž7ož——×Ñ£GÉpþرc~~~¤åGˆÜM±üÿF6°<¿‰å9ÎÂ… ³²²®^½jkkûË/¿üùÏ&ãz±XLFñ„å"°<ïÁò_|ñEvvöëׯ[ZZARRÎÇÄÄ466*•JXËðÁA†ê»wï.(( m®P(ärù;w˜Lfhh¨½½}bbâ›7o`y,À‡©u2~'#úžçïííí×®]Û¼yóÉ“'KKK;::Fr›€åtq:éwýhü›4;Ú“Q<ùLj äay`y¨‘=ÜÀån*hÞMm–¿&ò—wö,ð°5#ÂÌš3m^|ÎSß¼’å¯å¼›vµyCÎÙs¹ rÔ,€å)(^¦‹¬­øû²kÞv´5J]M:•ošåå]Š÷lyË`y0Ú‹¶sÇú^jìÿ‰²õIÒŠvæan;}ùñÇo•¤”ï­tä.ݵ–?Å•ã¶hg[5²9ƒÅb9…œ•ÄØré•-¾œqcˆ1æ–ìÙ±…2rZGMÆ+ƒ,béüÕí—ò–ÌņK *¥wI »c¯ÞŒðø¨ éî®l[ނħ2%ª–Àò ¯Ëß\_Ê2sYôõ™›¯;zeùöÞzÃó«ÜW]]¯î}=“á¸.§EIêÕ†`“ÊÚÈ•F"qò'çHñ‹.Ò÷Í•… ]ÝmÅqîVSwܬm—Ë*/FºØŠÎ¿.¢åmVعêò2¥$aÓç´÷ñay,ú{^Vž±g…ÿä dTmï½1­R•ÓDvoÕ»ð;oµÊ¿û…½ýò©× vw[{]Ncù¶‚h.‹¿çféëN¥nÞõŽ¶K2[´¡~M’'+(ãáú—¢µ<›y_é·Çð8º,€å¥îÛërü­Ì=Utu7‹¼Ý%š:K÷Lbð3šT·Ê?)nÐòÝŠ7Ž¬ñwc!X¡_ݬ—7‹æ„…­ƒ£;†²ô­áú•¢µ¼#/æ±f3:Jâ'±—õ\yXËjš¯c.ÍnÕÄòÙýcù^½jî°s7>4Öw®óeÁ!ÁxÛˆœ—y«W÷ ¹i–ÐSª–‡å°<*•);ãOg—H›::›«²öø1,f—1ôÛœµŽŒY{òåòÆü=^ ‡Õšçòú–ï(ùÊ)<ßпO~{ùå ¹UÍ]Je‡äü"{öŠ\™ìQì$¦gÌõj™¢[.kø1Süà•Üp ¥`yX˃¡"oÈÚî3‘9FÕ×ÝÜÎ#ôKq­æ6½òmñшi6æaf3-âHQ‹¶½žå»;«’—¹Y‘e­æœ¨Ñ¿½.+Œä±ÌUË´šwã…êÃÎq\ˆÓŒ ÆX²=Bb®“S –ÐBU –‡å°<–ÀòXË~¥6á`´ËúX0š€å€Q,ÿ>#¦Q¬o/úwcx—6LÛô Šë]ˆ9˸~׿—bÌõ}ƯƒåÀò#ÛòÆŸ›|\–—w)†lù¾eÀò#Eñ„£¢ï—ûUåe×ÈCqBî8‚;1¬7©k¿¼±}„øòæŽ@ž*åÍ8®ÿq\ky—û¢ý8 ³1VîKŽ9qrÏýõ×kR†\ESþþP7Gשž³Wo±{úØS,–X°<Ë–,XŒPè’Ë µÀò`ÄÐ^¼cÒ”/|•Êö=ÿºw€xyõaO‚`„‰ß(‡hyº%€åÁojùβ±ön¬ñ‹/°<Àò`i¾Äœ–X­iF–·š3Q=€¼z`çÅ1‘þžŸMrvô0–4V›­5<>*hº»+Û–· 7_í@–ïxºëS†ï‰Ì]<†ï™:íx7t‰bi¦Ãò˃ÁZ^•€õTe'oW'‡ØN¤O«ÉôÊ ;WÝA:·$aÓç´´oxycAzzQS»4+õz•nd¹¶ÂÍN–¾çêe%ñ¼q3JÔ…èÅÒM§\2X±¼½Í’ÛÚ40­w#&I«ÊÃŽÔæ}i/Žáq"d˜é¥5wõ„q~ç^(TÃéó>ñ8PÞe$Q,Ít°<´å ’ÆÆKK?ú½1”M™KYcÕ’WgÍá™ñ¾*é0’(–z:XPÓr+ÔÆã@E—æŸÌ,—ËÛYÏ7is7ÝcI;$Ë+¯„2‚0g©bœù·ó¶¢6ÚD±4ÓÀò€š®òŸ±üNJHÍwTœ2çËÛŸLÿ2÷µ\Þx?~º¥±¤±C±¼¢!5`œÕÜJë_h¨-H˜fÎ^Ÿ×J“(–6,XÐض)oO×–ã6Õ3xSœˆ­×ÇÞeÅ>Ó’ÆÁòòÚ“Þc¢òZõ½Ahe÷źD±´ dÀò` Í:X–`yX,X°<à£äà@Ý¡î`yÀ·ç_«ó:Ë_ˆ£Ü,#R«;aù÷VwJÙóó1ó¦N°PÕ ƒëýEBfݯµÿayðÛX~¬wjC—¼£¹B¼ÅÝÌ:Lü–°ü°žfe‰~ ÆçÛ/=})ëhªÌJÙ[¸Çä·(õ,ßU›éj912]Ú…XþýÕºj,<Ö¼WÙ(“½©ÎOݵ("µNËÃò£ÊòºP¢«âÀsÞ7Ïšt‡¢üåÍ<ÕEî8®ÿ±öØo¯Jß8ÑŠœjÎò?S­n•:j2v\äDK×à¯n¿”cçX^,6ŠÃÆ3Sê{Î eóݵl‹YÇ$rí XW•ád9)Jübäœ7ƒå5U#8+5ÜíJê¦O•©“4ÝÝ•mË[øT¦ìT“˃ßÔòŠ·Åßy[X Ó_Éz-Ÿ›’þ¨^&—·”œ ³c†f3Ê wð,'m¼\!S(ÚžÜ+VÝzl+Žs·šºãfm»\Vy1ÒÅVtþ<`y•îGÚYø¥½TêO»i7. ½±<ÍœD|6cêÖÌW#ê”ù,OU5j¨›¾{+mVعêÒë% Ó˜>§Õ—¦7‰°<øm,OcÆZZŽCÖÓ"S*;¨/8Û 7»8¯Ë—ÉòÖ°mg6ž,ùëm—d¶hÏš$OVPÆì^Ëww·\7~ñVƒ¯$þÓOf«‘©O@3Â|Æž¢VåˆÚuƒå©ª¦º¦Oey6;ò¾L¾Çð8êÑ$Âòà·±¼ÅÌã%õ oZåÚV¦çPT4^5‡gÏÏd±¬ÌæÒìÖf±€ÁûºÄ°{ 9Ñœ°°upÔà`Ç°ñO–"˜°¼.`<ÿªo,?VËõ>–3m,3(±D6‚DÿñÄò†U£†ªéSß±çŽ픔MåÑ$"–#Ìò²üõNŽ+o7‘Ê®E² kõ¡^¸ƒÇ˜²Y\%S(ÚžÞܨêUò(vÓ3æzµLÑ-—5ü˜)~ð ¡<€å5§Yé÷Þ–ê>ö¯Ú:š«nïÙÂÛz¯Y¿=yŽ]â}b¿ ¹²–ou×Q–èË7-*9¯ºI&k’*{Åu‰.l3uú `yË£î`ùÛïÿÓã=¹–°1 µ:ÏûøSq˜ÇþrհͽÏå ©¶ž‰U¯ÒCXî»KúŽÉMÐ6ÀÊ}ßsí¬²¼5®ZÐôÓ)¶°$ƒzf°üè=õÜ´¹)3ÉÊ¥'g1?ß9Åó@EgóÍ….þ;—q¹ªEª×ûNÙx%Ôþ³osÎXŸP_èCÐöÞJ¶SÔí›Ä"k'm,O3j éfËÈÓ[ž.“,9‡›¹™™s4)sEý™9 sÂjþ¦î>–ïV6e.`ëÎf¨Rh÷['UB[eKÎjŽCxFƒj-åGü,Gµ é¦So!ÝÌ`ùQêzcóÑçR¤Î$+»·Ê–`,¸Õ¬šãùÞI1;¹^ÑÏòÝÝ­9‘vÄoM_¾>P'´U4åïustê9+põ6»§=ÍtÊ-¤,†‹®ŠÌ€TÝ u`y0JP6çlr±_|ýû`y0zP4¤ ¬ ×¥g«»°7–,X˜Äp¦˜y«ËÌÈrΞËm«’Îp~}ÿÖòïg«Àò£<¨‡åÀò£Lîí6d.f$ ‹årV5” E^WÕ:.«làXÓsã•êŠ Q³ìÍ ÂÖgWÎõ«ñré•-¾œqcˆ1æ–ìÙ±…2jËS&Ÿí[–b«Àò`ÐÁuŸ¨™2¯+iy[ÂÚ/!¿±«­â”9–35üä-ò·Å žV“÷”vªƒo'çHñ‹.ÒÙÍ•… ]t–ïŸ|–¢,by€åÁp[ž&ó콕ì¿Ð$vU¾¾8gœK¬fl\yåï¦2…כɫƒ‚h.‹¿çféëN¥±Õé£K>KQ–X ·å©3϶èzOÎ2~úa‰ú.º¼6ÉÓšŸ¡ÊZ£xóàÈ7Ö‚åúÕÍz9õꨒÏR”…å–Ãby'îƇºXž2ó¬ÌËëè|YpH0Þ6Bù÷]eòYŠ²†[–C±|GÉWnLáy]vʼ®&X¾½üò…ܪæ.¥²Cr~‘={E®Œju”Ég)ÊöÙ*°<‚å»;«’—¹Y!«9'jä”y]M°¼¬0>Ç2'ÂÜjbPÜ4wì©’ÏR•í»U`yÀò€åË°<`yÀò€å0¼–Ç~ðƒüà?£õçÿg–•$ endstream endobj 363 0 obj << /Type /XObject /Subtype /Image /Width 675 /Height 562 /BitsPerComponent 8 /ColorSpace /DeviceRGB /Length 53139 /Filter /FlateDecode >> stream xÚì½XWÚ÷í}ö{ßo³Ï&›ìf“l6»Ùìn²%m›" Ñqï-v;všÄéNs@•nŠM³0`$:ØØÓ;’@ef¤Û€q)ÆN\Þ3*HIˆæBîÿu_\ôsŸsFç7÷™3sf¿øü ÿ}îùçŸyöÙÿüûßO¹1\ÀÀÀÀÀÀÀî^û÷¿žúÏ¿ÿùÌÓÿFdG|Ÿ3ûù_xöùçŸÖo]µr±U[mÝ–¬^eÅÖزÕz[jnkíØš¶ì%;¶v¤­3Ør+öÒH[?ÒVXØ: {Ù¦­¤l½É^qÄ^^¥·W±W ¶á•ÕÙ«”m4Ù{¶²×¬ØZ+¶‘²McØK&{í¥ÍŽÚºÍ›Ö½î°½l3²õŽØ›È^7·—íØ[z{c´½2ÚÞÖÛ›vìÕaÛ¢··± [‡íí±íƒm|gËø ¥;Ü‘„†½ÚB™Ñeùm ³]VºÂ|˺Y«»õ¨3³JwìRÑuu9~)¢ë–ºz»ÎÑ/ÂâB™ÍŸ’þçfý—8ê«ÿ!ñc7¶ ú&ÂÁöd¸ýq¤±nÙiÍ›MCCj£™Ñl«-[òÑM½u"è`1"vˆ³–2KBÙfÙš‘¦C¡ PZ¥*E[k¶…l=ПþÏ?Ÿ}æ߈ï³_|þ¿Ï?ƒÖ,\à63ÌÀúgÿóŸEË æû‚Í$Ó³þùçžЃÍHÐÿëŸÿ@çÏóqÐæÍõö÷÷ôóõôõõðñaiÞÞú¿îȼ¼X^^Lô×Û›åãÍö÷óœëï¥?§ã€9núçõµc‚²ŸŸ‡¯·»‡;éæB§9Ó\Ð_´ŒÖøz±Ð¦æãÅBæøîéÆbÒÜèÎt×9®.³Ñ_´ŒÖxy2Ñm€žøŽø6.sôå}<==Ü4§Ð—W¶±'ø½¡ÝÛÑ_´ŒÖÐiNh«¯{Ø|èQ ïÅb³è4×9k׬øø£÷‚„Ü„øô-£5h=ÚŠv»CYïïéëãáç?¡cýØL—Ù³\P)Àeöó2ôSpqž3›’“ Û]ØžÞ,š‹q‡¯—·›Óì9sèl¹ÓzýLæ›Y GµcˆÂÜlwÆJf7ï­+‘Û†"Þ Ûr9ä-ô-£5h=Úª£6B<ÛÛ ™»—'‹É¤yz²¸œ»wEDG…íŒ  ¿h­Aë=uûøxXï ùû±=˜®®Îs m§³3ƒÉôõórü c'áëF£{,X¾|‰{Ügöugz¬ß}Lr\&æ-c°¼§Ò1°;ÚüØ,æÂãsZ•òã¤\Ûz³„Á`ølN«íèDkŽ·glŸï±q׺>gû|&ÛÃz®ŸI]À```3Èô Gµc¾¾l¶Û:_Öyá[?6}°n´¡õhëjo77ýÓy//www"~hˆ "BÄ  ô†–Ñ´mõòD{ÒõOí1?o7:kéá)e-]¨å$;•õyBÞ\Äfúøz:x’1’@¤ç»¥ÔÉ˃Ó˜^ã;Ü›N_Ìk¤—ùÐ n~~SãØm^nÌWêŽëª^«”tcâ<]vëÖWµidûÞðZ±»E÷oñ6ÍÍw®ŸI^À```3É=B0“á¬þì¥o¾\uñã…¶lð«Õø'«ÝèNúN{/&æôõŽí!Á>oû—ó%'ð‹À€ÏÑ_´ŒÖðy_…sw|õ ÚíïÏ>LºÏû›äú¶Ôd5œ%t†7z¶'ÓÕØMêÌ`¹It7álì˜=ÇÙ™áîîg¸1ððaÓiÎú sœ\hlo7š×{e†vÒ…ZïÂðA{ú²Ù W§á“м|=<Φ®4¶uBO–Çú/ÃBß]Îvóòµ“ô›€óÃGQ[ç8ÿäV2e~”.iÃÖ9s\\™,Mß1ve²ü&x~Ç|öÕífLÝÉÙ™n8­qÜaKF»Ó\ »ÍqaêS±U6ü÷e1k’¥è²‘ä½ãá„r¥»ÑüvTP—hƒp1Mçƒé¶ðƒèü‚$ÁzwO=ÄG¤Îò°¸~ì—õrðó±qºÔ=| ûY€ÞÏ×Öùúxx°Âå—¶/9·•Lµ‘^³ÆI¿Œ¬á%çÎWiúe´`9µ¿·—;‹E[¹b1 Þ¾þ,àëí_ïøô«/?þò‹¾øü#ô-£5h=ÚŠöY±b±;“fÇ £±Ùtï÷Dm:¸—§~ñÊR?6ÛËå¦mŸ½îïÎfûù°˜4ï×øI•íhMKþ¾¯ÖxÐÝ||‘?K9éímúØJÓR¸ÿËl&ËׇíÍr›ûvHA]+uZu›¬|÷_WC;i´ÊÏç¹ÒX¬ïJ«hlÒêVbU‘ëÝXw´wvèvkm)Žùt‹åɤ17äâhMýžµtW7[I[æn"ΣYLº÷faJµŒ:JÝœ›ôé2w&ÓÑ“û¸³GgŠIó2îÎr_ÊIÍ«okÕužtÈ+RvòwæÕ4SùU”ç…lðb Ç{~ÍŸ½YL†ßÂäŠ6)•œ¦­£®0[°ÂƒéÆ´Y ;laîî öK_ì>ܨ»´°Ú#‘kÙnn6ªÃv™{Sõžƒ›]6ñëÝ—„7šß”~êã¾¥ˆZn\FwcYMiº~ [UlóÂðñ½ùL÷´r©Ó<¼=øéÝ­¦=j`혛›kÃ[óνãyæM²ck]<ùHÕÊÐrÝê=õHÙgý&´ÚíïåéΠ»¼÷Þ[;¾ú‘ÙçŸ}ðÙö÷·ºMoh­ÑoBû =tgûnPæÅ ù}\¡k¸ª¾žçBcxyQ£þ<Ùn,6Û×›íÁpߘPmhÙ4†&®6t9‹Îp5´œ¨}kÇÕúMM©™ wtÎJuk´íÉÞLçaÐk¤µ’ÖÚ¦ìmþ šé$¤„@ 5‹\¯$×wa-òö&>ÝæðUtW3Ð;¹ÙHù<œµ‰9Ïv÷ ³^«Ò­‘*$ú[ ¶Ü÷½Qä;î“gŠÎðòÖÈ4KZ¢1‡–Ð/´çlE.¸³Çu~š#>£sz¾™R¯ÛÚÙQ_ÞŽé»»ßf»Òlצ›C»³}L×ÛÓùÒÎ2Þj º!i YLwµ‘#“m«Ì™¦Ô»duè²i+ Z5 zE=Z#©Ý÷®—èL«©ÓÜL×Íuìä¬\Õ£/`:ÍÚ¥îêáí1öO ì®5#èÙ¶ÌÏǃNs:ùþü3¯»lrFvv›o͇«<ùð>Ÿ.zò‘Êm+Ͼï§ß„ö9ñþ|´¿·›î:A\õO?yoØ>ùø]óõ;|þÙ‡h;nFò³]ÝÖÅÊ©æ«"l)ÍÍÓÝbtà¿]wÐ"þÐÏuŽë’¹T‰elõp1#/m¶Ë¼¯RÍrß:7WWƪˆv]¨õå<ª?Ó•áéÅ0ëùtEëœi oõ¨“ÌÑuÂz{³X®NNÎt¶ßüÕ_éR/åÌwq¶úQIÓ©· 'å<ÍÕÕÿS}¢AËKÎó¶çQ‘lÅö¹® ïñŸÜ˜)ã–Yvò|C¬cVþ±·“³ß—Gtzú˨ ]??NsÄgO:mÁ—5:re¼íé<›¶<üØ(Ð.†‹#ÓÜ,Kþ£2 å’⯖2¨w†›n»Äœm—¹…cè²qrveÐ  /}×­prvqsaš@à;:u7óE“³,šëè ؃fåR÷v÷ëwvW›ôú§êVÍ×׃æ:çü6ßþW_¶w<ö¯óf?þ`ÒZÏw½L›6¼xæ=_´?Šè]]fñÅG}¸Ùǽƒè?ÚÐz´õö|þù‡h;n^Öc» ƒ>Þƒ5b+sS:Ä)’72\Ü=.ž[P;×ů¦91ÍC$Úª˜ŠEo¸»ÐÌ—šõÑPsU†ðƒe,˼D é’ðr3kWé(šöA”gønÙ•'·0P‰|s²zˤ×&é¼³3sSš†5b¡=æ%7W¯qŸ\—)3³Ì²‹‹ß׺Ç͵ó\]èF7ÞtG÷5ŸßÛÝÅ!ŸÝ™›ºsnpsa1è6@?²6rؼä‡ýQ¦¼Ætv÷¤ªÕ‹m'G®¶ý·¸Á£ Ó‹éfzWwtŲÜÍAïê<:uo/†y.Lnd9X»€­\êîLï±~w```wµéAïíånËü|Pdåtü ÷þõÏ÷­{VoÕ‹þ¹è©Gö¬ñ\ðä# þ9¼¾ÿåç»^gQ½7áþƒ÷·¾¿í-d¼¿eøѼÞôëÑúmºÞ߶íoÇ ƒyÐ]ç~V­kè*¿šëâŠb–—'˃Í`2™žÃ-¶ îæ3Ý‚•4sj0¼Üé4¯5ŸïÊl2ô0—E­b8{½{T×N†.A-6Ë‹z[Ђztww/–+ûít5Ú(I üà£Ï#j4c~DÒƬyMÌy'·ÍºgßUw…FG,Š÷ú–›çøOŽ2eQàž#@ïý™.f¬ãPÜ4wÃñó{¹;;ä3ËÅ°›bït•ÒlƒÞö&›[”¼ÑÊyg6µ’åɲSb4›þ3é#®oO¦Í ô,êûQ, Ð;NÝË2¢w09Ërp¥¾€ÑÉ­]êt6kìŸØ]kFгl™î9¾kÙZ—¾õÏžZó/dUóÿ±èɇËß^xú}¿šw—.øÇÃåóþ®ß„öA{º¹¹"Ð3èÎ^]÷î;o¼³uó{ï¾ñþ¶·Qä®îõQ<â;Z¶¢}О »n {§BÿP²4éÓu‹¼XL–ׂ՛?yw±ƒfìäl ïnÍ£ZH<ógg›M"íéáëî:gÎl'ÆŠZêäÍ!‹QØ–G=Õ•älñpB›\ž Ëfœ…R\Ä?J§ìc?§Ùs/g*Æzº×pÖ<èqÞÅ…¶àëZÝCÛ¢¯–±œP&æ¸zøû»»±=ÍÊÍá“£LY¸'Ã1лŒãü4G|öbÓh‹t±gö‡ótï7S†)fpŽÞTò(!ú€:ÝXÊœçÓæÌFî°<tÛ9rµåÿ¨+Ý)Y‚žéíÅ´=ÍÕZênƒ9KnT9Œ¾€YV/u:ÃÍkìŸØÝjzÐ{y2m™/5Ô¹Ã÷E°Ÿ\ñd-+ž«x{Aß&´Ü»qNõÖÅM+ŸÓoê[÷Üç>³==ÐQžî,ú‚ù¾[·l~ëÍo¿õZ@XßöÞ›ˆïè/ZFkÐú7ßØ€æÍõa³vÜ6OwWæò€2|TO¦bßëî(îs-±Ö8®É°©qçJ&õ­^óNN‹h”±&ZFÊ5’cuµºÐIV´ÍËÕÅï³Cú‘`X§DÛ¹†aÑSJg1©¸oC‚B—Š¢¦ðHi>v×½Eøé9œ5¦;}üλÒÜhž[Så ÑP¶g¿æ†˜â5þ“ÓY–îaþÈÛèi¬qœßéˆÏÔn>ˆZGV4•œÅXGËÚtuÈaó’G,¦{¿—Ñl>®ûл^.6rÄ Û,s¦EqQ™õ`0ÌAÏt£>ílþŒžn=uw³\¸:”ÜÈ ÃuôlýR§¹¹y:ðÓ»KMz„fÛ†‚zO7W§¦ÏžZùÔ‰¥=õÒÓ§^}-èíÔ+/œZ÷ µ°ê©¢ÅÏ¢vÝ×Ç™—'›NszõÕ—ÞxýÕÍ›^~ýõWôÄßòökè/ZFkÐz´uíÚtªßžm× “y0]ó7 Sreú—Ú:°¦¢œÝŸ¼äÍpGw.tŸ7‚Sku›´íÅ8ë½h46Û¢Itµ@ƒ û%aqƒþÛ;r­ä°(x£/ƒÎbÓsߊ-ªÖ=íP”E¬gY>e2Ñ-‹Î^¸÷X»ÌÐ+ëŠcÞ›ëj ô–IÓ<̳æ>çilî÷FPJE«D_Šzqô¶“9“S™2?Ší(è=Æu~}F»¹ÍÛ”Pp¸¸¢8#ïµsWþfwšíquÔa‹‹ŠAgÌßžV#ÑÕ£¦¥>íýù WºÕÙñßmTfÙ waºyz¸1ÍAOcXM}Ûü<%7âÂ`[¹€Ý­]êîŽýîÀÀÀîR3‚ÞÍŽùx³½c¶—'ËßÏ b=jüYLúÆ /½¶qþï¦×Ö#¸£¿†5¯­ßðê:¦› —Ó¾æNg¸:>'N}BÄÉ…æêÎÖmb3˜ôáÏ¥8¹2èúCØtÝ—OæÐt­;݉ÚÁ5øn,§áoŽ8¹ÐélÝ!lšñå³ç8»°Ø#O¢;­›îó,³®Ìvrqq±§í¤GæküÎ5ì* :ÓmÂ'u”ùV†›î[+Î †-7=¿>{¸3|ø³ / >,AwY¹bñ«¯¬}åå5/¯_­7´ŒÖ õ(uwÍËËíî6¦`?'c0\|?/£¢Q¼£ 7t\—Ó,”˜ƒ¦=Û1¦ùùz¢hNs~ŸõŸ|¿'äóC”GÑ2ZƒÖ{cy½Qßáô¥X¢ušëoöÒ¥ó×®]þÒK+Ñ_´ŒÖÐisPÔï­û*¾;‹îˆ`?£Ñó?Û_ØÖ!Õ"¦½tïU/®000°q˜ôtGÌÛ›…B{O&ÓÍ•AwBøFÑ2Z£ûò¶×Ó³Þ×ÇÓÇ‹PîFw¡é棧Qóѻ觼A7lŠò4}û9îâ4üü9ÎNt7(000°ñ™ô(švД½¨ïÑé'ÃòÓô7Ü{{èŒííøîîéÁÒŸ“åpê```````ã2=èYLØÌ3#è]ÁÀÀÀÀÀÀfžéAOÍ‚6ãLz000000°lj@3Qzô ô ô fè;²×>ú§—s:mn½ôè£ëÅS›ªüÇïñ{þç÷<÷E™txùƒðUvÒšO@ º£A¯(zë±Y³fýã£#JÃUùçÿBkÙço è;,ýãë2¥„å2Ör ³~ý­6“:¸êa¸=@ ÐÝ úÿ}ìwÿØR¬Ð­P~ÿŸôè‚-kÿþÛcÞÿ¸Ï—E(yI²ÿƒh¯Y÷Üwß}°‚k0¬&öÕ9ÿÿ¿˜õ‹_þêÁÿ¾žÓîŒÊ·¼Á?,×Ý%™-ƒ@ t§€Þ5úàûÿüóÚƒêNñ+?¹õànw=è¥k¾çé­šTª¦Œwž¹ç¡ÕéB’¾ò¿~áãœvLUŸüÊÿ3ëa=è;óßüÛ¯ŸÚ´·ZŽu”G-ÿÓýî;ë1« ïÈÙôø¯žX·«¬Ç; 3ò›j­¦eõœDKÔìÿ}VXGEñæËÃiY=ÿðVë~f,ýí¬û¼C*è.¡ä£Ýû¢ Ñ£ÿùãòøzt߀µ—‹sЂ,mþ½¿bí‘PÛÍ—A ºc@ß\öù3Z–^`åcÿùähó^=è;2–ýþû¥ú]¥i‹~÷»ÅéËüýât™nÑáü+=è;®yèþ¹ÉÈaUœ§ïcìnµúŽ¬•þÖ?¹ÝþÖÒj²zα@oõü¦­ÖýÌXúàƒË3;ôý ù›Xÿ%è;Å/?zŸó¶½‡›•\Z »H\4ã4Nз#à½ðGæF÷?>¿£“@ßïzÏãï–†é)o{âçÝUñ®¿þÛ¥Êah?ü¨ôíñ´_Îú?÷ÿá!½þðÀ=¿uÖH­€íyÏãïO;¼ÒjZÖΉz«ç7ßj園Œ¥=¾¹@ßé®(ÙòăóÓd£@¯Æ[³w¬tùë}¿˜uß“ž[÷Öa@ÐÝú› hfi¼ WcuÁN¿œõK]µÄ<¢_˜:2¢_úà#k³õ±¯º-ÞýÑGôY+zh…!&…×ýPLm%¢™V“Õs:Ñ>¿ÙV«~Úý#®c lAûÍý  Ð]z(hFþ¢½šè¨<¸ÿ`…n Ú^ã3úôUÝóܶ¬ k9¸íù{þ°"]‚KÒW<ü‡»©ÕGw8ýŸYéŸÑwä¾þĽOoN¬ìÀÕXGCqr|v“ígô÷üã•øŠ—7e´PÏè­¤EX=§CÏèGß´ÕªŸ6@¯(Ùú×{Ýv6èGõËîŠÉÚœhc0"fyä+ÎþjÖ¬YÿßoŸ\¬uo--kçôVÏo¶ÕšŸ6@¯VVçÿõ׿˜5ë׳¹U’œ-ôÇïû%5˜ÿ×f¼¹§ƒÁx =º£Aô @ô @ô @ô @ô h¢ Ÿî÷3Yz@¯ýtÇKúbK^cŸãzèg út«ªªj4èÑ@èô Ð ½þ— ôzh†Þ|2 = Ѓ@3ôæ?v= è¹;EMxï¥îœÐ[Õñ›Îõ–Fqnu:-é SNw© ½pIñ.  4… 17¥yP ôGôÂôNL<5 Δ¶¥ íîÂ.lmÈ›ZàÞ®t…©R,7,ðš>Ðþ½èA ýí}ˆ“ÞúÖñv¥  M#èG‡óæA=€ÐÛ=?:»N¥Ñ¨q•¼³½òÀîL Ð b²k•To´Z£ÕÈ«2#-z£9áéÕØɾ“ݤ†TË[eDM"_1ǶÝC—OwiôÂ[r"¹æ‡†eÔ£dµ§.ô–ï⎷󥽃ÇKD•8†á„F#=œ(´xÄ]'“+T®Övide)!ÆÐ|Réê6'µàj]ÿ»ZÕ+ tÔ+zù¹SZRK’Ú®.usN HšJÐÛúÉèA ½mÐs²$Ú¦¬z|s·ñ“çµFÐs"rä'ÚEQ¸ó£sÚ»Û2B†cVaš¼·i_á¿ ”²~|‘5Ovt4pƒ2Èζd†sõlÎ’¯O´À¦)næD*[ö ¦"]~\¹/K2Ü7pÃöWi°âh®ƒ^qBöŇr ÷"b)Y•¨M èm…óÃA=€Ð[=/®’¨OXŒR«0Æã%Ôu·î³ fPš„,ßÍr‹U]­Å鉻"‚¸ëB· z%v0Øôo¦…˜(+ª’¡¸[«”2õ­8d Ò )°üó÷ܘRµ4-È!¯Fˆ9‚7%éASzûß¿Ѓ@zë ¦wª²ƒ-›©"  ¦«.ök s©Éžž¦d³;ƒ@ADBzÞÑzNj‰¶’áÔ>+ÈÔ)~@Ž #•·«”ÐÖgÅ Ýb ŸУÛ¬*Þr¤¼ ¹+1ŽÎ·çÕh…Š±Ît!À 4eƒñì@è'Ñs£«šRl£*PÆ7ëCÍêè©KàM3蹻ˉ†½f‘ò(ÞN"¢Wb;-"úèÃjYzPÀ@/Ü׬©ˆå¬@zt;ŸÑ‡fIºš†Øq#2ê{.~gz^¸· kË3 Hã…'å¦ßJ ‘}µIAÆ‚ØRòxe¬ ôÂýÒîcñ|ã]AdrúžPΤAž£Ä‹£¸†¡uû+5ƒWz,"ú‰§Ë¯ ‰òdã3úð´+‰á:ÒÏr MÝ’»K`¸éÙ_«•‹Ã9À*€ÝNÐèGÝcZ-©!p\Qw0. »4 •?ö ÞSÔ¢¢:í L©ËŠ’†‡¸gÈ”552²» Mvu)k²,¸pÂ3ëɾ¾­Ç0¬³6'ÖØy`ûN~ºp‚¤UÉ:ˆ†¤Öে†N㵩ÔÀ?nTž¤÷òPgqŒá~$RÔ¤¦’Ôvwáy)by,)ŠáM.]c~‹Û’Ú@’X£é&gL¯¨’̬ìÔïÖjµ]IqB¼iЃ@ é½;‹nô @îVÐ#Ê{z0ô €šGhF‚QÞÇ›  ôÐ<‚@3ôˆòsý¼ô €šGhF‚Q~Á<_= ‡æš‘ G”_¼h€ÐCóÍHÐ#Ê/[º@è¡yf$èåW­X ôÐ<‚@3ôˆòkW/Ѓ@zhA  zDùõ/­Ѓ@zhA  zDùW_^ë(è¹;EMxï%Ó·î§]üĦs½¥Q·ü“ðÓ˜.7|ß ¡Æ FQ±f¯èA Ðt‚Q~Ó†õã‰è…é˜xj@œ)mKµO:ntakCFØÔ÷v¥‹$H8ªî‘äÇ á3÷ ýÌÔMÐ$×Ït€QþÍnè™.v:t»Ò à'ÖtËs"`Ê:€~Fƒ aÂx‚B˜Ð#Ê¿ýæ&» §f¯Si4j\%ïl¯<°;Óô‚˜ìZ¥šš½N£ÕÈ«2#yæí'<½;Ùw²›Ôjy«Œ¨IÔÏË)l'º†.ŸîÒè…·äXLmg˜GN{êBoùˆéb¹ÑùÒÞÁã%¢ÊNè>péáD¡Å#†ì:™\¡Âpµ¶K#+KžPoRéê6'µàj*Ç„ZÕkš¢wL¯)m]Uq<À@ èoèåßÙò†mÐs²$Ú¦¬†ù_·ñ“çµFÐs"rä'ÚEQvñ£sÚ»Û2B†»¥…iòÞ¦}Æ9^ƒRÈú=üñEÖ6ç£'8Û’ÎÕ³9Kz¼>‘o¾ƒ©kœY¨lÙ'˜Štùqå¼Ì8/7l•+Žæ:æ'ªDÙ¹ïˆÝʦ‚xèÀèô ý4ƒQþýw߶ z^\%QŸ$°è}®0Æã%Ôu·î³ fPš„,ßÍr‹U]­Å鉻"‚¸ëB· z%v0Øôo¦…˜(+ª’¡¸[«”2õ­8d Ò )°|‹žwnL©ZšäWÜØr\ÞAHŠ(ÀsBöUk  ЃôÓ zDùÞÇ&è…éªì`KÄfªè…骋ýZÂ\j²§§)ÙìÎ P‘žw´^†“Z¢­$A8u Ï 2 8 dž‘ÊÛUJhë³b†n1†O èÑmVoÙõ.HnÆJŒ£óíyEõöœnHšú;dêâH =@ è§ôˆòŸ~´m‚=7ú°ª)Å61áa|³>ôЬŽžºÞ4ƒž»»œhØkÖ?‚·“Šè•XÁN‹ˆ>ú°Z–àèQiV(Êc‡ó/Ø×FŽЃôz€~A(ÿùöí<£Í’t54 ±ãFdÔ÷\üÎô½poÖ–gÆ OÊ-L5¾•""ûj“‚Œ±¥äñJè„û¥ÝÇâùÆ»‚Èäô=¡œIƒ>0± úý¨{L«%5Ž+êÆea—´Òâ]z`s‚÷µ¨¨N{S*䲆¢¤á!îÁ2eMŒìîBG“]]Êš,‹îœðÌz²¯¯G«Æ1 ë¬Í‰5vƾS£ß‡.œ ©EEU²¢!©5øé¡¡Óxm*5ð•'é½<ÔßYc¸‰5©©$µÝ]xc^Š˜@KŠbx“Kטßâ6‚¤6$ÖhºÉÓ+üI¨·HRU›˜èô·X¸´2?ÿ˜n“Cëôw=èåw|nô @º‹@¯¨/ÊÎiÅôzJˆò¼À¯ô €šÇ™zB^W(Ê-ÈÉ)mQz½Z(/äèA =43ôDGuøPc{u¾øP“Ò@z“Õ)g‹Äù‡«jänk=€~FQ>TÈЃ@zhgèqYeEx¼ñ¾¸A¡[‰I*rÅ%u2Œ Tí•¢ì<=Ðm­ÐÏ,Ð#ÊG„ô €šÇz\Rž‹8¯"Ô"½¨°NNèVæå–I W5ënk=€~†Q>*"@è¡y¼ûAµÍé8¯Ö“¾ º“Pcm‡ÄµrÂø,¾"Ot[ëô3 ôˆò»¢Âô €šÇ»ôXëq6’X/ZÌ?ÖSa~^…Ô€p¬­ÔÑ[] Ÿa G”Û  ôÐ<Þå '¨Îwqis§R¥—BVW”[)Å1]}+èò†CfÏè­®'Çqêó(úzDù=q1z@ÍãÝ zBÑX"¢°n¶JÙrXœ[ÖŽ˜´îHA^~aqñáòc¥¹¦Q÷ÖÖ£{„7ÞxC*•"Üègèå“cô €šÇ»ÿýÔ¨³³]WkÖ¬inn¾•¬ÐOèå÷'%8 zîNQÞ{Éô­ûi?±é\oiÔ-ŸùeÒ ÊP_ê'Mýiû/cM­èô·\2™ ]Wüã×­[×ÒÒrËX Ÿ&Ð#ʧïÛ;žˆ^˜Þ‰‰§ôÁ™Ò¶Tû¨ãF¶6d„M-èoOº¼Ø‚¢xÓ¤z¼„¢zÄœ· €@§€^«Õ>øàƒ›6mjkk»5¬ÐOèå³ÒSnè™.v:t»ÒµÄ~…ådº €@'þƆ=ðÀ[·nmoo¿¬ÐOèåE™©vAOÍ^§ÒhÔ¸JÞÙ^y`w¦è1ÙµJj¦Z£ÕÈ«2#-bTNxz5v²ïd7©!ÕòVQ“¨g7¦°èº|ºK£Þ’c1µa9í© ½å#¦‹åFçK{—ˆ*;q à Fz8Qhñˆ!»N&WP³Äi»4²²”á õ&•®nspbQ‹~X*¡V5䚦èÓ+‹,Ä!ZR@*€@‡‚þ§Ÿ~Byéèè¸ÿþû?üðC‰D2ݬÐOèås³3lƒž–%Ñ6eí4ÌÿŠ¸Ÿ<¯5‚ž‘#?Ñ.Š2ÀÓÞÝ–8ý§É{›öçx Ji ë÷ðÇYÛœžüálKf8WÏæ,éñúD‹ð8ЄæÈBeË>AÀT¤Ë+×àeƉx¹aû«4Xq4×a¯L>ãí©ð€ Ðß± BA=ÊNKKË}÷Ý÷ùçŸ#ÖOë;wúi=¢¼=Ðóâ*‰ú$Å(µúã`<^B]wë> \¥IÈòÝÜá ·XÕÕZœž¸+"ˆ;±.t› WbƒMÿf*qQˆ‰òA±¢*ÝÛ#¸JÙ)SŸÑŠC¦ ÝÀP‘ËàXDæ¥jiZC^™q>"_Õ‘ œèôw,è¿ÿþ{}PTWW‡X(•J§õúÛzaz§*ÛHA™*Âzaºêb¿–0—šìéiJ6»3D$¤ç­—ᤖh+IN賂Lȱa¤òv•Úú¬A ¡BŒáSztÛƒU?'HnÆJŒ£óíye~Ã&VȳCS =€þÎý7ß|såÊ•›F•——ß{ï½AAAhë4±@'FôÜèê¦ÛÄ „‡ñÍúÐC³:zêxÓ zîîrËan£x;‰ˆ^‰ì´ˆè£«eéAã}pf§27<0ÐÏLÐsïr}õÕW躺pá‚yPTRRò›ßüfÅŠh‡3IÃõs@OÁYÒÕ|Ð0ÄŽ‘Qßsñ;Ó{ô½UX[ži@/<)·0ÕøVZˆˆì«M 2nÄ–’Ç+cM î—v3¾q(ˆLNßÊ™4èÃs”xq×0´n¥fðJED?ñtùñ$Qžl|FžVCb%1Ü€q>(]‚îä¥@zÐÑŸ9sfppÐ<¨G‰D¿ýíowïÞÝÙÙyë¿‘ šÐèGÝcZ-©!p\Qw0. »4 •ïÒ›¼§¨EEU8)rYCQÒð÷à ™²¦FFvw¡£É®.eM–Å÷Nxf=Ù××£Uã†uÖæÄ; cß©ÑïƒCNÔ¢¢*YÑÀÔüôÐÐi¼6•øÇÊ“ô^êï,Ž1ÜDŠšÔT’Úî.¼1/EL %E1¼É¥kÌoqARHk4ÝäŒéÕð€„Š¶‚Hà<@º£Aßßß?:¨GJKK{àår9°~¦€èA?;П:uêìÙ³ú þúõëÏ<ó̯úÅ/~qÿý÷gee©T*(.=ЃîFÐ?~õ8Ž'%%]»v-**êïÿûŽ;BCC#""bbbòóóQPÅ @zÐÝz­VÛÖÖöç?ÿù¾ûî;wîÜ?üðÀp8œÚÚÚööv‰DÒÙÙy‹§·èA =€4U ¯®®þãÿèááñüóÏýõ×(¨G—œ““=ð@ô »ô?ü°¯¯oJJJ\\Üý÷ßáÂ…K—.¡è>>>ÅòPJz ÝÕ _´hQjjjss³D"¡ÑhõL&sóæÍ---0Þ@ô »T(`÷ÝwÅb±~2z Ãöïßï½÷zzzþîw¿ûòË/[[[ôz Ý¥BdGQ<Šë‡ŸÅËåò„„„W^y…Çã>|X¡P@)èA €t— Eëñæ1;ZFpG>ŠåÑ„óz @w%è¹;EMxï%Ó·î§]üĦs½¥Q·üƒ±Ó“.7üÀ1I}?W«‘M ïà‚ô3N7A“\?wFD/LïÄÄSúàLi[ªýYä¸Ñ…­ aSKÄÛ”.7¦”PíÒO¦#L,WËs#õ ýÌ=„ñ…0ã@ïÈt±Ó¡Û”npF˜fÑãD"Ú÷ €U =€ ¿Í §f¯Si4j\%ïl¯<°;Óô‚˜ìZ%5RC­ÑjäU™‘SµqÂÓ«±“}'»I ©–·ÊˆšD¾!º-l'º†.ŸîÒè…·äXLmg˜GN{êBoùˆéb¹ÑùÒÞÁã%¢ÊNÃpB£‘NZ‘o¾ƒ‰ œÈBeË>AÀT¤Ë+×àeƉx¹aû«4Xq4×a¯Ðñ§~ì;1ðÝ÷Z³ã@ =€ ¿= çÅUõI‹QjõÆÁx¼„ºîÖ}Ä J“廹ÃAn±ª«µ8=qWDwb]è6A¯Ä›þÍTâ¢åƒbEU2êP\¥ì”©ÏhM]æ“H70T¤Àò-ž«scJÕÒ´ ‡¼BäoÅjÓu·¼¨ì&¼!5"z€@ÐßNÐ Ó;UÙÁ–ˆÍTÐ ÓUûµ„¹ÔdOOS²ÙA  "!=ïh½ 'µD[I‚pê@Ÿd8p@Ž #•·«”ÐÖgÅ Ýb ŸУÛ¬*ÞâÙD€ ¹+1ŽÎ·çU'ªkN1Mp†L™¤èô ýÑs£«šRl3PÆ7ëCÍêè©KàM3蹻ˉ†½f=æ–¼\D¯Ä vZDôчղô n?¸»Êps¿„©íxáNvÐèAúÛúŒ>4KÒÕ|Ð0ÄŽ‘Qßsñ;Ó{ô½UX[ži@/<)·0ÕøVZˆˆì«M 2nÄ–’Ç+cM î—v‹çï "“Ó÷„r& úÀð%^Å5t—ï¯Ô ^鱈è'ž.?¾‚$Ê“ÏèÃÓjH¬$†ëH?ÕÍWŠè.ì­TËó€ó =€ ¿½ кǴZRCยî`\vi@+-Þ¥6'xOQ‹Šê´'0¥B.k(JâNuN×ÔÈÈî.t4ÙÕ¥¬É²àÀ Ϭ'ûúz´jjÚ„ÎÚœXÁðólÝØwjôûàÐ…$µ¨¨JÖA40$µ?=4t¯M¥þq£ò$½—‡ú;‹c ÷#‘¢&5•¤¶» oÌKÈcIQ oréó[ÜFPß¼Ñ$ÖhºÉÓ+ÝÙS+:©æZ´4 >˜ÐèÑQŸm”8ÿpU«‚Ѓ¦ô @º@/*iVQA•¢­ºH$.mQèAz@ ŸY ×ÿ£¨+ÌίéÔÅô„²µêp¾XèW·£[5.)ÏË/¯«:”'ÊÎÎ)®lU(Z*Šs©Jª%˜¡+ïh(×­å•5vàj¢³¦0çp‹Ò¸%·B‚[M@ @zý´€žÀ•’š"QNYFÑXV] .¬jSà.o)ÏÏ)EQ?BtN¶øP#pEãáQ^aYc'†v®+êopIež¸¨Z¢"•¤¦XœWîäu…9‡šõ Ç%e¹y•RÜzz=Ðè§ë}¶¨¨FŠéh,­ÌË9ªC¾.Òo,Ê9ÜŠ!Ðçæ–Iô³É«ZJÄùUúäõ(f§v×ïÒŽƒ÷2Ý„¼¾(çPÓcíe¹ùÇd¸$ôz ÐOGDOôµ%âܲ6*¬ÆÚS]ï¹yzåæŠs5*1Iy^‘íXëᜢz…ÓÅâC¦±¶Cbcç¿šê´/ЯW4å”4*Ñö#¹ùºYO‚ÐèA €@?]ÏèåuE¢‚ê*Ü.ÏË+—â–;ãcƒÞ"èŽèu{” ÒËZJsPº>~«I@D @zýt^M¨ZäämU¸¬ª §X÷ûªÎ¶6©îýX Gl¯È×RO0)ÕEP.ÑõÇʦqÞ¡Bq1Þ·š€@ôúi=5d¾*_TT§@ìU´U—䈨1ñy…¥UmŽžxCYQŽaÔ}ƒÌøØ‘þZYX+¦¹•$ôz ЃôzèÎT —ËЃôz½QüĦs½¥QÓ×2òãvTïáOé)'áó´ç×®¸;EMxï%ӌӮەßiI—ŸP­ªÞ+°¾‘´û@q­L}êBoån.€ ÐÏxÐsc eýC—OwQŸ|×P£0Z ã…£gVåF¶6d„M¸Q%Úžò)nw'ãó4ç× Ó;1ñÔ€>8SÚ–*¼õù½]éRÓï.¶z9 ãsŠ²÷î ÙSƒ•èAzýÏ#¢>ˆ)3Œsºp‚÷–ê#»y·6®ÛÛ4 Í ‡™Û§ôŽLÚ;º]é†ågZ“¶÷à%èAzýÏôºèº“‹ÂŒÔ5Ìö¦=u¡·|ôd²ÜÙu2¹B…ájm—FV–b9_›0.¿'5jÇU-IU´R³Iæô¤HÎwç†à<'8±¨…zç“š8OÕkÖÑÀ)l'º.á‚C’H ÙÓ¢5;‚ãˆÏö½²s,7:_Ú;x¼DTÙ‰cNh4ÒÉS2jA”°WÉ;Û+ìδ½ &»VI•…Z£ÕÈ«2#-Êž^ì;ÙMjHµ¼UFÔ$òÍ j¸ÏF£Á[r,&œL~m×þ¤Òµ_ûŽÕB¨X{¡cŸ@ Ðèô£AÀ‰(ÀÚ÷h#mÌOEO¦# •-¦Æ•.î$kSõÁs‚“+´—~8ž;:HåÅV œ©Žµ¤??®\ƒ—çyå†í¯Ò`Å–±!"Í€VÛ--ØM%Ɇ‡ñðÙA¯lä7(ƒüálKf8WϪ,éñúÄ©X€N%Ñ6eí4Ì‹¸Ÿ<¯5‚ž‘#?Ñ.Šâo rÚ»Û2B†K^˜&ïmÚg¬Ã ”²Þ|¸ƒ#‘õDók³ö'•îµï@-pwUžyQè¡yÐèô†g«£ÖØA`P¬H÷ Eˆ*;eê3Zqˆ±)¯"jÍyË«éï¶zÁ~ÅàÉ|‹ŽûÀP‘Ë0ïàÆ”ª¥iA!{ëR‚ÇG½² >%v0Øôo¦…L¾5çÅUõI‹QjõÆÁx¼„ºîÖ}Ä J“&Nq£‹U]­Å鉻"‚¸ëBŸP~íÔþdÒ³ö©…ÐœžAeª@ Ðèô“‰èy»J m}VŒ ЈŠ1|¸©¦wªÌZc}ûLXH.Ø×ñÍ©B‹v«Š· ÇÉÍX‰ùøì19bÍgG½²¾,Sa ȱ©=åUv°eJ™*£[ÂtÕÅ~-a.5ÙÓÓdþ :P‘žw´^†“Z¢­$A8u ·•_»µ?™tǬ}j!0,ïÄ7ûôúé=4QÁõs'>£·Ó sw— {Í¢cKÄ#êÌ_r²Ñs£öŸkHàYÆtJ¬`§EL}X-K ˜$èõj¢  Þ[R]W[)ŽÏxFû=ʺª)ÅvNáa|³>ôЬŽž:³òœÐÛ¯ýIFôökßÐóâkÏô—Åpôz@?bÔ}èþjRstô¨{« rxŽ/Žâž2ï¯Ô ^é1ÅtœpQY—¦’²¯’è¿ØeíÕp^BÃÙS‡"-†ññã+H¢<Ùø”6<­†ÄJb¸“½£^Mô!YšŸnÞlIß =gIWóAÃ;nDF}ÏÅïLïÑ ÷Vamy¦i¼ð¤ÜÂTã[i!"²¯6)ȸQ[J¯4{<-Ü/í>Ï7ÞD&§ï åL:¿cÔþ¤Ò£öÇ=ggщ³‰<Œ ~ö ·|^£!‰¶¢„ À‘c£©ÑуCNÔ¢¢*ÙØÈr#EMj-©!µÝ]xc^Š˜¸4 •˜†Õ bó[­–$0eG½81«³ú ^lUßÉ‘ï=s‚÷·$•$Ibf˜ àÆIÒlP·Åî1|¶ë•½cCRkðÓCC§ñÚTj 7*OÒ{y¨¿³Ø½ ÓWn^Á³ÃÆÿª 5êC^iWÔŒËÂPYJ‹wñŒ¥QÔ¢¢:í‘× ¹¬¡(ixˆ{p†LYS##»»¨ªèêRÖdY pG77™õd__VcÖY›+˜ŠüŽUûL×~í;R œÈC=}UÖ†âqÂèÖžúfèâ -ZRV¥èA ÐŒŽèo‚³T6†®†‰T½MI‚ÛÑ’Úöjü$K¾»Ö{$æ.ˆg²ø{êO*Eaœ™=º{@Ï‹*+]½£ÔGå7–%ðoy;jß«qŠu¨«§:Q¨½­âŪGΛ-=ºÃAÏ Ë¨SêljkTÕÂïŒÖ÷Îô Ѓ@ »»ëèA €èA €ÐèA =€ÐèA =Ѓ@ =Ѓ@ =ºK@ŸîxÝ"Ðó›Îõ–FMß·ÆøñG;ª÷Lés&ãó´ç×®¸;EMxï¥n«ß žžâ¿Mù–tù ժ꽶'®‹7©¨O(¤²&3’÷³ýMt«¿¿’ ·üÖ½šÀU-…ñÂÑjçF¶6dLÛGE¹Q%Úž©žcd2>Os~0½O èƒ3¥m©Â[ŸßÛ•nwWÙñî‘s'èÅ ;(ÕÔî×ÏÀ‹Ê•jªùz4ƒA0röºà½e„úÈî[æð÷6 hóÂ@ÓzG¦‹Ý®t©¹rÈ3­ÉV‚zÞ®œÜ8Ózîî2¼~@~> 5½až1í© ½å£§1åîÌ®“É* Wk»4²²”‹àL—ߌ“5Žãª–‚¤ƒª‹ZiÑÈ©Ú)’óݹ#'|ã'µàjµîKµª†\³Ùë¸1…íD÷À%ü@pHò)¡!{úO´fGpñÙ¾WvŽåFçK{—ˆ*;q à Fz8qÊ@FÍ^‡Vã*yg{åÝ™ Äd×*Õºöj5òª=ÎœðôjìdßÉnRCªå­2¢Æ¦ Êl~B‹‰þ&™_Ûµ?©tí×¾cµ*Ö^èØ7æ¬Á$„õÐ@f.è8Xûþm¤ùÊ©èÉt`d¡²ÅÔ¸rÂÅdmª~ rNpr…öÒÇsG©¼Øª3Õ#æåÇ•kð2ãT¬Ü°ýU¬8zÄ|ôš­¶[Z°›J’# ã;ೃ^ÙœŸüálK¦þëøœ°,éñú)éùE§’h›²væEÜÆOž×AωȑŸhEñŒ·9íÝm!Ã%/L“÷6í3ÖaPJi¥:YO4¿6kRéŽQûÔwWåéêX»=S¼˜ÂN¼r¯0 @?];{ôËÕ‹Σ?éL\q­ì厰—w´þ0¥§lÛöO§¨îŸnõ±“×'²ß_ìô¸gæÙ·ªøoS~§%ÝË’¯—Ýü­²û±ï0g¿Ÿÿܹ¾s_ ¯9wmZAoeMÅŠªd(òB¨²S¦>£g|åÅWµæ¼áÅÕôw[½`¿bðd¾EÇ}`¨HåG˜÷pcJÕÒ´ Ž½u)Áよ£^ÙŸ;lú7S9%SÜòâ*‰z‹Yzù‰õÆÁx¼„ºîÖ}< J“¦1 ÜèbUWkqz⮈ îĺÐ'”_;µ?™tǬ}Gj!4§gP™j;¤çFŠZ‰†ôÐ;}.Û[zUÔâ'~ý›¿½Yýíp‹3xló¿ùõ‹£±+Ž7b­Û—…[ýÅ¢oU|kíˆëçmw{àþçüßH&¯Þ¼J¦¼áÿÜý¿wÛ^yþúƒù»b}YAØÔÞ¿\%“¶|\}gÑW’¡ñD+ÖAã¼hÙÚÃ6ú¦xå²üKÃÿ^ÈY¶æÐàÎÿмí9ß´þë7AÓzû×Æôév¥KÝÜžÎðæíúï­Ý}ä/]tðœ±\êÞ½ðõcßݲgôvdîîr¢a¯YtlɸcDùKN¶"znôÑþs <˘N‰ì´ˆé¢«eéA“½£^MôÁ{Kªëj+Å1ãÏh?¢GYW5¥ØÎi  <ŒoÖ‡šÕÑSgVžÓzûµ?ɈÞ~í;z^|í™þ²k$Ä’«ëÒîøXþ–‚~™ áÈ» ‚”TU4Ëᆠe:Ð_UïÙì÷Ü_êùûcwú–ÅÞÿ}Ì+ûü±AU³÷õ…žÏÞÿ»ÿzÏÕkñ;™=?:úkÛ÷l]2ÁÂ…‹.XñÑÅ·:ä^?W¶ÝûïO-zm‰§‡»+ëõăµ>^î®î[óú~ãXCSÿ­|ß»‹ýç.X¼dÉò­)¹ËŸxÖçµDõˆŒïk7=é‘~zçíùª& eù¹Çºp®nçk çù1_t{;§×àÖµÓ%Ÿ,Ÿ;××åïÖsØ÷ÊαWµû_sú;VÄ_¿dÙ²% çÎÝÓ>8E½ë7.“9Ÿ­˜7oÁâk׿Î;¤*4ýïÕâ/VÍGE1®ïÜ—Šz,Êð§þÒ¯—ÑçÐ<ýçúÍéí ‹¾l»ìàµ1™üþxBüÙ†µ«W,[<ß×{Þ«!5ç¯9|MÚ¯#;µïP-Ü8“åóWª¾]Î?õf¿µ>´þu©\í-ýrõ‡GÎ]Ÿ&ÐsB÷W“š££GÝ[mÃs”xq×ð”y¥fðJ)¦ã„‹:Ⱥ´0ý“Ô}•DÿÅ.k¯†óΞ:iÑðòã+H¢<Ùø”6<­†ÄJb¸“½£^Mô!YTOƒ-Éã{ò‹à,éj>hbÇȨï¹øé=záÞ*¬-Ï4 ž”[˜j|+-DDöÕ&7 bKÉã•f§…û¥ÝÇâùÆ»‚Èäô=£(7þüŽQû“JwŒÚôœE'Î6&òF?jÚ[®Â*“‡øðã‹ byúeBÅ·xØ¢­•ß\¬Ú:_ üV¼l8¢ÁßÁÃk—‰ý$#úËÁó–„6Úè*æ-‹'õ) I·¿Èâ·{ãæ7å¯üÝ9 5¨ßÕ}°8 ¿:æ±7¯õe­÷ÿ¢âŒŽ¾×.Ô |þú ûÀ™Q`¼¢ xö™ KúÛ=³1ËsŸõõõx-ûó§ÁþË7Æ.+½²QÎßûýþß[‹útȺ6P°ÑýR'9c `³ïû'ô…pµïÈŽÅô§| ÿéTÆZ·7DÝC7 ·™¯{¾YrnøÆçÛ£k·U£ßùÆ75û}Òry|‘õó{ãúpÁýt|ïªwª¿»9éŽQûÔÂUœÿßgG^TƳ˃\ÿØl·ç~È7A{uêãY¾G¯ÑD[QBPàȱÑÔèèÁ¡ 'HjQQ•dzÐÙ¤Ö’RÛÝ…7楈‰KZ‰iX½ 6¿…ÐjISvÔ‹³:0«ß€áÅVõ9ø™¼§¸ ©$Ik4Ã\7¦HBfƒº-†tá³]¯ì’ZƒŸ:צRá¸Qy’ÞËCýŦè]˜†*ð ž6þW©Q÷òJCยî`\†ÊRZ¼‹g,¢Õi¼VÈe EIÃC܃3dÊšÙÝEUEW—²&Ëb€;º¹É¬'ûúz´jðÎÚœXÁTäw¬ÚŸ`ºökß‘ZàDê髲ð Lòû =¤Æ(²kàâ€@¿L¨¼rý´hýê”–Ô5«Óû®]UÝ"ÐÏûË‹~ ò}áñEÆ„®Ÿ­^¶ï”ùCÒ5‘ó_;úô_¯ÔCýŠB°’/×uE+õ>Û?öJGàÂÏÍ‡× u~õ‚§ÐW¹úOô}÷öÏlÌr¶ŸÓöÚ‹×ÇG½² ¾•Kó.šþ-ZµDt~òœ¿"ç-ü´Ñ<þ¼ÜþésžzÐI>óx»Êâa÷7e›ü‚‡Ç4ü¨MXáýv‘6üÔ¥«7&Ô…>¡ü^ÿF! Ü°dþüK–¯Zÿêü§}Mâ“HwÌÚw nœÍ`þiUÅèþÚéü-K¿(íCIÞêÎyñG¦0¢¿u ÎRÙº&Rõ6% nGKjÛ«ñK,ùîZ»àÙïLOýI¥(Œ3òr AÀÖðÁì=ÿvÕ Õ»#ú+rþ²™å(ïëÞ[« môö,]¿"÷Â7fŒý÷Õ/ÿÅuoï5 ðÙ;³£±VVŽze|æ%i½Ž&ð@y•sÁ2¥¢ £î¿=²üo/ø.4×&k›ù3èëߟ’–¦†}²a‰ŸÏ¢7ã¤ßÞ˜*ÐÛÌï3•§ýkcÂù½ª Zôq³YßÈࡵK³G‚þЄ"ú1jßÐ_éøüéC4£Þ_ÅC›{ýmÅëKöž¸vw€ž–Q§Ôרª„ß­ïét»AoÖ옃þ‡úw|¾jÓµA7¾S¦½õ߇<²ýÍo+7zr:.[þãõGZÎ^sì}‡ÐQ°~tõx¸ì+ÿ¥±ã3z{ ·ìÍkýÙ¯øm?zZ×Øþt®š¿èÅ¿y[‰èoI>úë®ã/SÛ=ó$@ï¨Wýóùó~5kÖŸÞ©ûv\qþµ3›½ßËëÑëÕS%Ÿ°þöÈð{ô7¾m\úfªÒ8 íÆ•¾†I§õ…sã\¶ßœÏ¿1nü^åïÎ7»Òì^ÎïõþŒ•Kºô(ýiàÞc÷±²Ì@?‰tǨý±Aíd"ã?¶Y{½.fÉò]rÝ[ö×/5ñ¬M=ùÓÝØuî<Зo÷~üþ'èþkøMƒúwx›ùküOÜÿ¸÷gúWÚ<)~Ç›ååëãí½àõ¨¢ä¹yq}‚~0øµÓÅÔej”òc÷ÿÝÍ-Í[XwÉÔºýÔWô‰ÿœ9,ßK–-[ºþ‹ ¥þ‹!×ÏW~µä¿÷xþw/,ùê˜áÙäÍkZâßZH}Ad®¿ÿÒÒÌGÝ?þÀ?æq›¾½qE8ûŸ~;j/ݸJ»<5ì³c ÌùN¹ÿ…>¾þ —­zùÓ¬öü——eZýMêä´H­eøeçÌÔ; ýÍu/1Ò=VYÙóÊÞ±×ÏW|I•äó‹¿(§Â]íJÝäü»û_0Ö‘Ž9e«ï›uïbñÀ¸_¤FÝo_æëë?oÑ’%k>Ë“,ûë³¾¯Å†—.´&n]¾hÑ‹–­Z³öÕ†¿ôr±xê/¿zÕßÃË×®Ÿ—×ʯ z,ÐiëÚ˜d~¯Ï~JÒÇÓkɇiµ=þœï&³—*&–®ýÚw¤nþt|7sv€õ¡x覤B¸Žú`Ž¿ïÜQ ç§ôƒ9 ègу.æ¯X":g-Ò½> Záü~ã÷w–Wã×÷õ›þ¿N;5?BeßNýÐú)}eöéq~†@èô“YÞ•òÒJcWïèÕ®ý†J/ßa^S?uïöf}Ýö= ö¶jH¾ëëƒÇÇ]¥z@ Ÿ€®.Ù¾jÑÂ… -˜·bÇá¾Á+Ð*= Ѓ@z= Ѓ@z @?SÐó›Îõ–FMß·ÆøñG;ª÷Lés&ãó´ç×®¸;EMxï¥n«ß žžâ¿Mù–tù ժ꽶>[Ä I*iŵZ’$»ºÇÒÂ9z4“Aoù­{5«Z ã…£?ÔÎ.lmȘ¶Šr£J´=¦ùէ褓ðyšó뀄é˜xj@œ)mKÞúüÞ®t¸»ÊŽwœ;Á ^Lþ±¢½ÁúñG Ua$@ftDo9{]ðÞ2B}d÷­Ï‹¿·i@›šÐ;2]ìtèv¥KÍCžiMû[ÄœÈb¼}Ÿ@~> 5½až1í© ½å£§1åîÌ®“É* Wk»4²²”‹èH—ߌ“5Žãª–‚¤ƒª‹ZiÑÈ©Ú)’óݹ#'|ã'µàjµîKµª†\³Ùë¸1…íD÷À%ü@pHò)¡!{úO´fGpñÙ¾WvŽåFçK{—ˆ*;q à Fz8qÊ@FÍ^‡Vã*yg{åÝ™ Äd×*Õºöj5òªÌH‹2䄧Wc'ûNv“R-o•5‰|ó‚2›ŸÐb¢¿Iæ×víO*]ûµïX-„Šµ:Æ"87ü@ƒ¦-#$@~V àD`íûG´‘6æ+§¢'³ð¨PÙbj\9áâN²6U?9'8¹B{é‡ã¹£ƒT^lÕÀ™êsŠòãÊ5x™q*VnØþ* V=b>zÍ€VÛ--ØM%Ɇ‡ñðÙA¯lÎÏNþp¶%Sÿu|NX–ôx}âT ,@§’h›²væEÜÆOž×AωȑŸhEñŒ·9íÝæ|¦É{›öë0(¥¬7îàHd=ÑüÚ¬ýI¥;Fí;P Ü]•§ªmÎ4Ï.èèûîÚÕ“ wøDz@?  ·²Æ&ƒbEU2y¡@TÙ)SŸÑŠ3¾ò⫈ZsÞðâjú»­€^°_1x2ߢã>0T¤Àò#Ì{¸1¥jiZGÈÞº”àÀqAÄQ¯l‚O‰ 6ý›©œ’)nyq•D½Å,½üÄúã`<^B]wë> b¥IHÓ˜nt±ª«µ8=qWDwb]èʯڟLºcÖ¾#µšÓ3¨LµKqNpBA«ºMÏèA Dô6@ÀÛUJhë³b†@TŒáÃM½0½SeÖëÛgÂÊ@rÁ¾ŽoNZ´¶|XU¼e8&HnÆJÌÇgÉk>;ê•ðe™ Kx@ŽMè)¯²ƒ-SÊTݦ«.ök s©Éžž&ógЂˆ„ô¼£õ2œÔm% ©½­üÚ­ýɤ;fí;P ay'¾éûé;u‹U—ÈЃ@ Ÿñ3z; 2ww9Ѱ׬‘´ä@Ü1¢Îü%'[=7úhÿ¹†žeL§Ä vZÄtчղô €I‚ÞQ¯& úÀà½%Õuµ•â˜ñŒg´Ñ£¬«šRlç4PÆ7ëCÍêè©3+Ïé½ýÚŸdDo¿ö=/¾öLY̘ïqp"‹‰Qwµz4ƒAÏ Ý_MjŽŽuoµAÏQâÅQ\ÃSæý•šÁ+=¦˜Ž.ê ëÒÂôORCöUý»¬½ÎKh8{êåKNüø ’(O6>¥ O«!±’nÀdAï¨W}H–槛7[’Ç7HÁYÒÕ|Ð0ÄŽ‘Qßsñ;Ó{ô½UX[ži@/<)·0ÕøVZˆˆì«M 2nÄ–’Ç+ÍO ÷K»Åów‘Éé{B9“Îïµ?©tǨý±AÏÙYtâlcâè›-ážCe⸠ý‰…qÅ mõuf0è-ߣ×hH¢­(!(päØhjtôàÐ…$µ¨¨J66²ÜHQ“ZKjHmwÞ˜—"&. h%¦aõ‚ØüB«% LÙQ/NÌêÀ¬~†[Õwrä{Ïœà=ÅmI%I’X£æ¸1E‚4Ôm1¤{ ŸízeïØÀÔüôÐÐi¼6•ÇÊ“ô^êï,6EïÂ4Å•›Wðì°ñä¦FÝcÈ+ ㊺ƒqY*Kiñ.ž±4ŠZTT§=òZ!—5% qÎ)kjddwU]]Êš,‹îèæ&³žìëëѪq Ã:ksbS‘ß±j‚éÚ¯}GjÅé‡zúª¬ÅãE8Ú†k©¢ÒšŽ²Ôpn@€ÍèˆþÖ)8KecèZ`˜HÕÛ”t["+Û^_‚dÉw×zÄÜÙì˜ñâï©?©…qfB^ô €þî=/ª@®4võŽ~P•ßX–pëEÙ÷jœâDêê©Njo«xq‡ªÅ‘3äf @èïpÐsÂ2ê”úqâUõ;¤›ôÎô Ѓ@ »»ëèA €èA €ÐèA =€ÐèA =Ѓ@ =Ѓ@ =Ѓ@ » ôüĦs½¥QÓ÷­1~üÑŽê=SúÁœÉø<íùµ«ÿÇÞ™ÿ¥­¬ üO—„]PEEEQTÜ—Z JQY²DE´‹Zßs[{νŸOß D„—ö<ߟÚÓ›ÌL†;ßÌä™yÐNWˆÈÝdªžü<ÿ•Úû,åª|˜o¨Î±Eˆa:vþõÿØi-ˆ€?Zôâ³î)’ÀÂKýšÇµ£ÝKû»“Ïv¨(jZe³åüê-ºiu~æöÊ@ãˆãîÖˆ^7=Ó¼|{_«Ü6´gó8S™;AŒÖ¾›ÚžÁ„ÉqÞ¶èå  zÑËÈ^§Ú$©÷½ÊÌTC¡KvÞ¨hžCôrÒÅ>¯Un!wÂ,óißVsR¯¶~HùÕYðÞ¼èð8ˆ@ômuóÑ—òŒ±gW¹­ÇiLÑΙ@,‘Äp‚bÓtlsD/šœiú> M…‡§±ëK6º\™ª]=ùš™«Lø†è¬Ëa‚¢îOªÅvçÙëÚPóÒ!™¹¼!&tzÛû(I3ÙüÉþL"§ÎÒµ’¸í^ˆænW]ž8ãIÓÑwÖ–‰¬½Ž+˜"°DüÐ3Ñ;%½Ú<ãOQ÷‡ö²tÂ;Õ%z†ˆÑáÃOÏO3 ÍP‰ý¹cU ” ?¡(Ñ_“í­ÝûM•+Ýûòz¡ÝÍ^V7½²w=¹7¦{”îDÀ¿DômHÇ"~h¯#kä+/ÌžÊv-¥ÂåÁ1ºãŒ¬˜‚ÑÙ¶Ù›ïÇs'©J‹÷ò“¯"§¨ªo‹&6ùT¬¨Áî¥ñ•îŠ|ôô%Ëf¢‹½…"Ñ ’Qg™µª™Ÿùþ9±n>㫲ßKú…¾Qöíä3UD¯¶'oOD ÷ŠvW_è® æu*:®y„ÉFtŠ'IDn­jŠ/…OëÊJµ$Å­²ÏCEYzUÖà%Œ§döGEÆÔŽG˜rLÚ½‚¥÷WÖž-ÚØzCí•èýfÊ­Ûûrz¡}6{›S?^ÔwîǺ‡û€èAô3z ({ÖI6è4«¥‰¨'†z#Ž Fãâ¸JV $W};[ëœøpo¿x:¦¶}ÄW…ñÙu=R­Îrk%!>4 ¼¢/ÔjF'.i ã«¥q`×y–B1ÙlHø Z¡îpÌocÃ’«šÖ‰¾V{%{¿™rëö¾Œ^PæO¾Å¯Ý«GöOŽéÒSdÎüÈE»Q=ÿæoô2Ú»Eî fÇbô} ÂMNµfôh÷FþËî€R<§Ká‹¢9]÷;*æж5)z¹µjTô ÝЪ/à÷¸ÍO‰g”žÑsMÇB#µ[ªP *Áz»ó(<Ïç½tï79£—î}¢Wöû?å7ÍÒ ‡=ˆþ…¢GÚí>†Þxu_u@6ΦˆZúÊl÷з?³å9bt1qCñKª~ÔCæ¯ÓÕ¶†+v?Ÿ­u‰ÂøTýÛ ¹eã¿ÒÇw|ÕŒ¶5+z¹µjTôz'ýß_¿nö§éqrŽ¤?N—BìÐŽÉ`öú¯ò>zÍ?˜/¤)ÃsKcü®4½‹9÷kùT[Ö™càó´ÆÍ|èWño]6Ç`;Òt{ëô~SåÖéýú¢G:—O>ïYë¼lèAôð/½x=M3äÁò€VQ]ˆŽ¾½»:a Lzmüàˆv¹BËÐ ›I{ó#nòæ’”ÃêÕ–…0ɲ ‰§Ž‚n«ó¯zŒÒâ=?­Ü÷ŒèWH¦P$Ãà{͵¡æåÉ‚ºE!Ýuê,Y+©kú±ââîî‚ðáPÓ|$÷ã._)ÏÞ5ãÉŸ¿~3†§o,DÝã\­h’ ’é>'Î=ËèJ’Ëa¬°hÏÕ:™ˆí.?„¸ë&c©“Iº"Ní8EîÜËÍT9?ϲãxÜ?kQ·¢½õz¿Ár¥{_N/´!]kÙs¯EÂóh÷™æîs[2ÃÒ=öŒþåÐ9±¡k ƒ Ë…†Õ¯1’Ö®ÕÓQÛ"ý/÷þM›ã_€j0xšr½æqG zþ¢Wš)~©÷ñ Ë´°·9 zñqTºVO1­¥³>«Tûª(ûÖ|î®?äe D ú7.zÄ0HãÄiÌ7a|£ïÛ¬€èø½—îÑ¢Dˆ@ô zуèD¢= z@ô€èÑ¢à÷½Êú’[7=ßYcªþ#ß`KÌi¦ÎÏÞ^IÐNWˆÈÝdªžü<ÿ•Úû,åª|˜o¨Ö±Eš±äÕEš*%°#"ü!à zþLыϺ§H /õkÔŽv/íïN>Û¡¢¨i•Í–ó«·è¦MÔù™Û+#Ž»[#zÝTô`Lóòí}­rÛОÍãÌz쳚±(>×@.=¿íŒ^œ½N7´IRï{_v’£ ]²óÆßfðýÍD/']ìsðZår'Ì2Ÿömj=ˆ@ômuóÑ—òŒ±gW¹­ÇiLÑΙ@,‘Äp‚bÓtlsD/šœiú> M…‡§±ëK6º\™ª]=ùšy4ö":ër˜ î×X)lwN½® 5/’™ËbB§·½’4“ÍŸìÏt rê,]+‰kÑî…hîöxÕå‰8N4}gm™È Ù븂)KÄ=½S"Ñ«Í3þuh/K'¼S]¢gˆ>üôü4ÃÐ •Ø‘;V•ðA òŠý5ÙÞÚ½ßT¹Ò½/¯ÚÝìÕѨºªè_ÎX†e6¦>ΚU zþe¢oC:ñC{ÅY#_yaöT¾°k).®ˆÑgücÅäˆÎ¶ÍÞ|?ž{ÔŽì2Aa¸ƒœ™u£í­ÙûM•[§÷eôÚ㹸ôUKT‹èFûÛÑÒ›Š;ÊxT zþ]¢¯ò_jŠ@¡µ¸¼1næÅMDSñõ‰uó_•ý^Ò/ô²o'Ÿ©"zµ=y{º Z¸W´»’øB‡pu5¯SÑq­È#L.0¢Sgùai&x+D_¨ÕŒN\ÒÆWKãÀ®ó,)„b²Ùð´BÝ1à˜ßƆ%V4­}­öJö~3åÖí}½ 0ÌŸ|‹ÖÏÜîÆã ˆ€ñ7z‰íÝ"w‡³!±ú>á&§Z3z´{#ÿew@)žÓ¥ðÅNÑœ®ûshÛš½ÜZ5*z…nhÕð{Üæ§Ä3JÏ蹦c¡‘Ú-U¨•` ½Ýy” žçóˆ^º÷›œÑK÷¾ Ñ+ûýŸò›æºû84£ém ¢à_#z¤ÝîcèÇQ÷UdãlŠX1¡¥¯Ìv}û3[žÓ!F×7¿¤êG=dþ:]mk¸r`÷óÙZ—(ŒOÕ¿Í[6þ+­q|‡ÁWÍh[³¢—[«FE¯wÒÿýõë6l{Ú$‘“s$ýqºb‡vL³×•÷Ñk†¼øÁ|9 Miž[ãw¥é]̹XËÿ£Ú²Î{Ÿ§5öhæC¿Š+è²9Û‘¦Û[§÷›*·Nï×=Ò¹|òyÏúøeK?q@…çzÔü¯ÝÏ&ÜF¤ DÀ+zñ>zšfȃå­¢26º}{wuÂþ˜ôÚøAír…(–¡6“&öæGÜäÍ%)‡Õ«- a’eOÝVç^õ ¥Å{~Z¹ïÑ ®L¡H†Á÷škCÍË’u‹BºëÔY²VR×*ôc;ÄÅÝÝá+¡¦ùHîÇ]>¾Rž½kÆ“?ý$fؾUˆºÇ¹ZÑ$A$Ó}Nœ{–Qþ8îi,‡±Â¢=Wëd"¶»<üâ®›Œ¥vvbL&]èŠt:µã¸s/7SAæü<ËRŽãqÿ¬EÝŠöÖëýË•î}9½Ð†t­eϽeõç<剳Ç–eÓtdEøkÑðgÎè_«º¦0¸°\hXý#iíZ=µ-ò×ÿrïÍhðŠ¨ƒ§)—ùÚ¢ýï#z¥i1‘â—z¨7-ìm¾Â>'éZ=Ä´–Îú¬jPí«¢ì[ó¹»þ—-=€è߸èÃd UŒ§1ß„ñmŒ¾o³V¢à÷^ºDˆ= zуèD¢=ˆ@ô€èÑ¢Dˆ€ßKô*kèKnÝô|g©ú7Ž|ƒ-=0§™:?{{%A;]!"w“©z^ðó<þWjï³”«ða¾!éc‹PãèûI$M'·!{®èÅgÝS$…—ú5OÿF»—öw'ŸíPQÔ´ÊfËùÕ[tÓ&êüÌí•ÆÇÝ­½n*z0¦yùö¾V¹mhÏæq¦2w‚õÀ•,ôiÞö1÷ zž%{nh“¤Þ÷*_t0S ….Ùyão0ìþ–¢—“.ö9x­r ¹f™Oû6uU„Lb¶ã79 D úgÏG_Ê3Æž]嶧1E;g±Dà ŠMӱͽhüÔô-|$š" /Oc×—lt¹2U»z$ò53W™ð ÑY—ÃEÝŸT‹íÎ ²×µ¡æ¥C2syCLèô¶÷Q’f²ù“ý~ð–®³t­$®E»¢¹ÛãU—'NàxaÕ7úÎÚ2‘²×qS–ˆz&z§D¢W›gü)êþÐ^–Nx§ºDÏ1:|øéùi†¡*±#w¬*áƒä'%úk²½µ{¿©r¥{_^/´»Ù«£Ñj¦W¤½}Êßå…D úÖ‹¾ éXÄícd|å…ÙSù®¥T¸<¸"FwœñS#:Û6{óýxîñ$Uiñ^~òUäUõmÑÄ&ŸŠ5ؽ4¾Ò]‘ž¾dÙLt±·P$¢1T2ê,³V5ó³3ß?‡§Š§ã#gô8hmE`w«rv–ò¿rÞ&N¿²¼è‘ŽÙÄɡˤä_ f3“ú‡'¯OäB£|jGv™ 0ÜAÎ̺ÑöÖìý¦Ê­Óû2zíñ\\úª$ªEL«©][×èû(÷ÆD¦B‹ýo{D úg}•ÿRS ­Ååq3/n"šŠÇ¨O¬›Ïøªì÷’~¡o”};ùLÑ«íÉÛÓѽ¢Ý•ÄD««¨yŠŽkEarâI‘[«šâKáÓºò_§R-Iq«ìóAQ–^•5xÉã)™ýQ‘1µã¦Ó€v¯`éý‡µ§C‹6¶„ÞP{%z¿™rëö¾œ^hŸÍÞ¦ÆOéQË‘8"#ËÁ#úQý¶—ñAô¢Õ½²gdƒN³ZQšˆºqâa¨×8â˜`4.ŽÏd•@rõèÑ·³%Ñ`ˉ÷ö‹§cjÛG|UŸ]×#Õê,·Vâs––f"·Bô…ZÍèÄ%Ma|µ4ì:Ï’B(&› ¿A+ÔŽù`Œ`Xò`u@Ó:Ñ×j¯dï7SnÝÞ—Ñ ÃüÉ·x•µ{¤c){±;¬)¯†Ä¨•.DÀ¿÷½Ä€Œön‘»C‚Ù±Ø}È€p“S­=Ú½‘ÿ²; ÏéRøb§hN×ýŽŠ9´mMŠ^n­½B7´ê ø=nóS¾KÏ蹦c¡‘Ú-U¨•` ½Ýy” žçóˆ^º÷›œÑK÷¾ Ñ+ûýŸò›f´Ú³ÞNn•×ôÕ£ä»n=ÿÑ#ívCo<Žº¯: gSÄŠ -}e¶{èÛŸÙòœ1ºŽ˜À¸¡ø%U?ê!ó×éj[Õ»ŸÏÖij*Uÿ6CnÙø¯´Æñ_5£mÍŠ^n­½ÞIÿ÷ׯ۰íiAzœœ#éÓ¥;´c2˜½þ«¼^3äÅæËiJãðÜÒ¿+MïbÎýÃZþÕ–uæØ#ø<­±G3úUü[A—Í1ØŽ4ÝÞ:½ßT¹uz¿¾è‘Îå“Ï{Öª/[Hç\ìhµ÷¾^JóR’X3£m zþTÑ‹÷ÑÓ4C,h•±Ñ…èèÛ»«¦ðǤ×Ʋh—+D± Í°™4±7?â&o.ÙH9¬^mY“,Ëxê(è¶:ðªgÀ(-ÞóÓÊ}ψnpå€d E2 ¾'Ð\j^ŽŒ ¨[Ò]§Î’µ’ºV¡Û!.îî.ÿX!5ÍGr?îòñ•òì]3žüùë'1cxz€W!êçjE“‘ L÷9qîYFWz”üÓXc…E{®ÖÉDlwyø!Ä]7KíìĘLºÐétjÇ) pç^n¦‚Ìùy–¥ÇãþY‹ºí­×û –+Ýûrz¡ éZËž{-5U4k‘Â^†ÁüSoXó zÑÿ^GàêœXÐ5…Á…åBÃê×Ik×êé¨m‘¿þ—{o~ÓêøóQ OS.ò'´D úßGôJÓb"Å/õ>þPoZØÛP½ø8*]«'‚˜ÖÒYŸU ª}U”}k>w×ò²¢ý=b˜ ¤Šqâ4æ›0¾Ñ÷mÖ @ôüÞK÷€èÑ¢D z=€èAô¢Ñˆ= z@ô€èÑð{‰^e }É­›žï¬1UÿÆ‘o°¥æ4Sçgo¯$h§+Dän2UÏ ~žÇÿJí}–rU>Ì7TýØ"í$u“gÊiÙü 9¡ÑðçŠ^|Ö=EXx©_óø v´{iwòÙEM«l¶œ_½E7m¢ÎÏÜ^hqÜÝÑ릢cš—oïk•Û†ölg*s'QZ—ûËï“ÊÒׯÑðGÏèÅÙëtC›$õ¾÷e‡>ÕPè’7*ڀ罜t±ÏÁk•[È0Ë|Ú·Õ=‹XiÙ§ÚÑðÇ‹¾íQ>úRž1öì*·õ8)Ú9ˆ%’…L`lšŽmŽèE“3MßÂG‚¡)‚ °ðâð4v}ÉF—+Sµ«G"_3s• ßu9LPÔýIµØîœ {]j^:$3—7Ä„No{%i&›?ÙŸé@äÔYºV×¢Ý ÑÜíñªË'pœ i:úÎÚ2‘²×qS–ˆz&z§D¢W›gü)êþÐ^–Nx§ºDÏ1:|øéùi†¡*±#w¬*áƒä'%úk²½µ{¿©r¥{_^/´»Ù«£Ñ:¦GÍïÉðÈÛNM¢}ëE߆t,â‡öŠá¯F¾òÂì©|a×R*\\£;ÎøÇŠ)Èm›½ù~<×^eZå½üä«È)ªêÛ¢‰M>+j°{i|¥»"=}ɲ™èbo¡HDc4¨dÔYf­jægg¾OOÇG ÎèqÐÚŠ !w«rv–ò¿rÞ&N¿²¼è‘ŽÙÄɡˤä_ f3“ú‡'¯OäB£|jGv™ 0ÜAÎ̺ÑöÖìý¦Ê­Óû2zíñ\\ú,’+SH× q8¦iÑðo}•ÿRS ­Ååq3/n"šŠÇ¨O¬›Ïøªì÷’~¡o”};ùLÑ«íÉÛÓѽ¢Ý•Ä:ÑìkŠŽkEarâI‘[«šâKáÓºò_§R-Iq«ìóAQ–^•5xÉã)™ýQ‘´ã¦Ó€v¯`éý‡µ§C‹6¶„ÞP{%z¿™rëö¾œ^hŸÍަƤfëHÇv4©ƒ¨{HfôµE ìY'Ù Ó¬V”&¢nœxê5Ž86­«°Y%\=zôílI4®sâý1RjÛG|UŸ]×#Õê,·Vâs––f"·Bô…Z͈•£ÂøjiØuž%…PL6~ƒV¨;óÁÁ°äÁꀦu¢¯Õ^ÉÞo¦Üº½/£†ù“oq©µ{…ÁLÌ´·èa €û7z‰í݇2‰=Ð÷ 79՚ѣÝù/»Jñœ.…/vŠætÝ礼CÛÖ¤èåÖªQÑ+tC«¾€ßã6?%žQzFÏ5 Ôn©Bm4¨kèíΣl@ð<ŸGôÒ½ßäŒ^º÷eˆ^Ùïÿ”ß4£R{⩹· ¢}«£îÛí>†Þxu_u@6ΦˆZúÊl÷з?³å9bt1qCñKª~ÔCæ¯ÓÕ¶†+v?Ÿ­u‰ÂøTýÛ ¹eã¿ÒÇw|ÕŒ¶5+z¹µjTôz'ýß_¿nö§}úåäIœ.…Ø¡“Áìõ_å}ôš!/~0_HS‡ç–Æø]izsîÖòÿ¨¶¬3ÇÁçi=šùÀï)S¨»lŽÁv¤éöÖéý¦Ê­ÓûõEt.Ÿ|Þ³J¼li|©iуèà½x=M3äÁò€VQ]ˆŽ¾½»:a Lzmü ‹v¹BËÐ ›I{ó#nòæ’”ÃêÕ–…0ɲ ‰§Ž‚n«ó¯zŒÒâ=?­Ü÷ŒèWH¦P$Ãà{͵¡æåÉ‚ºE!Ýuê,Y+©kú±ââîî‚ðáPÓ|$÷ã._)ÏÞ5ãÉŸ¿~3†§Ï Q÷8W+š$ˆd`ºÏ‰sÏ2ºÒ£äŸÆr+,ÚsµN&b»ËÃ!îºÉXjg'ÆdÒ…®H§S;NQ€;÷r3dÎϳ,Eà8÷ÏZÔ­ho½Þo°\éÞ—Ó mH×ZöÜ+Š§Ø>XìzûžÑˆþ·:WçÄj„®) .,~•NµkõtÔ¶È_ÿ˽7£mÀ+¢ ž¦\äOh ˆ@ô¿è•¦ÅDŠ_ê}ü¡Þ´°·9ðòG—H×ê‰ ¦µtÖgUƒj_eßšÏÝõ‡¼lèDÿÆE&©bœ8ù&Œocô}›µ=¿÷Ò= z@ô€èшD z=€èAô¢Dˆ= z@ôü^¢WYC_rë¦ç;kLÕ¿qälé9ÍÔùÙÛ+ Úé ¹›Ìì‹åV{­ö>K¹ªæªyljœød §ë²tlcD€èø“E/>ëž" ,¼Ô¯y|P;Ú½´¿;ùl‡Š¢¦U6[ίޢ›6Qçgn¯ 4Ž8înèuSу1ÍË·÷µÊmC{63•¹ø3¯“©åžâ+¥ÆºE%æ:=ôŒ^œ½N7´IRï{•/:˜©†B—ìüÛOú›Š^NºØçàµÊ-äN˜e>íÛªMêuNœdX4­‘‡v5ˆ€èÛå£/åcÏ®r[Ó˜¢3X"‰áŦéØfÅB¨¦oá#ÁÐAXxqx»¾d£Ë•©ÚÕ#‘¯™¹Ê„oˆÎº&(êþ¤ZlwN½Ž›–-’™ËbB§·½’4“ÍŸìÏðS3é:K×JâZ´{!š»=^uyâŽ$MGßY[&²Bö:®`ŠÀñCÏDï”HôjóŒ?EÝÚËÒ ïT—è"F‡?=?Í04C%öcäŽU%|P‚ü„¢DM¶·vï7U®tïËë…v7{u4ZÅàH‡ë ¾YÌ€‹v8v’{½DÀ¿JôÜP¸ˆ?šåÔÈW^˜=•/ìZJ…˃+btÇÿX19¢³m³7ßçOR•ïå'_ENQUßMlò©XQƒÝKã+ÝùèéK–ÍD{ E"£A%£Î2kU3?;óýsxªx:>bpFƒÖVp·Š°!gg)ÿ+çmâô+Ë‹é˜MœºLJþ•`ö0s0Yö“f<‘ ò}¨Ùe‚Âp93ëFÛ[³÷›*·NïËè´Çsqé«ž¨–»ûÙ÷Ûó“Ë¿þît¿ñD zÑ?ƒè«ü—š"Ph-.oŒ›yqÑTâ«Âø캩Vg¹µ’Ÿ³ü°4 ¼¢/ÔjF'.i ã«¥q`×y–B1ÙlHø Z¡îpÌocÃ’«šÖ‰¾V{%{¿™rëö¾Œ^PæO¾Å¯Ý#¦UüãHù?ë&c©YƒDÀ¿ø½Ä€Œön‘»C‚Ù±Ø}È€p“S­=Ú½‘ÿ²; ÏéRøb§hN×ýŽŠ9´mMŠ^n­½B7´ê ø=nóSâ¥gô\Ó±ÐHí–*ÔFƒJ°†Þî<ÊÏóyD/ÝûMÎè¥{_†è•ýþOùM3Z% ŸÖZ3vH,u" zþ5¢GÚí>†Þxu_u@6ΦˆZZµ{èÛŸYAD³ÑuÄÆ Å/©úQ™¿NWÛ®Øý|¶Ö%mUýÛ ¹eã¿ÒÇw|ÕŒ¶5+z¹µjTôz'ýß_¿nö§éqrŽ¤?N—BìÐŽÉ`öú¯ò>zÍ?˜/¤)ÃsKcü®4½‹9÷kùT[Ö™càó´ÆÍ|èWño]6Ç`;Òt{ëô~SåÖéýú¢G:—O>ïY«¼l¡æw¶Ö§.~zòP‰ùNØ^À,zñ>zšfȃå­¢26º}{wuÂþ˜ôÚøAír…(–¡6“&öæGÜäÍ%)‡Õ«- a’eOÝVç^õ ¥Å{~Z¹ïÑ ®…sMh†Á÷šãëåÉ‚ºE!Ýuê,Y+©kú±ââîî‚ðáPÓ|$÷ã._)ÏÞ5ãÉŸ¿~3 ,¢îq®V4IÉÀtŸçžet¥GÉ?å0VX´çjLÄv—‡BÜ ËÏ;;1&“.tE:ÚqŠܹ—›© s~že)Çñ¸Ö¢nE{ëõ~ƒåJ÷¾œ^hCºÖ²çÞê¡x\ÙcÛñÂ9 KGׇáÀHàŸÑ¿:'V#tMapa¹Ðð«ìg®]«§£¶Eþú_î½ùGrÿ騃§)—ùÚ¢ýï#z¥i1‘â—z¯¨šö6T/>ŽJ×ê‰ ¦µtÖgUƒj_eßšÏÝõ‡¼lèDÿÆE&©bœ8ù&Œocô}›µ=¿÷Ò= z@ô€èшD z=€èAô¢Dˆ= zC¢ˆ€·$z¤}|'I’ôé×ÜVÏŸ·]Õ¿qälé1>*kèKnÝ„¼ôµÍƒvºBDî&Sõãçyü¯ÔÞg)W5àÃ|C5SB´½+þuv•óô¢ zZ(úÞuâúŸ¿¯ÙÎ5öÈõßÿ\ïzåç_«‘¥M7=«šÞE¡·N~Ü]2‡‹ÝhÚ½xÈ\Þ}? Øõo'e(jZe³[½­}A»—öw';žµ™k[ƒÆÇÝ­}íßÆó¶÷µÊ-$Î;ÎTftàŸkÿìòÌP§~pßÑÐÚ½rÀG„DrÎXÖã\’ †É§¤Y­%zÉs9Èt3xrBÓövP ….ÙyãÎUþ[‹^NÛçàµÊ-dt˜e>íÛ$NHV€èxÑãÛÃŽÔÖ}®/´w‹ÚŸÞÂïE_ÈÇ\~#‹ã"jšÛ'Ò·éÇ)`ªˆ5/’‚s)æê‹ÑY—ÃEÝŸT‹íΕò—)ôãôÕW2Jd2ÇgÙÃe÷v‚MŸ‡§Hky[õ-|$š" /Oc×—lt¹2¼z$ò53W™†NêÎ¥&ß:½í}”¤™lþd¦£T­RÆ6öìªêgÉZI\‹v/Ds·Ç«.OœÀq‚¤éè;kËDVÈ©ÇLX"~è™è‰^mžñ§¨û£„Y:áê=CÄèðá§ç§†f¨Ä~ŒÜ±ªdþ6ši/Ú9ˆ%’NPlšŽmŽ<ä§kª\éÞ—× ínöêhT ¢àÅEoÑôla»V¡ SÜß,%ÑWño•ÜßÏ2£WõmÑÄ&ŸC5ؽ4¾RZöTòYOaÕŽ]弃šBÞ¸=‚¯Ôµœ}ÜqÆ?VLŒŽèlÛìÍ÷ã¹Ç“T¥Å{ùÉW‘éTòÎ|“éK–ÍD{ ã9¢1T2ž•ÌZÕÌÏ|ÿž*žÙœÑ㠵ܭ"lÈÙYÊJËy›8ýÊò¢G:f'‡.“’%˜=ÌL–?½hƹÐ(ßÁÚ‘]&( w3³n´½‚<Ë]K©°H­—[§÷eôÚ㹸ôY” z^\ôJ…ÁO.Î' køÜ´þeDOË3$›¿!ùÿ­¢Ý•Ä:Ñ Á:ׇC_ªXšÒ²òôÝ/Eô¼Oë\çÚ~/éúFÙ·“ÏT½Úž¼=]-ÜKßù¡ÉL.0¢SÈc3:qIS_-»Î³¤ŠÉfCÂoÐ uÇ€c~##–è¹ß.¹;$Xyô0š˜ÑK÷¾ Ñ+ûýŸò›fDÀ«ˆ^0~ E¯²…Y>ˆJm™?¸øOF¾è5öhæC¿Šù»lŽÁvDÞ7úþm†Ü²ñßCã; ¾jæ¿ÑK‰^úÚ6Äè:bã†â—Tý¨‡Ì_§«m Wì~>[ëí±’¼s¢—[«FE¯wÒÿýõë6l{Z'çHúãt)Äí˜ f¯ÿ*ï£× yñƒùr@šÒ8<·4ÆïJÓ»˜sÿ°–ÿGµe9ö~i’¿†Û«0ΦˆZŠ1°{èÛŸYÑŒ¾ñrëô~}Ñ#Ë'Ÿ÷¬JÆàE?HßÜ]Ÿ2IϦ4x{’ÌÉõÝM:0VŒ«B;gÂélšeÓiêp}j‘þ–¯g†¥åB”òíÝÕ SøcÒkÓ 6dÎϳ,Eà8÷ÏZŠD…Þ¾C\ÜUì£ÿ‘'vìüçmD7¸r@2…›2 ¾7/Œº¿ùqEàj¬´xÏ¿0¾-'ú­³¯u®qíƒsÂ$Ë2$ž: º­Î#¼ê0…›ŸVî{–¸saÉ‚ºE!ÝuŸ•D­¤®UèÇîŸäá¿owד»D z}#ê3LRÅ8qóMQ¨ðFшD z=€èAô¢Ñˆ= z@ô€èÑ¢Dˆ@ô zÑ¿´è‘öñ$IÒ§_s[=ÞÆoUÿÆ‘o°¥樬¡/¹uòÒ×6Úé ¹›LÕó‚Ÿçñ¿R{Ÿ¥\Õ€ó ÕN\×ía…#&µ3Õ¥ÑÐ"Ñ÷®×ÿü}} È+®±G®ÿþçšx×+?ÓY­¤6º©èÁXÕD* ½=pò£â¬ûï'»^ñf4šVÙlEŽ‘æoÚ½´¿;i@^úÚÖ qÄqwkD_û·ñ¼í}­rÛОÍãLeî„Òû²a:JûíÅ<JÓ\”ö–òHèDߊìuD8@$çŒÅa 1Î%É`˜ÜxJBÓV§©}Óù¡Ð%;oT´Ï!z9ébŸƒ×*·;a–ù´o«2©WöÌÎõ•ÿ;Ú»IK I zÑ·$Mí°ã#µuŸU íÝ¢ö'†Kij ùؘËodq\DMsûDúâ6-+M-j^:$©Ü*’¹Õ=¢³.‡ Šº?Ûf¯»úJF‰Læø,{¸ìÞN°éã³ãð´©s-o«¾…CSA`áÅáiìú’.W¦jWD¾fæ*¾IݹÔäbB§·½’4“ÍŸìÏt ü´í>7{vUõ3‡d­$®E»¢¹ÛãU—'Nà8AÒtôµe"+d¯ã ¦,?ôLôN‰D¯6ÏøSÔý¡½,ð Vœ‹/~z~šah†JìÇH>ÝqýßF3íE;g±Dà ŠMӱ͑‡„zM•+Ýûòz¡ÝÍ^ÖÍ: ›ˆüÔD úÖä£×ôla»V¡ SÜß,‚|ôþ­’eûYfôª¾-šØä“ž¢»—ÆWºù|ô|ÖSFµcGW9ï ¦¡màk u-gwœñS#:Û6{óýxîñ$Uiñ^~òU䕼3ßdú’e3ÑÅÞÂxŽhŒ•Œg%³V5ó³3ß?‡§Š§ã#gô8hmE`w«rv–ò¿rÞ&N¿²¼è‘ŽÙÄɡˤä_ f3“åO/šñD.4Êw°vd— Ãä̬moùå éZJ…Ejm¼Ü:½/£ÐÏÅ¥Ï"¹X¦4/Å >k4ˆ@ô­½RapÅ“‹ƒóÉÂ>7­ÑÓßò ÉÃæoHþ«hw%ñ…D´B°NEǵEÑûRÅÒ”–픧ï~)¢ç}ªXç:×ö{I¿Ð7ʾ|¦ŠèÕöäíé‚há^úÎMfrâI‘[«šâKáÓºò_§R-Iq«ìóAQ–^•5xÉãq¯[™ýQQçjÇ#L9¦í^ÁÒû+kO‡ml ½¡ö*´—7ÆÍ»)KÅcÔ'Ö­oA¹u{_N/´ÏfoScµ§ôh—kŸÜu´#u@«EωmxïüËÅýÈýb¢¯9£çƒ{ûŵí#¾z --zék5Ž8&‹Õ «’«G¾-‰Æué;ËõHµg%·Vâ<Éê}ÔÐylF'.i ã«¥q`×y–B1ÙlHø Z¡îpÌocÃ’«šÖ‰¾V{•=ë$tšÕŠÒ2„'Z"úº½/£†ù“oñZk÷JóB”þ8Ý!*D ú–‰¾ Ñ÷Zï§D?ùÒ¢çfO)|±S4{ê~GÅòfôR×ö} ÂMNµfôh÷FþË®($QúÎ͈^n­½B7´ê ø=n³²e3z®éXh¤vKj£A%XCowe‚çù<¢ç~»äî`mäÑÃhbF/Ýû2D¯ì÷Êoš«Þ«ûÖT`¼½2ÔD úÖ‰^0~ E¯²…Y>ˆJm™?¸øOF¾è5öhæC¿Šù»lŽÁòH&ý¾›!·lü÷Pãøƒ¯šùoôR¢—¾¶ 1ºŽ˜À¸¡ø%U?ê!ó×éj[Õ»ŸÏÖºD¯ä›½ÜZ5*z½“þï¯_·aÛÓ‚ô89GÒ§K!vhÇd0{ýWy½fÈ‹Ì—Ò”Æá¹¥1~WšÞÅœû‡µü?ª-ë̱GðK“üm4Ü^…q6E¬˜ÐRŒÝCßþÌŠfô—[§÷ë‹é\>ù¼g}ô²¥Ðma¸ÇöðÁGÕ¿¼XzT zÑ7)ú±@úæîú”IòÑ?š!O’9¹¾»IÆŠqUhçL8M³l:M®O-Òßòñ•âÌ°£\ˆR¾½»:a LzmZ¡À¦‚Ìùy–¥ÇãþYKq‚¨ÐÛwˆ‹»Š}ô?òÄŽíÝàÊÉnÊ0øÞ¼0êþæÇý«±Òâ=ÿÂøF´œè·Î¾>ԹƵÎY“,Ëxê(è¶:ðªgÀn~Z¹ïYâÎ…= $#ê…t×}Vµ’ºV¡»’„ÿ¾ñ¨i>’ûq÷ÐGÅ°¸äÏ_?‰ÃÓ· ¢îq®V4IÉÀtŸ¿¹d£+=Jþi,‡±Â¢=Wëd"¶»<üâ®›Œ¥vvbL&Í]ͤө§(À½æo£Éö¢]®U(’ͤ‰½ù7ÉÕ8"ØTÑX¹Ò½/§Ú®µì¹·J(žvŠùÏU–¡y˜ôÃV=€èáÜæPqb5B×– «ßV­žŽÚùë¹÷f:û5Q OS®§Ã¢=ˆ¾ÉÈrÓb"Å/õ>þPoZØÛP½±Z=u“œi-õYÕ Ú×ý¥õ­ùÜ]OîR=€èAô¨Ï0HãÄiÌ7aD¡VÀD z=€èAô¢ÑˆD z@ô€èÑ¢Dˆ= zуèDÿÒ¢GÚÇw’$IŸ~Ímõüy¿UýG¾Á–˜£²†¾äÖMÈK_ÛÌ7TëØ"D?¼ºO°,Ã0étòøÑÐÑ÷®×ÿü}} È+®±G®ÿþçšx×+?ÓY­¤6º©èÁXÕD* ½=pò£â¬ûï'»^ñf4šVÙl9¿z‹nÚ½´¿;i@^úÚÖ qÄqwkD_û·ñ¼í}­rÛОÍãLeî„Jó‡å!]±Duÿ‰-•Ò)èDß‚ìuD8@$çJSÄ8—$ƒargà) M[¦ömLç‡B—ì¼QÑ<‡è夋}^«ÜBî„YæÓ¾­þYÄH× qXÊZ¢}KÒÔ;>R[÷YµÐÞ-jb¸”¦¶yH¤…šæö‰ôÅmZVšZÔ¼tH R¹U$s«/zDg]u&,¶;'Ì^wõ•Œ™ÌñYöpÙ½`ÓÇgÇáé‡åÎZ×ò¶ê[øH04E^žÆ®/ÙèreªvõHäkf®2á›ÔKM¾!&tzÛû(I3ÙüÉþLG©Z¥ÜhìÙUÕÏ’µ’¸í^ˆænW]ž8ãIÓÑwÖ–‰¬½Ž+˜"°DüÐ3Ñ;%½Ú<ãOQ÷‡ö²tÂ;Õ%z†ˆÑáÃOÏO3 ÍP‰ýɧ;®ÿÛh¦½hçL –Hb8A±i:¶9òP¯©r¥{_^/´»Ù«£Ñ:¦G»ôÁdii D úÖä£×ôla»V¡ SÜß,‚|ôþ­’eûYfôª¾-šØä“ž¢»—ÆWºù|ô|ÖSFµcGW9ï ¦¡màk u-gwœñS#:Û6{óýxîñ$Uiñ^~òU䕼3ßdú’e3ÑÅÞÂxŽhŒ•Œg%³V5ó³3ß?‡§Š§ã#gô8hmE`w«rv–ò¿rÞ&N¿²¼è‘ŽÙÄɡˤä_ f3~*üTƹÐ(ßÁÚ‘]&( w3³n´½ Áäx)©µñrëô¾Œ^@{<—>‹²æ'ƒÅ£ó¿þ÷÷éö€‚ñh­è• ƒ+ž\œOÖð¹iýˈžþ–gH6ÃçànS´»’øB"Z!X§¢ãÚ¢è}©biJËvÊÓw¿Ñó>U¬skû½¤_èeßN>SEôj{òötA´p/}ç‡&3¹ÀˆNñ$‰È­UMñ¥ði]ù¯S©–¤¸UöyÈ (K¯Ê¼äƒñ¸×­Ìþ¨¨sµã¦Ó€v¯`éý‡µ§C‹6¶„ÞP{Z‹ËãæÝ¥â1êëÖ· Üº½/§Úg³·©1É)=¢Xܧ\ð€–ŠžÛðÞù—‹û‘ûÅD_sFÏ)÷ö‹'>jÛG|õ>ZZôÒ×jqL0«AV $W};[ëÒw–ë‘jÏJn­$Ä'x’Õû¨¡òØŒN\ÒÆWKãÀ®ó,)„b²Ùð´BÝ1à˜ßƆ%V4­}­ö*{ÖI6è4«¥e7N´Dôu{_F/( ó'ßâ£j9¯Xârˆ@ô-}¢ïµÞ¯?ýäK‹ž›=¥ðÅNÑì©ûsÈ›ÑK]Û÷ 79՚ѣÝù/»¢Dé;7#z¹µjTô ÝЪ/à÷¸ÍÊ–Í蹦c¡‘Ú-U¨•` ½Ýy” žçóˆžûí’»C‚µ‘G£‰½tï˽²ßÿ)¿i®»éZ!íj=­½`üŠ^e ³|•Ú2pñŸŒ|ÑkìÑ̇~?òwك툼oôýÛ ¹e㿇Çw|ÕÌ£—½ôµmˆÑuÄÆ Å/©úQ™¿NWÛ®Øý|¶Ö%Úc%yç&D/·VŠ^ï¤ÿûë×mØö´ =NΑôÇéRˆÚ1Ì^ÿUÞG¯òâóå€4¥qxniŒß•¦w1çþa-ÿjË:sìüÒ$ ·WaœM+&´c`÷з?³¢}ãåÖéýú¢G:—O>ïY¿li×6Ý}Úâš¾•$넨{Z"ú±@úæîú”Iz†4¥ÁÛ“dN®ïnÒ±b\Ú9NgÓ,›NS‡ëS‹ô·||¥83,Å(¢”oï®N˜Â“^›V(°© s~že)Çñ¸ÖR¾zûqqW±þGžØ±óŸ·ÝàÊÉnÊ0øÞ¼0êþæÇý«±Òâ=ÿÂøF´œè·Î¾>ԹƵÎY“,Ëxê(è¶:ðªgÀn~Z¹ïYâÎ…= $#ê…t×}Vµ’ºV¡»’„ÿ¾ñ¨i>’ûq÷ÐGÅ°¸äÏ_?‰ÃÓ· ¢îq®V4IÉÀtŸ¿¹d£+=Jþi,‡±Â¢=Wëd"¶»<üâ®›Œ¥vvbL&Í]ͤө§(À½æo£Éö¢]®U(’ͤ‰½ù7ÉÕ8"ØTÑX¹Ò½/§¸yúZöÜ[5OÙ5±q@°…GE3ôÑ昅`<Z4£tN¬FèšÂàÂr¡aõÛªÕÓQÛ"ý/÷ÞŒBg¿&ªÁàiÊõÔcx@ô¢Ñ7YnZL¤ø¥ÞÇêM {›ª7V«§n’3­¥³>«Tûº¿´¾5Ÿ»ëÉ] ¢=ˆ¾õ&©bœ8ù&Œ(Ô x£€èD¢=ˆ@ô zуèDˆ= z@ô€èÑ¢D z=€è_ZôHûøN’$éÓ¯¹­ž?oã·ªãÈ7ØÒsTÖЗܺ yék›ít…ˆÜM¦êyÁÏóø_©½ÏR®jÀ‡ù†ê[„¦cç_ÿÖ‚èh‘è{׉ëþ¾>ä×Ø#×ÿsM¼ë•Ÿé¬VRÝTô`¬j"…Þ8ùQqÖý÷“€]¯x3šGM«l¶œ_½E7í^Úß4 /}mkÐ8⸻5¢¯ýÛxÞö¾V¹mhÏæq¦2w‚­}7µ1<ƒ ’ã€èDß‚ìuD8@$çŒÅa 1Î%É`˜ÜxJBÓV§©}Óù¡Ð%;oT´Ï!z9ébŸƒ×*·;a–ù´o«9©W[?¤|ƒjñÿ/@ô¢oIšÚaÇGjë>«Ú»EíO —ÒÔò±1—ßÈ⸈šæö‰ôÅmZVšZÔ¼tH R¹U$s«/zDg]u&,¶;'Ì^wõ•Œ™ÌñYöpÙ½`ÓÇgÇái#RçZÞV}  †¦‚À‹ÃÓØõ%]®LÕ®‰|ÍÌU&|“ºs©É7Ä„No{%i&›?ÙŸé@øeÙûÜhìÙUÕÏ’µ’¸í^ˆænW]ž8ãIÓÑwÖ–‰¬½Ž+˜"°DüÐ3Ñ;%½Ú<ãOQ÷‡ö²tÂ;Õ%z†ˆÑáÃOÏO3 ÍP‰ýɧ;®ÿÛh¦½hçL –Hb8A±i:¶9òP¯©r¥{_^/´»Ù«£Ñê¦Wö®'÷Æt•ÿ¿Ñˆ¾5ùè5=dØ®UhFÃ÷7‹ }…«dÙ~–½ªo‹&6ù¤§¨Áî¥ñ•n>} Ÿõ†QíØÑUÎ;¨)dhÛ#øH]ËÙÇgücÅäˆÎ¶ÍÞ|?ž{YXÃç¦õ/#zú[ž!yØü Éÿoí®$¾ÐˆV֩踶(z_ªXšÒ²òôÝ/Eô¼Oë\çÚ~/éúFÙ·“ÏT½Úž¼=]-ÜKßù¡ÉL.0¢SïYë¼lèh©èÇé›»ëS&éÒ”oO’9¹¾»IÆŠqUhçL8M³l:M®O-Òßòñ•âÌ°£\ˆR¾½»:a LzmZ¡À¦‚Ìùy–¥ÇãþYKq‚¨ÐÛwˆ‹»Š}ô?òÄŽÿ¼èWH¦pS†Á÷æ…Q÷7?®è\•ïùÆ7¢åD¿uöõ¡Î5®}pÎB˜dY†ÄSGA·Õy„W=¦póÓÊ}Ïw.ìQ AP·(¤»î³’¨•Ôµ ýØý“¼ ü÷GMó‘Ü»‡>*†Å%þúIÌž¾U°usµ¢I‚H¦ûœøÍ%]éQòOc9Œí¹Z'±Ýåá‡wÝd,µ³c2iîj&Ní8Eî5M¶ír…¨B‘l&Mì͸I®ÆÁ¦ŠÆÊ•î}9½Ð†t­eϽ¥Ôþ…2ÍÝç<¶d†¥{~Á¸­9@ʼnÕ]S\X.4¬~[µz:j[ä¯ÿåÞ›Qèì×D5Ì7TëØ"ÍXòê"M•Øþa=€è›}ï:qýÏßׂ¼â{äúﮉw½ò3ÕJj£›ŠŒUM¤¢ÐÛ'?*κÿ~°ëoFó¨i•Í–ó«·è¦ÝKû»“䥯m Gw·FôµÏÛÞ×*· íÙ<ÎTæN(‹>ŠÏUËE¢} ²×á‘œ3‡5Ä8—$ƒargà) M[¦ömLç‡B—ì¼QÑ<‡è夋}^«ÜBî„YæÓ¾M ¢à¥Eo;>R[÷YµÐÞ-jb¸”¦¶¹üFÇEÔ4·O¤/nÓ²ÒÔ¢æ¥CRÊ­"™[}Ñ#:ër˜ îW3)lwN˜½îê+%2™ã³ìá²{;Á¦ÏŽÃÓF¤Îµü Ú·ð‘`hŠ ,¼8<]_²ÑåÊTíê‘È×Ì£±WêÎ¥&ß:½í}”¤™lþd¦£T­Rn4öìªêgÉZI\‹v/Ds·Ç«.OœÀq‚¤éè;kËDVÈ^ÇLX"~è™è‰^mžñ§¨ûC{Y:áê=CÄèðá§ç§†f¨Ä~ŒäÓ×ÿm4Ó^´s&K$1œ Ø4ÛyH¨×T¹Ò½/¯ÚÝìÕѨºªè_ÎX†e˜B–ȳfˆ€VŠÞ¢éÙ Ãv­B3¦¸¿Yùè+ü[%Ëö³ÌèU}[4±É'=E v/¯tóùèù¬§0ŒjÇŽ®rÞAM!CÛÁ×@êZÎ>î8ã+¦ Gt¶möæûñÜãIªÒâ½üä«È)*yg¾Éô%Ëf¢‹½…ñÑ *ÏJf­jægg¾OOÇG ÎèqÐÚŠÀîV6äì,åå¼Mœ~eyÑ#³‰“C—IÉ¿Ìf&ËŸ^4ã‰\h”ï`íÈ.†;È™Y7ÚÞòËÒµ” ‹ÔÚx¹uz_F/ =ž‹K_µDµˆ~`´¿-½©¸£Œ·”K D úÖˆ^©0¸âÉÅÁùda Ÿ›Ö¿Œèéoy†äaó7$ÿ¿U´»’øB"Z!X§¢ãÚ¢è}©biJËvÊÓw¿Ñó>U¬skû½¤_èeßN>SEôj{òötA´p/}ç‡&3¹ÀˆNñ$‰È­UMñ¥ði]ù¯S©–¤¸UöyÈ (K¯Ê¼äƒñ¸×­Ìþ¨¨sµã¦Ó€v¯`éý‡µ§C‹6¶„ÞP{Z‹ËãæÝ¥â1êëÖ· Üº½/§Úg³·©±º‰Qó{"4¬ÑÐBÑsbÞ;ÿrq?r¿˜èkÎè9ÅàÞ~ñÄGmûˆ¯ÞGBK‹^úZ#Ž Fãb5È*äêÑ£ogK¢q]úÎr=RíYÉ­•„øO²z5ôA›Ñ‰KšÂøjiØuž%…PL6~ƒV¨;óÁÁ°äÁꀦu¢¯Õ^eÏ:Éfµ¢´ áƉ–ˆ¾nïËè…aþä[|´~Æàv7wh@ô´Tômˆ¾Ôz¿vüHô“/-znö”Â;E³§îwTÌ!oF/umß2 ÜäTkFvoä¿ìŠB¥ïÜŒèåÖªQÑ+tC«¾€ßã6+[6£çšŽ…Fj·T¡6T‚5ôvçQ6 xžÏ#zî·Kî ÖF=Œ&fôÒ½/CôÊ~ÿ§ü¦¹î>ÍèGzÛ‚‚èh­èã—Pô*[˜åƒ¨Ô–ùƒ‹ÿdä‹^cf>ô«ø‘¿ËælGä}£ïßfÈ-ÿ=Ô8¾Ãà«fþ½”襯mCŒ®#&0n(~IÕzÈüuºÚÖpåÀîç³µ.Ñ+É;7!z¹µjTôz'ýß_¿nö§éqrŽ¤?N—BìÐŽÉ`öú¯ò>zÍ?˜/¤)ÃsKcü®4½‹9÷kùT[Ö™cà—&ùÛh¸½ ãlŠX1¡¥»‡¾ý™Íè/·Nï×=Ò¹|òyÏúøeK?q@…çzÔ¥§n÷³ w)´D ú&E?HßÜ]Ÿ2IϦ4x{’ÌÉõÝM:0VŒ«B;gÂélš-D®O-Òßòñ•âÌ°£\ˆR¾½»:a LzmZ¡À¦‚Ìùy–¥ÇãþYKq0Sèí;ÄÅ]Å>úybÇÎÞFtƒ+$S¸)Ãà{ó¨û›Wô®ÆJ‹÷ü ãÑr¢ß:ûúPç×>8g!L²,Câ©£ Ûê<«žS¸ùiå¾g‰;ö(Œ ¨[Ò]÷YIÔJêZ…~ìþI^þûÆ£¦ùHîÇÝCÃâ’?ý$f Oß*XˆºÇ¹ZÑ$A$Ó}Nüæ’òǹpOc9Œí¹Z'±Ýåá‡wÝd,µ³c2iîj&Ní8Eî5M¶ír…¨B‘l&Mì͸I®ÆÁ¦ŠÆÊ•î}9½Ð†t­eϽeõç<剳Ç–ûÿYxxAшŽÀm«º¦0¸°\hXý¶jõtÔ¶È_ÿ˽7£ÐÙ¯‰j0xšr=õ=€èAôMF–›)~©÷ñ‡zÓÂÞæ€êÕê©›äLké¬ÏªÕ¾î/­oÍçîzr—‚èD¢oD}†É@ª'Nc¾ # µÞ( zуèD¢=ˆ@ô zÑ¢Dˆ= z@ô€èшD ú—=Ò>¾“$Iúôkn«çÏÛø­êß8ò ¶ôÀ•5ô%·nB^úÚæA;]!"w“©z^ðó<þWjï³”«ða¾!éc‹PãèûI$M'·!{­}ï:qýÏßׂ¼â{äúﮉw½ò3ÕJj£›ŠŒUM¤¢ÐÛ'?*κÿ~°ëoFó¨i•Í–ó«·è¦ÝKû»“䥯m Gw·FôµÏÛÞ×*· íÙ<ÎTæN Ø ²‘…>Q^=€è[‘½Žˆä\)YbœK’Á0¹3𔄦­NSû6¦óC¡KvÞ¨hžCôrÒÅ>¯Un!wÂ,óiߦ®±Š°“IÌv °tÀ3ˆßv|¤¶î³j¡½[ÔþÄp)Mm!sù,Ž‹¨inŸH_ܦe¥©EÍK‡¤ •[E2·ú¢GtÖå0AQ÷gÂb»sÂìuW_É(‘ÉŸe—ÝÛ 6}|vž6"u®åmÕ·ð‘`hŠ ,¼8<]_²ÑåÊTíê‘È×Ì\eÂ7©;—š|CLèô¶÷Q’f²ù“ý~ð.åFcÏ®ª~欕ĵh÷B4w{¼êòÄ /¬úFßY[&²Bö:®`ŠÀñCÏDï”HôjóŒ?EÝÚËÒ ïT—è"F‡?=?Í04C%öc$Ÿî¸þo£™ö¢3X"‰áŦéØæÈCB½¦Ê•î}y½Ðîf¯ŽF«™^=röö)á=Ï$z‹¦gƒ Ûµ Íh˜âþf䣯ðo•,ÛÏ2£WõmÑÄ&Ÿô5ؽ4¾ÒÍç£䳞Â0ª;ºÊy 3Õ¶=‚¯Ôµœ}ÜqÆ?VLAŽèlÛìÍ÷ã¹Ç“T¥Å{ùÉW‘STòÎ|“éK–ÍD{ ã9¢1T2ž•ÌZÕÌÏÎ|ÿž*žŽœÑ㠵ܭ"lÈÙYÊÿÊy›8ýÊò¢G:f'‡.“’%˜=ÌL–?½hƹÐ(ßÁÚ‘]&( w3³n´½å—/¤k)©µñrëô¾Œ^@{<—¾*‰jÓjj×Ö5ú>ʽ1‘©Ðb?ÿ ¢}kD¯T\ñäâà|²°†ÏMë_Fôô·›½M=žÒ£–-"qDF– ‚Gô£>š_Æш¾E¢çÄ6¼wþåâ~ä~1ÑלÑsŠÁ½ýâ‰Úö_½„–½ôµGŒÆÅjUÉÕ£GßΖDãºôåz¤Ú³’[+ ñ ždõ>jèƒ<6£—4…ñÕÒ8°ëûºüñ¶WÀ•;u‚vq¥ííDUºâ¨û_~ÿ©­XÔb'ÿl/õ‰¾à»»ÚLQ¶Ë95‚ŽŽvÁåT¤F%qÒ9`Ä•Ûó½g5‹ßQ´+ êVÒÝç¾RÑ*Ue=ã${ò{~™dãY¡é ÿ£«¤7‚[ÿóçø·ÎöÿUAñ¨{BÔª6ŸßZž|)‰ø凎Æ;½å{#»†'¾i/jukKSeöÕ®!î‰M\6»©ýa§¨t{g'—¤4ÀòØäö²Î§T Å«ìxØɯJMˆZÜ ðRÅÀÖ«º÷Õé…3^çï>zRAySÅïòÝñ»íí¼²›!, Æü‰)p‡f•$ÅÐ5ϳ)¼ÇÕW}‡W«úoLÿþïqn ý"ñ¹Rñ-7¥¿Óð@ô@ôý G–‡f¶på·z{?¨ͨʿì3ÌZÕß—äBïv>*ò…j_ì‘véniêù~w)DDÑD}g˹Òqâm¼Ò„`Z†)==DDÑÑCô@ô==¢@ôˆÑ zD€è==DDÿ¼EïtƒÝ*´}ûãã‚‹/ß‹ß>‘yœÒ+C:aŽOTõ?ßõzÞeë\J5ÿñ/Iç ~6»ÿmï3Y¯ÏåR^i4ù´Eþ‰Â_26+¦l†è€è)úðûüŸÿç¿®SÈ+îw½áçÿþŸŸù÷ÂÕÏtF•Ô&àfc]i"ÏÀëåßüÞc®ûS~=ÐsØhžšÓñ¨;¿úUz!«¶2ñ¬×ó.;4øÅ7©C#zêcãÙnï‹ZïÖÅü¯öÌ Å;"3;²ûzÒû2[P*ˈ ÑÑAö:~M9¿õv°ô´æ|»UPQ#`_îOBÓ¡NS;<Âùèê:Òƒ=Ï€g!zuÒÅ> ^ÔzŹÒÚÿ^Óç\ÄÞ…Ý©v!z ú!IS{5þPš ›^ ¬M¸*KS+ÎÇÖþï²»ˆ¬ÐÛµüÎïëT+M-+,«^ Ê­G2·¾Eï•]à %sÂò*o+f¯ûéGA#ÿáï¿{TŸZØÒÑùõw_×${õQVn«KøímB>ŸÏ«É¼šÌûù‡ŽÆìž©Ú}c~|x»gÂ7U5Ë6ù~B@`Ln£ ­ýÑÓojo…Èš%ËÖñÝO¤9T¶JEYÖ…ŒÆÇ¿}“RÔÌ'¾ ­­ñ^Ô‰Lœ½N´b!Ÿ×Ò\_”~SIô¾a·Ê¸Bɤ½m-%7Ï+íC¯àøRâÛ'ß>lok¶Ô6 äéŽû>6³½¬s·Ê›ZZÅyà::Ûšòc»ê j½ª{_½^Jíø‰s­Ó³Âr5±¾xFJÑGø]ÌÔ\÷÷ô»V#ý/B!}ÿ’dÙ~&½Ï¥‚6~¾<é)ëìõ’6âÎy>úò§ŠÄ§Qÿ8ÎOK®ø‰3´Uñå-PUVdŸÔæö²8i r¯€˜ÂŽ_þýõí ’°ªä‡¿—öÈ)ª²fù&·ýÐÑñ°13\|ªöò >ë£Æ¾R³U”ùÙÛÿýš›ÒÙñ½Î&5~]5 DU5tT'“åy›ÿírÑ{…¤µ|SŸê-¿$H«X—ØýèÅïFËãêkòö­l¯Pî Nd=Ðíí¾øò:ŸÅ­QRëÀ×ÛGï«Ñ ¬‹EßÿP¡òf™×ù;üúîg =ýЈÞÛólJskæ•ôVñ=|QXÿ|Dßö«Â$…Hg<ƒRZ‰Œ/¥8ç¾°ñ†¿Tô¥\éÚ¼# ¹E—$·".ær¥mî£ld‰ LÑ7Þ—ØO’ˆÞ÷zëoßf(ݸW]s×&·?. ðì—DÔm¥ø¸Dr@÷or‡$Å­÷¥"A…R–^Ÿ¨ŠäƒñD—[k¯)u®ÿ†öî1 ¬ wxµwâ£.†ø³v }@Ûëé‘RÒ$Š»…|·¹Iø÷ŽÔÀ!XoŸ½¯N/¥=ú§*¤÷ Éàq0ê0Ô¢‰íjÕ“~/9s?7ÑSFô"Å%‘ÊoÌ"G2ZµèU—õ‹oæ)œ¥Í $÷½Æùõ»,¥óºêšÕõÙ¾R·U*ħ°'Éûh@äy·”×t“'o–_<ïç§E„íU+>ƒöô ¹ŸžWÑÄoïÔå\ö:ÑSm¯÷Åû‚ŽŠ¤0_OÙmˆT‚?$¢ï³÷ÕèϳéßüÚ¬êÞ½çÙÔÖ–[ ÑÑ™èÏxF^‹’Ü;î%úÄç-zQôÄ%2Ï)EOî ›âÕ‹èU•½T,(W|ɉ*¢g]È{úÏJ¥!‰ªkŒèÕmÕ@EïSZ^V”æ=d½hÓyÕ±Ô[êé|ÖGázPçQ¹Âþ|6¢»ÝÙHwÆ "zÕ½¯†è½#Ëþþ4?Œ¥ê€fîmÅûH=ýЉ^áü¥(zŸ˜šù *߈ôºïÿë¡ú¢÷»Þø°XþÞ§ïù˜ø+A^ê=£,lÄÈŸ‡ß`·9aògôªD¯ºì¯àN{ù³Ò'©×ŠOî${5Üûrå?¾»{^é+•5Bôê¶j ¢Ljûß?ÿü­&¦ƒôDrnè|,bÇ I¬xôó¿ºß£÷‹.!êÒ»¤y_½'+-0¥ýIÙUù}#î·]¤p¤©<6¼½žÁi\þP–lŒÁõ¢¶ßþóH)¢øzûèý¾Eïu.û›TE©¸Øòo ²”.& z úAŠ>®¼ó—?~þ¶½µ(ÚOvò.jmÿæç?~é,“Ž«b»UÓù¨³££³SXÿffÛ¯O›ïH#CÙeñ(åßþøé›vñ?[Kbüv³¢ýÉ“GB>AÍeiÒÑ3ð:›ÿý=Þ£ÿý)Ÿ}]þxÛ+àÊ:A»¸Òöv¢*]qÔý/¿ÿÔV,j±wDÉ“¶—Æú‹D_ðÝ]m¦(Û圌AGG»€àr*R£’8é0âÊ¿íùÞ³ŠšÅï(Úu+ éîs_©h•ª²žq’=ù=¿L²ñ¬Ðô†Ç¿ÿÑÕGÒaq­ÿùó?ü[gûÿª xÔ=!jU›€Ïo-O¾”DüòCGã‹Þò½‘]Ãß´µºµ¥©2ûj×÷€Ä&.›ÝÔþ°STº½³“ËNRàNyl r{YçSª…âUv<ìäW¥Ç¦ D-nPx©b`ëUÝûêô¯ów=)Q5Ïûra]¦òµ%DD)pK@bèšçÙÞã꫾ëUýÇ7¦á_ÿ÷87Œ…Î~‘ø\©ø–›Òßix z zˆ~#ËC3[¸ò[½½Ô‡fTå_öf­êïKr¡w;•FùBµ/öH»t·4õ|¿»¢¢‡è¢¾³‰å\é8ñ6^iB0 ­Ãˆˆ¢¢‡è€è!z zˆˆÑ^&Ñ‹~———ã7~ã÷ðü ÑщèÃÑ×¢¢¤èq.`8ÑÑ#¢=DDÿ¬"zö. GûÙæ³'¿t'ÚÂkËw_)Ò*£7¿>ëã”Òç]vbË4¯-ŒY&Ögî”=¯Ýÿ‚¶÷™¬·0f7swT>Õ¾+É8»ÏÅ–nkgG·[ùAx=`È"úIšZ¦ë.æuqr/¬™¤¥9‰q$A}q^ÙÆü€Dô÷V­Í#½Bȹ°mŽÎ¸é¶ïü-©¸¼8éËwl§?gÛ…®3Ü0¸kzë3úÜC C{ýRœôÅ÷À,öó.;4ä†,czè©g»½/j½åʼnó¬?N*!ýÛÍÎÇ Å{¶,ïêa×SélDô€!‰è‹bw36nsr9•!=­±3N¹0ÞscìŠ)êO´B.ú²Óõl.•4—0ýîwý÷®Ó%(w…óQ›§ÓOf²ËÁ3½êcãÙñ¢Ö+þne¶¶þÙ ¢{~‹ž9òýZšò©ãÚ ùˆèCÑïfŒ qs8tCìéâøCöÎFbJD_|óó5ôéÆŽRÿ–¤œÚÀ°±4\à•SÖ·èEQÑZGëiãt-mì¤0ÜΤ–¨/zö½«Ÿot²wptd8:,Þz*.-½pn›ÙdÆj'këyVs×÷<èºpÁ<«y}3Jû(+;ÕçÅ}åÆ°µs`8917~éÃœ4máêã7{\§„¯1·>Õ3¬SU³t“§›0‚îæD|´ÚÑž>wÆœõ>é¥òó|;ÓÎŽþ–ÙcÕ­RQ¶8ù«Õ³ çÆ:°Ì‰Étr´³[õÉÕÜ!º»^V˜ä³}±½½c±ë²µûƒâýE_VpÓ{‡³½hWØÛÑí–î HUjZifðnæ왳­míèöKׯdìŒ.TóØÌö–¤yo_é겘ɰ§Ûد|¿û6ø Ö«º÷Õê…²lÏ…æ+Âò{ïçÒt¯õË<.刕âôà.ï†ä ¢ YDÏ<——àÁØx!—¶Ñþ༇™]}ÿæžue²Ôý #úÂë‡í<"eçè’¬ {ì™G“¤k(ŠÝ6cî«yeå¹çW˜ÍÚsEtBÍØÂðÝeWU¶œá¹ÌvGh¶Ä¾ì»—.4?ÿTv/1ßØ3mÚî8åmRY³|“í¦ÑéÖ«¿L(ÕYš›™UXÖ÷¾R³Uû97>Þbc@F‰ÔUþ«æ¹Ë”:ØÔ }‹št…Å!»³'Óå¢/M?í:çVJQ™ì’àÌZëuAÝ^òB\gm¾(“]Yn¸;ÝýJaÿ"ëno»kÇ•¦~áìv1¿|(ÖÛGï«Ñ Å ,{TòÚãY7|sÎtý ôcÉÅd£îY*è€èUDô7ŠÙY¬e._^9á"¾‡_ÿœDoo<ƒî(‡n) …eŸeg³\˜_¥+FÓ%IٯɕŠ~÷©Ô‹ã.9 9o'~¸DÚfÕe‹¯íuÜ¡8¼®èú.KkÑç_p1œý•Ò{Õ5Ë7Ù‹>k[ø=v¿$¢n«(Å·d‘ï½îÿ8;±r†à¡rÜ~Ç÷"ãÏ«ïM·>-}QÌvëõaJ»sÏ­¡îÓP’|l±Íúc!Ñ é÷‹Ët }@ÛËÎcí]édoïàÄt^¶Ò~*ÝSá€ðzûì}5z¡ìÎ鹆Ρù$—T~íέ²¬(Åg ckèÝý¾€è€èUDôâ»öù‘[Þ´xc}X®øþý‹Žè‹ã0÷Ä*(ˆØÄüL2ZµèU—Í ^¶Øçnf8žî-ú‚‹Ë­¾P:¯«®Y]í+u[E->Å=IÞGx ß«U"{-v”ºÏ ašZv_§‰q°;w³â3hvAzlð ÷•Nô…ŒuGcóʆJô”Û[œð‘#ý=¿¤|IÏ••dx2C"ú>{_^`g˜c¼¬÷½ûÒ[Ÿ17…w-.»´rÉiéS#ň¾O z zêˆ^òð;öâÕ;’‡„Ê¢_48Ñ @ôìl/çE_ÞVŽžŽØ¯ ¾¯ND¯ªlqÜ>ƶ(5"ú’$[•†$ª®y0¢W·U=û^ÔÑÝÛvð¼YÔψÞ])¢/PˆèK’>an §¾Î.ÈÌ*`+ì<¿ås·wïOÕÇÆ€··8þÃ]qOæ¹.òê)ú Eô}ô¾¢/¾¶cêŒ÷{¼/NðPju^èZ§/ÒØ=#zÇ­{ ú¾#zÅÓŽ¢è .¹-Ü-9•åß8¹Þr‚µ—º¢/Ï»°ÊzßµBù™?õRÈ•;lõžÑ_;hË8,$~<|n—í¢ÏºžÑ«½ê²åìL¯ôm!Y’“miÎÅŒ¦6$}yQÌÖ)VŸ¤*½L­²æAˆ^ÝV Pôe9~ö£i4C·ˆ¼~Åùìlÿ56›|S¥»µ8=È}®©~×{ôeyQ{­;qC> ­¬8#òÔ¡Y¥]Ï0fîˆÌ•ÿ± îcÛyŽ4•ÇÆ€·—yz‰Ó±[R•–f]8`o8v®bD?ˆõöÑû}‹ž}ûøÛSÞ½ZDözÝ'NÌOâ$oÙ³ï_>ààzâv©ò3z7çUÿ¢½É¸I³m]\Ε¾ÃuÀÅöíIãLl¶‡JG•Üöv³™»€¾ÐÆÆaídz3ž±ì˜t08;+P"ŸªÊáJÿĽwdª=¼N|Ïüõý-}Š^ȯ½uj³íÝ4]Kç¢Ë‰úH{ Ú(=C#)†´ÆÛú±•kèUŠRôFû3¥ÍhÍñ˜j°ºëâ@ôˆPRå4FgU\“4¢_×;¢ï6¬ôV»é¶UCé¸U©_ÙÓ[_•´ÉÈhSÏÀ›¢†®RM=D€èÁ àôðËa×µrë c,Ô5ϳXI7Æo1Òšs$©† j’ÌÕ2Ü$}F¯(úÖœ¦è0ÎUö¥ß’wñ|Ba=O h->·\ß`}‡sûÀT«ýQE¾àTÞ‰‰¼UM(×@R ¢‡è=Deì‘5Ö“uFˆG¿kL°t~?²Tz¿^ИqzíÌñ4ÚÈñ3מJoºW½[è·zŠ¶¨¬öV‰â}vNš‡ƒ…®†¸NíÉŽ‡®TˆÿÈ-‰<´hŠÎHm„¦å¢ýQ¢¥J54•‚è!zD€è=¢@ô€g}Z`^ z@oÑ^& zàU¢ÎqSý(Ös\ýàÚÚ†¨M·ÜMÍv¤!@ôiÅ+ÞTQ¥û³ÛAô úa/zÕÏP^-Ñ<þ€Eß³,ˆ~Yž&A¥ë{%Š«ÙÜí«C Ó14ÚèÉ®Ý`{%™íáĪè}ãÅ™qƘÚî‹,#d¢7[ÿ鎅ÆZ#GhO[y*«IZ[ï¸Ü{G¦ë8œ—ϲה¸ÉÈÐM<nëí;®¨~}êç.¯kÒFj™Úïýt¹q—èI*ïYÇ@ôÃQô4Tõ½#z=šÖüã+¹¼ªØífZó¥`I“Ì*‹>!(ìv9‡ r¼]'è8‡?Hkeå‘ð€ j“>´Ò2Û•*žÕ–,.7Žm°Ôô7ÖoMæô½^uÒé6Äo6Ôž,­žàUÄìœ:’f,=yæ\å² ú—LôâD´Òd6-é{Ì$`É“ÌR5¢9m§Ù¤wEšÕ¦?Î%ªNº¸>f¹žéδfŠ¸¼ü¿ÍÒ±ñ§¬«]3Ñt‡èš Ïõª“NWôƒ‰ëd±¹ *„®)=Eæ\åT¼ úa)zõnÝ«~Fß=ÃIÚll*=yå Ê©x@ôÃØõj ÆSLK!zi’YZ2Zèڞ̨#$wv›M‰^Ñ_Qˆè'íGô½ràJ‚ê¹:s=3.:·8”Õ*Tc½ê¤ÓýÏÀÄý–¬’ºHæXiDOž9w¾)=è¦Ï™‹”è‘j–*1i’Y¥ë…­&F®Õ Ä ñ.»ÐÆÊD¯G=ûCq²Ûºä#³µLw¦HžÑ“åÀ-'ؾÖcMm§ú¡,Hïk½ê¤Ó4Äo26\^)*Øšwjá(š‘ô=iå=@ô5Ý«ü rªYê °dIfà•œÛleh`1óíynûW›êËDod¶þ“í6ÆZ#iZS–ŸHo”º'É+†_H-œÞu£½Ïõª•N—_—|ÌyŠ‘ù «9›ö0 ºFÝ“TÑDˆ€èÑ¢Dþ2På A:€èÁp¢%cßÔ7Þ¿Sãl`sîA÷$òDÑI+MË5²V0@ÑSÕ ¢/ZôÜÜ£–£õ§èŽc„Vñ!z€èÁÐRÙQgæ‰"é$4œ¤MÆ“%“ÌK朙´b¿›­Õ›S'YªÊ0+KíºÆÃÝqÖ4s=‹¥ÝÉmû}köá×µlX1‡-´lüËdSáPe•¥XÑD zq¶VŸ®(ê.ò[¤7™:ì4-¬®k`Q«H»9GgêXû²{æ'jRÃÂÒëZرÁQ…òyçšÓvšhÚ–sr<,ÆÌ>],)D•U–j9iÍ z ZôúãW^“e‹iº±v¢Š ³â´1n²ô0-û-ŒÝT$¢é¦)aÓÄ1 +øâ)÷-^³<žÇS‘U–b9ˆ DôJf¨Ê0K=C¾*u1«tGKò®»âª«"Ãì€D/¨‰pÖ¢Ñhc4ÅŒ)ú÷¤=éÍ”Ye)–€è%¼¼ãoê.ô.™¾5ß›¡E›Ô%z=Úk³ÞOx@57=fiªÊ0;Ñó+ƒíÆh/>{¯¼BJiêÑ™[“š(²ÊRf›ÑjáÖ%q4Õ3ž2ÃÊiû!¦Â¨{³õŸª—av¢'J½ç6tOjRTÿy†ö„w®7Pe•¥Ì6 ¢ýiÙ@ô=ÑCô z@ô€è@(ü ôú¢ Ñc' ‹…¾CßAôˆ²@ß¡ï zdY ï z€, ôDD z€, zôD€èD¾ƒè_98)nOEj8ÑC軿ީ¢J´f6§ÉÑ™²èýè ¢ý#hÊ Øá`1Nü=c2wÝ—‰ùp´õ¹’$ÐDE¤ûÍIkƒ‹¸ýsë;çþ¹ýKfL%î-Óùï){Vû¢/Lô£çWòˆÖúüÈ]ÓFŽu¬…èD?¤_³Ü µ´ÞÞ{!»ŠÓZW{”©?jÚþä‚èy¥anæš“ÝÂØ€­¼Ûä§>qNOÓ5¦™èY,=‘ÍûuJ„èÁ‹=¿1ãóù£4aÕœnÑ'…Ý.çDCŽ·ëçpÑ9iû,4§n»˜Ïáó›+³3Ä÷ ›3MÓž±/º´…à„º™é1ÏU@õ¢ûá¦Û„Q CªŠËÝ&Œ± «i}Gš0éZ3vÇT«¯Ì« z²®‘@~êKÜ0ž¦ëXÔ*R{ÎÑ™:Ö¾’+õO‰=xa¢§ÑFŒÖÔ=‚F;Ó-¨ •ü²³9m§Ù¤w“9œ¤ÍãWÄÔ+_’·é­Œi}IJ¼¬tÃk±{D/6D9·âz“Ò/Çãõ×Þ:S‘|GÒ4Þ:’Þ$V»îU=Y×ôB~ê‹ÞÀÀí&GÄgì·0–ü§§Dˆ¼0Ñší™S^YYÛDÈN4]G#¿.õäÆú:ãttuµGÒtVÅ5ÕGÚkY|˜£ÚEW¸çwÒM5i4Úkã-×H‡˜rK"-š¢3’F¡i`¹hãˆ^¿!Ëo›ÝëcÅïj2|{íçñU½ß£åánf¯®./Ò¿:ïÑß Þ·Ør‚†¸k&L¡o9_Á#=õQ‰^ýS"D/«èÿ‚`f<ôD€è! ôú¢@ú¢È²@ßAôY@ô¢ÒŽîâ9­r(”rn¹›šíHkVwy¿*=€èÑwý_^ñ*ðœtш²@ßAôýséâ?É€èD zôDÿ²ZžÒõÔ‰{åXä—ûÏ×]TÉ ›oï2¡i8„‰§ùà$o™d¾ëv³\ôõ÷?³ÔµLLS‚›¡Ñ¦¥¬äÙoùõ©Ÿ»¼®I©ej¿÷ÓåÆrGS-'MŒKùaðÊŠž ú}Ñ¿$¢'M¼H–c±ùÞ'oè/©_Zéë®orïycâŠØ†îˆž(·7¬Žçs룗™Ù\mjú^ŠÄ¥ ƒñ5Îúo~ïo¯ûæqÉU€"äÙo7˜¸ß’­±.’9ÖDÑS,'k!Õ‡@ô/ ýç©EO•vVô‰)#GNÚ!ò9¿Ü–MÛåJ°‡è…‚º˜Uõ¦h‰Sn÷Z'Yö[ACü&cÃ5á•âµäZ8Šf$q4ÕròR}Dÿòê^Õç¨/’§å$nÔ£i-½Z/þÄýO¦ÒhóüÊù½D/6Å»M ˜/Ú×òì·üºäcÎSŒÌgXÍqØ´‡iÐ5êžb9i )? ¢C/ÿ¸¥Ž]°ü-;=xyÔÇo7Ó_U+À¾ˆ¼Tð+Cìµi£ÌWñ°7¢Dˆô¡ÌD3üV¢Å!*ã* qnãg¯àþŠþù´ Dÿò‡ö=ˆþåó»Ô¼•1+ h"´tuuMˆg¹!I+ž]Çlãñ½vÆ£h4«mEùçÝçèkÐhzÖ‡ãk%¯Ììˆ]6ÆcFÐFhhÌ;Æ!=i¦ÚžeIZ¢ ±{ÄΤI`E¢×£]x4¹†×œïÃÐm·\ß`}‡lu¤™jIÊöhˆ LôBn¡ßê)Ú#h4í¬‚4 ¬¢ç¤y8XèjÐh4 íÉŽ‡®TPܺ'ËTKV¶g«@ô€èÑ¢ zˆ€èÑ¢Às=~ðƒüà?øyYþ¨‘'; endstream endobj 379 0 obj << /Length 2179 /Filter /FlateDecode >> stream xÚ­ÙnäÆñ]_1€B%¦v7¹p xíÝØ×°³z°^ Ù3"ÄcÌC#-Œ|»«ºšÇŒ¸’ d¦¯bÝ]G¿¾¾øûÛXn’0ÑBo®wÃC-ÍÆh2Í7×ùæCp}k/?^ÿ Ù’›Ps¾acË¢*ê´!—8£$”‰!w6í‡Ö^n—AÑÑ8tv7”W°,Ø5-îŠÀ^r<¤Õ¡´îˆÇ[[ÓY5”}Q7U‘–ÛÖv‡¦î<ΪÉméñ¦#¡_}osúøæq$[Ô{šö·òhË…Ølc&Jl¶GEÜßÕÍ¥PÁ¹Hd` \ßã_ZÚ:³tpcû£µjÁ+m¤52’DÀ–bYSçE_4uZÂ’ÄOMÑuÿ¾,î.9Cß6MÞ…—ÛH&Á÷þ´²[?+ú!EL­{ØçÁÈ4‰^šF€½ãh´Í/ŒëÆ¡É“¶îIA‘!O5ÔõÀB×Y¬¨/#†úaASÞ£AmG') ‡¦„ šò‘6Ê´½äq°·´¬‡êƶ4ovä!h͇+§ƒS ÃFKf¢EßôiÙ92^÷¸{ë)dMK”“7 Si›V¶·íÈkët@²j’ÕñcâœF:¾ñAò dŠpÙ®ÓI°Å[BïvŨ×”ÛvT)”'WžÚ˘ó†àê¦_‹ÃP¿O‰±ð óÐbåŸ?ÉðÝPUPÛÚî4ç"ߢltHÏ/Ó¸Hß@·õj”AH!F?DtP€@å–ž„›ÿo(tóçž‹L Ú˳ÿl$$g?s‚ÑWßúÀCðÿ¶]‘Ð’$øf|Öç¿C Fk½ŸÁ6&ï =IÇ] »‘`,x—>x_Á²a¬“1ôá Ö`‹Å;b^3)ÁZ,|u¤ 7óK¥| sn©o»Û®KŠc¯IÂE¢³å TÕd˜÷}ÒìMÛ6¾ÚøäóNZì'WÑõÛ§ßfM¯!G^€-ìkEñ$ƒ‹Ä aÜ*èe'a¿vK¯OzMÀ(gŒB©8Ò3F¡¤Jaö?+5btß øóî÷ã×ØOþ›hãvÉÅ›ë‹_/ðUŽmør¿Nä†CS¬M´Éª‹Ù&‡3 Ê$Þd  ÇTSnÞ_ü|ñÚ? b§l(H±p¡KŸ—ôÀ§’>‚*ý;¨ù­#ÑÙÝ‹Ìøü€ùÌŸœ¾õIÓã œ@ài»q³èióˆá»Êœ–®•Ñ§=Ä—5ø‘Ñ“›¦\ £kvFf¦¯(0Ã4/Zè^]º„ìF}ÈRBˆÑó3ÃØå®é’æüJêbk €§ôdSUØéå>–nÁ$ #íµ}H³;ŸO†R …4DFfzëkÛ¯0¡]Òñ‚L‰xÐë¡_h¦§Æ dO½¡(—â„PšLŒÿwñLiðí'»û @‚¡Ç©îì“I×Æ›ÄLƳezÓø·$ÜÇ‚À%R¯«5ÉQt³VE-$‡[ ¦Jk‡9žÍЃڻç-Wh~}†^£µ_¼è2ÆfÑ·Æ `HÆYBƼњúêÔËĉ>öª W]P \»$ŒÇO‘Âïbï endstream endobj 392 0 obj << /Length 2803 /Filter /FlateDecode >> stream xÚ­YI—㶾ϯP|R¿7‚±“ÌÍN<‰óbg™¾Ù>°IJ‚‡"’šöøק pQS­±“‹€`ªPõÕ¯ß|ùNp»ÉXf¥Ý<î7Æ2›©MbãVlËÍÛ¹ávý¹êz×6;#Íöœwù©ªŽ¦û6 έëû¸iŸŸ\ý‰ÆCþ¡ ËCKÿOý ` ~züÛ†ovR±,ð}?”Œv|Óu‘ëéÿû¯èÿùXuNÑVû½+\Õ ´pÌûÈ)ò.Ú¦ºÜ5UI ×t/+€Û;Ø1?S8Ô¿«Þ•—¼¦ÝeõÑåMQý‘¦R[Él 8Š,³q󡫪À§ÝÁRÙž"+!Xf ±úêÛ?ºBr™®hèûËé)^A$èàRò.ª/k|ñËw©š]3’N"%:y=ÛËáø°“îçX…WøÜ2`¿„ý ýåtÊ»O+”UÆŒÍâ>0‘c[YRëmNóº=¸ù’Uª™¶*ÒxvÑUµ;¹&ª–‰a©Mâöá˜Ä È› P{8ÔUIÜû6J÷¡¼þÑÈ$œx4òÞë~'A©È–·”£ùI÷®)êKéßÅÙžþK°—®CF B³Lëåû«Ú)“&Šôšð†3®lܹ¯òáâ”$[¯oøw§s]À¨ýÙp¡ „ ÖGnYbu¤thN+Ü$g&Îõ Ìö—¡jÊÀ‰4 ƒ*ïj‡–‰“çÁÁÏ>àÌz;ÅÕ¿æï®9ÄMy÷®‡r Bä\{„ m¸ GIâ]Øp¤ó§®ÍKO&E÷= Ûp'§¶¬ê°F|a€¨ä7¡Pñ(H¾ÝVà~—ë а«Î]ÕƒrÉç) îZoSŠY©è”æa'8çÛï¼s\Ÿ#/Š 4œ~i Ox¢ëm'Ò•ì$ t™¸¢¼êø`Z6}ÕöfaÇ«†D1K<%œY¤ °ÁåÙ6‹íÓÏU1К¿X£«Y9€2 »D ´‚/à_¼%DûâPOcÆøq}Fž±T‚@,‘Y†2{Á³Ñ ÝöÕƒ/±ŸžÆë-`ù˜£UÐ^zæ£ e*<ú-ü¹sÀñG&‚¥LdÐGO«äb¡ôJ©Gˆñ×Ä…kÓ£´Àg…Du>‚ø:`vÄïxltá¼pµ†¼®” ^ûLˆÅ°F_Ý«Doÿq–{`Èí¡jÀÁŠÀm4z?u í™âÒ·v Ð@¸X ÁSÞ¯¡¥‹HD”îí 9Ã9†ž¤]ó")™Ö#tÝÏ1ÛcÖ””Jƒ6:—?ÕQº:%EaGGÐ* ‰;åCç~ §úX¯QèÚóš@É$µô`_7Ãl¸ÿ°ñ\£û™ÔêísåÇ¡¿}÷!îòb€Ú }4pÝUéŠá¥K©L2ÁèwxÃK¿»€òLÿ¿|éR¡ø‡cF‡sÂd(o¡¢„¸·ýK ÖJØy°Æ=®¡åõ`-4DN>†Åÿ%Jgö3¢´¹£Ì~cžσt,(QvŠ=‚:¸b Þ®¼{ ;IïÆÈe¯bÕ˜; Ú$višï‡¼ëbù2ϯ ŸU°‹BÑ„¼ãºbÂzœFµk Ð åSÕb™ÕXZ³] #Xµiˆ:ÚY*_³V¡´…Y‰"Hà æšJùºß éM‹—µe2iUì†8,Å1›¢ºkÞ#Ë ›âþ\¬ `Ïhæ>Ñ÷w#¸ù“+ÔÚ?UDkUþƒÐt©×Îï…ÓYÔô‡ü|Î×ð¼´¦; p TÞx~^!ðvcÄïPçËkEöu(»¾×›($!Kôq»UBrB| B0õ Äoƒ`‰’ÿ²Éç€æê~© Ôoƒ!=¹ÿ †PT+رV+žÕ öºV€•X+ØY­ÀÍô;e¯rð›%)šÅ«%gIšÝ+IµQ/mßêAY|U‰ÈDGL{É,Í ºËL¬W¸wO5ò*$¾$ ‰­ÊÆ-“ ês»ŽÙ HY­æoñd‚Y­£2g>‚H²mz<6Œp6÷˜ÆJ#¼ìÂÿèP@HbÍ¢o9‹Ô ©½ŽãSÝì‘8èøtÍ•U8RNÏöžé cù?!öT$¨D]§Éu]•+­#i£ÅaPú'å_mׯ·Žm¦¯ZÇR…^5”¤SV)Â`jÞàdO[o6o´aÜÞÑ xŠ0ömæd4KfM¹[î&¦Þû„5‹ÓºŠWÝ GjÖ4S\l.ê`jõÌS«Q3d°SK-Ô&×µäöÁ‹‰ì¬Q[åÑŸñ&—}Zâƒs¼àcüy9ßWááû±CH¬®òš<<¡Ì=Gg_žÀ !©îvÿ»¶÷†“cˆð¥?â´†ËCIpF9sFKú£O0x ë”Aâˆ,3õàПh•zì—º\¾v£C‚¡dwmIr{º'“½…ÜrúŒ1Cn%·˜æ×({TÆéÒ‡J[AÝ äU·Âg×Râ§E”Þ‡-H |~ºòý*F¡¡ŠüÈÖ]XMNôÖ—á^éžÕ1ð+ZÈÓvŬäïBùŠ'‰u+¾ðìêšVËê\a³W ÔGºb;ûØGÌ—ÞªðéÈžêâ˜SbKæêKÈ·Þ°ý'¬Á—ÒtÍ·>hÚíØ ð[*ü¸ÚÓp7‡h§Áì\s GÁßÌ}ÑqÓX²Ã˜DGr¾2ðkO bµ g Ž¹øôæçd€+¨¶û裿ç%MÕî„m‡?ÜtÞ÷îä@?T³¼]±HÕ„ˆÐö®®n…t&[©:ù‚S>b0È°wÍðyô¼XZFEjÀÚó¹vJHO<<Âÿ(8m›ôé=ñŸ~üÆe×÷"dí‚meëAö¾ì·@ rà?§?Åá2Ãw² 2¬fª³é³Y} Çr¾ß…S²‹,´ÝÂ{û®=Ñè?—¼Ä¯ÛE໪_Ø•ZùÙ–M{sÈ~Oç!†BM*ª" ‘0˜"A¢‘…‚Q$±¬ú/ädY endstream endobj 400 0 obj << /Length 1300 /Filter /FlateDecode >> stream xÚµWYoã6~ϯPƒ>ÈmÌðÔ´ºhl=\ìƒãZf¢²äHrŽ>ô·w¨¡lIQ¶Ý,A@r8šÎ|søÝâäü*AJÒˆGÁâ&P‰RÄ#4bÁb,C]×Ï[ÓTϳ¹ IXV¸j îš²±ž×æNÏ8 l¹÷,¶pk ŸlŒ'5%®•iöUá%áRïg«ÅÏçWiÔ3ˆ Ax$Š¶\S!×Àì„pÙ±dÖ JÜUå5¥<7x¼9˜ŽfÝÚWáÃŒ©ÐxSvºÒðV°V©Ìæ2‘áµÿÊó÷Á“Þîrsæl æ\2¢¨挑T)oGYÔ¶}7W4lî nnÊB<“$•Ò‡9FKÞƒ\&k20µ·ºÈ ^Ù×û½ÞTºÍ]ÇØbK¶‰á–Qb8ŽcbH:¬Êô¨§—g­~|ˆÛ>Ú·‚OÙ‚ÏÝŒÀçH=ðµœ^˜Æe> /ExtGState << /GS1 407 0 R /GS257 408 0 R >>>> /Length 8744 >> stream q Q q 49.00 258.21 123.52 87.29 re W n Q q 0.00 0.00 395.25 394.50 re W n /GS1 gs 0.000 0.000 0.000 RG 0.75 w [] 0 d 1 J 1 j 10.00 M 70.08 258.21 m 156.96 258.21 l S 70.08 258.21 m 70.08 252.23 l S 91.80 258.21 m 91.80 252.23 l S 113.52 258.21 m 113.52 252.23 l S 135.24 258.21 m 135.24 252.23 l S 156.96 258.21 m 156.96 252.23 l S BT /GS257 gs 0.000 0.000 0.000 rg /F2 1 Tf 10.00 0.00 0.00 10.00 60.21 236.69 Tm (-0.6) Tj /F2 1 Tf 10.00 0.00 0.00 10.00 103.65 236.69 Tm (-0.2) Tj /F2 1 Tf 10.00 0.00 0.00 10.00 150.01 236.69 Tm (0.2) Tj ET 49.00 270.41 m 49.00 339.93 l S 49.00 270.41 m 43.03 270.41 l S 49.00 284.31 m 43.03 284.31 l S 49.00 298.22 m 43.03 298.22 l S 49.00 312.12 m 43.03 312.12 l S 49.00 326.03 m 43.03 326.03 l S 49.00 339.93 m 43.03 339.93 l S BT /F2 1 Tf 0.00 10.00 -10.00 0.00 34.66 264.71 Tm (-2) Tj /F2 1 Tf 0.00 10.00 -10.00 0.00 34.66 295.44 Tm (0) Tj /F2 1 Tf 0.00 10.00 -10.00 0.00 34.66 309.34 Tm (1) Tj /F2 1 Tf 0.00 10.00 -10.00 0.00 34.66 323.25 Tm (2) Tj /F2 1 Tf 0.00 10.00 -10.00 0.00 34.66 337.15 Tm (3) Tj ET 49.00 258.21 m 172.53 258.21 l 172.53 345.50 l 49.00 345.50 l 49.00 258.21 l S Q q 0.00 197.25 197.62 197.25 re W n BT /GS257 gs 0.000 0.000 0.000 rg /F2 1 Tf 10.00 0.00 0.00 10.00 47.61 212.79 Tm (Mult\(Exp\(.\), orig:dest\).educ2) Tj /F2 1 Tf 0.00 10.00 -10.00 0.00 10.76 299.35 Tm (z) Tj ET Q q 49.00 258.21 123.52 87.29 re W n /GS1 gs 0.000 0.000 0.000 RG 0.75 w [] 0 d 1 J 1 j 10.00 M 107.60 298.22 m 113.93 298.22 l S 110.76 295.05 m 110.76 301.39 l S 53.58 261.44 m 57.15 263.38 l 60.73 265.37 l 64.30 267.41 l 67.87 269.51 l 71.45 271.65 l 75.02 273.84 l 78.60 276.08 l 82.17 278.37 l 85.75 280.70 l 89.32 283.08 l 92.89 285.51 l 96.47 287.97 l 100.04 290.48 l 103.62 293.02 l 107.19 295.60 l 110.76 298.22 l 114.34 300.86 l 117.91 303.54 l 121.49 306.23 l 125.06 308.96 l 128.64 311.70 l 132.21 314.46 l 135.78 317.23 l 139.36 320.01 l 142.93 322.80 l 146.51 325.60 l 150.08 328.39 l 153.65 331.18 l 157.23 333.97 l 160.80 336.75 l 164.38 339.51 l 167.95 342.26 l S Q q 246.63 258.21 123.52 87.29 re W n Q q 0.00 0.00 395.25 394.50 re W n /GS1 gs 0.000 0.000 0.000 RG 0.75 w [] 0 d 1 J 1 j 10.00 M 260.61 258.21 m 364.24 258.21 l S 260.61 258.21 m 260.61 252.23 l S 295.15 258.21 m 295.15 252.23 l S 329.69 258.21 m 329.69 252.23 l S 364.24 258.21 m 364.24 252.23 l S BT /GS257 gs 0.000 0.000 0.000 rg /F2 1 Tf 10.00 0.00 0.00 10.00 250.74 236.69 Tm (-1.5) Tj /F2 1 Tf 10.00 0.00 0.00 10.00 285.28 236.69 Tm (-1.0) Tj /F2 1 Tf 10.00 0.00 0.00 10.00 319.82 236.69 Tm (-0.5) Tj /F2 1 Tf 10.00 0.00 0.00 10.00 357.29 236.69 Tm (0.0) Tj ET 246.63 265.03 m 246.63 341.28 l S 246.63 265.03 m 240.65 265.03 l S 246.63 280.28 m 240.65 280.28 l S 246.63 295.53 m 240.65 295.53 l S 246.63 310.78 m 240.65 310.78 l S 246.63 326.03 m 240.65 326.03 l S 246.63 341.28 m 240.65 341.28 l S BT /F2 1 Tf 0.00 10.00 -10.00 0.00 232.29 259.33 Tm (-2) Tj /F2 1 Tf 0.00 10.00 -10.00 0.00 232.29 292.75 Tm (0) Tj /F2 1 Tf 0.00 10.00 -10.00 0.00 232.29 308.00 Tm (1) Tj /F2 1 Tf 0.00 10.00 -10.00 0.00 232.29 323.25 Tm (2) Tj /F2 1 Tf 0.00 10.00 -10.00 0.00 232.29 338.50 Tm (3) Tj ET 246.63 258.21 m 370.15 258.21 l 370.15 345.50 l 246.63 345.50 l 246.63 258.21 l S Q q 197.62 197.25 197.62 197.25 re W n BT /GS257 gs 0.000 0.000 0.000 rg /F2 1 Tf 10.00 0.00 0.00 10.00 245.24 212.79 Tm (Mult\(Exp\(.\), orig:dest\).educ3) Tj /F2 1 Tf 0.00 10.00 -10.00 0.00 208.38 299.35 Tm (z) Tj ET Q q 246.63 258.21 123.52 87.29 re W n /GS1 gs 0.000 0.000 0.000 RG 0.75 w [] 0 d 1 J 1 j 10.00 M 309.71 295.53 m 316.05 295.53 l S 312.88 292.36 m 312.88 298.70 l S 251.20 261.44 m 254.21 262.58 l 257.22 263.77 l 260.23 265.00 l 263.24 266.29 l 266.25 267.62 l 269.26 269.01 l 272.27 270.45 l 275.28 271.94 l 278.29 273.49 l 281.30 275.10 l 284.31 276.76 l 287.32 278.48 l 290.33 280.26 l 293.34 282.10 l 296.35 284.00 l 299.36 285.97 l 302.37 287.99 l 305.38 290.08 l 308.39 292.22 l 311.40 294.43 l 314.41 296.69 l 317.42 299.01 l 320.43 301.40 l 323.44 303.83 l 326.45 306.32 l 329.46 308.86 l 332.47 311.45 l 335.48 314.09 l 338.49 316.77 l 341.50 319.49 l 344.51 322.25 l 347.52 325.04 l 350.53 327.86 l 353.54 330.71 l 356.55 333.58 l 359.56 336.46 l 362.57 339.36 l 365.58 342.26 l S Q q 49.00 60.96 123.52 87.29 re W n Q q 0.00 0.00 395.25 394.50 re W n /GS1 gs 0.000 0.000 0.000 RG 0.75 w [] 0 d 1 J 1 j 10.00 M 67.42 60.96 m 164.72 60.96 l S 67.42 60.96 m 67.42 54.98 l S 86.88 60.96 m 86.88 54.98 l S 106.34 60.96 m 106.34 54.98 l S 125.80 60.96 m 125.80 54.98 l S 145.26 60.96 m 145.26 54.98 l S 164.72 60.96 m 164.72 54.98 l S BT /GS257 gs 0.000 0.000 0.000 rg /F2 1 Tf 10.00 0.00 0.00 10.00 57.55 39.44 Tm (-2.5) Tj /F2 1 Tf 10.00 0.00 0.00 10.00 96.47 39.44 Tm (-1.5) Tj /F2 1 Tf 10.00 0.00 0.00 10.00 135.39 39.44 Tm (-0.5) Tj ET 49.00 65.60 m 49.00 144.92 l S 49.00 65.60 m 43.03 65.60 l S 49.00 81.47 m 43.03 81.47 l S 49.00 97.33 m 43.03 97.33 l S 49.00 113.19 m 43.03 113.19 l S 49.00 129.06 m 43.03 129.06 l S 49.00 144.92 m 43.03 144.92 l S BT /F2 1 Tf 0.00 10.00 -10.00 0.00 34.66 59.90 Tm (-2) Tj /F2 1 Tf 0.00 10.00 -10.00 0.00 34.66 94.55 Tm (0) Tj /F2 1 Tf 0.00 10.00 -10.00 0.00 34.66 110.41 Tm (1) Tj /F2 1 Tf 0.00 10.00 -10.00 0.00 34.66 126.28 Tm (2) Tj /F2 1 Tf 0.00 10.00 -10.00 0.00 34.66 142.14 Tm (3) Tj ET 49.00 60.96 m 172.53 60.96 l 172.53 148.25 l 49.00 148.25 l 49.00 60.96 l S Q q 0.00 0.00 197.62 197.25 re W n BT /GS257 gs 0.000 0.000 0.000 rg /F2 1 Tf 10.00 0.00 0.00 10.00 47.61 15.54 Tm (Mult\(Exp\(.\), orig:dest\).educ4) Tj /F2 1 Tf 0.00 10.00 -10.00 0.00 10.76 102.10 Tm (z) Tj ET Q q 49.00 60.96 123.52 87.29 re W n /GS1 gs 0.000 0.000 0.000 RG 0.75 w [] 0 d 1 J 1 j 10.00 M 121.12 97.33 m 127.45 97.33 l S 124.28 94.16 m 124.28 100.50 l S 53.58 64.19 m 56.18 64.63 l 58.78 65.10 l 61.38 65.61 l 63.98 66.16 l 66.58 66.75 l 69.17 67.38 l 71.77 68.05 l 74.37 68.78 l 76.97 69.56 l 79.57 70.39 l 82.17 71.29 l 84.77 72.25 l 87.37 73.28 l 89.97 74.38 l 92.57 75.56 l 95.17 76.82 l 97.77 78.17 l 100.37 79.60 l 102.97 81.12 l 105.57 82.74 l 108.17 84.46 l 110.76 86.27 l 113.36 88.19 l 115.96 90.21 l 118.56 92.32 l 121.16 94.54 l 123.76 96.85 l 126.36 99.26 l 128.96 101.76 l 131.56 104.35 l 134.16 107.02 l 136.76 109.76 l 139.36 112.57 l 141.96 115.44 l 144.56 118.37 l 147.16 121.34 l 149.76 124.35 l 152.35 127.39 l 154.95 130.45 l 157.55 133.49 l 160.15 136.50 l 162.75 139.43 l 165.35 142.28 l 167.95 145.01 l S Q q 246.63 60.96 123.52 87.29 re W n Q q 0.00 0.00 395.25 394.50 re W n /GS1 gs 0.000 0.000 0.000 RG 0.75 w [] 0 d 1 J 1 j 10.00 M 256.84 60.96 m 362.65 60.96 l S 256.84 60.96 m 256.84 54.98 l S 283.29 60.96 m 283.29 54.98 l S 309.74 60.96 m 309.74 54.98 l S 336.20 60.96 m 336.20 54.98 l S 362.65 60.96 m 362.65 54.98 l S BT /GS257 gs 0.000 0.000 0.000 rg /F2 1 Tf 10.00 0.00 0.00 10.00 251.14 39.44 Tm (-8) Tj /F2 1 Tf 10.00 0.00 0.00 10.00 277.59 39.44 Tm (-6) Tj /F2 1 Tf 10.00 0.00 0.00 10.00 304.04 39.44 Tm (-4) Tj /F2 1 Tf 10.00 0.00 0.00 10.00 330.50 39.44 Tm (-2) Tj /F2 1 Tf 10.00 0.00 0.00 10.00 359.87 39.44 Tm (0) Tj ET 246.63 62.64 m 246.63 128.34 l S 246.63 62.64 m 240.65 62.64 l S 246.63 84.54 m 240.65 84.54 l S 246.63 106.44 m 240.65 106.44 l S 246.63 128.34 m 240.65 128.34 l S BT /F2 1 Tf 0.00 10.00 -10.00 0.00 232.29 56.94 Tm (-1) Tj /F2 1 Tf 0.00 10.00 -10.00 0.00 232.29 81.76 Tm (0) Tj /F2 1 Tf 0.00 10.00 -10.00 0.00 232.29 103.66 Tm (1) Tj /F2 1 Tf 0.00 10.00 -10.00 0.00 232.29 125.56 Tm (2) Tj ET 246.63 60.96 m 370.15 60.96 l 370.15 148.25 l 246.63 148.25 l 246.63 60.96 l S Q q 197.62 0.00 197.62 197.25 re W n BT /GS257 gs 0.000 0.000 0.000 rg /F2 1 Tf 10.00 0.00 0.00 10.00 245.24 15.54 Tm (Mult\(Exp\(.\), orig:dest\).educ5) Tj /F2 1 Tf 0.00 10.00 -10.00 0.00 208.38 102.10 Tm (z) Tj ET Q q 246.63 60.96 123.52 87.29 re W n /GS1 gs 0.000 0.000 0.000 RG 0.75 w [] 0 d 1 J 1 j 10.00 M 329.73 84.54 m 336.07 84.54 l S 332.90 81.37 m 332.90 87.71 l S 251.20 64.19 m 253.80 64.20 l 256.40 64.21 l 259.00 64.22 l 261.60 64.24 l 264.20 64.26 l 266.80 64.29 l 269.40 64.32 l 272.00 64.36 l 274.60 64.41 l 277.20 64.46 l 279.80 64.53 l 282.40 64.62 l 285.00 64.72 l 287.59 64.84 l 290.19 64.99 l 292.79 65.18 l 295.39 65.40 l 297.99 65.68 l 300.59 66.01 l 303.19 66.41 l 305.79 66.90 l 308.39 67.50 l 310.99 68.22 l 313.59 69.10 l 316.19 70.16 l 318.79 71.45 l 321.39 73.00 l 323.99 74.88 l 326.59 77.13 l 329.18 79.82 l 331.78 83.00 l 334.38 86.76 l 336.98 91.11 l 339.58 96.06 l 342.18 101.63 l 344.78 107.67 l 347.38 113.90 l 349.98 120.03 l 352.58 125.82 l 355.18 131.14 l 357.78 135.86 l 360.38 139.85 l 362.98 142.95 l 365.58 145.01 l S Q q 0.00 0.00 395.25 394.50 re W n BT /GS257 gs 0.000 0.000 0.000 rg /F3 1 Tf 11.00 0.00 0.00 11.00 56.26 361.03 Tm (Profile traces for the multipliers of the orig:dest association) Tj ET Q endstream endobj 415 0 obj << /Length 2673 /Filter /FlateDecode >> stream xÚ¥]oÛ8ò}EPàsEQ_,î{½è¡»wئOihIŽy•%¯$'ÍËýö›á õáÈInëQ£áp†ó=þûõ?½Ë‚ -t&×Û‹8‰Vi"EÈ‹ëââfõ×ËuÆ+Ó÷íç m³½¹‰¿|¹"ø+Ó=ì}Ó—¯à³¼ürýÏŸÞÉ ™P .Ö¡ZfDðF~¹\K µºþíÓ[¢òîçßÒÞLM¶ÊHè(šï¿ÞÙîrÂ~[o›vozÛÔ àÇ®,hµy`ªS1¥ E_Ž^ÞÔ[[÷â¶Þ/p"K¥Çí"Û–ý±õ‡ÖŸƒ ¬m_Ò{e÷¶gF€?ZÀ!ˆU”u^úm}ÙÞ]âÕVÝÁ o3¸æCkM_¾^à-³Td:†Ãx¦(ÇëŠÕT•weEË¿Ð#:>§1™)05¿÷PÄ´óOÀdOŽ ‚lñ6 Ûbfë—cå˜yûí€g2SMko_e‡ß¥(‹c./×Q _~ýy\ éà;h†ôaˆX§™$[_ d†Q”~÷ ŠO"ÊT’ÅÑpb¨’P¥ÙwŸà)‚R*'èXEa0;!ù'€&U’É÷õv žA’ešLDË â?ôN‰ØG,B6ÔpŠ&`§Ú{O³}FÇ.QLÉÔ㚺X$ 0QÆ9Øü뛦Ü.¸-\D w• QŠ„™|߃aä„ Õê÷£sZåÍ~Q×Îiqa­ë!Ät½ÍMEð}S”¼Ä€€Ï¹ ƒÕÝ¥ŒW%í2ô¡2í¥ÌV·|N}ÜoÊ–0š-Á¦5û.Ã(sµAZÇþ„›¦®f¤»ã¦+ËÓêweW.™ÛxFC¸qG¨÷v¿+}të}Û—¼ÕA·å¡ÐÖ·üeÇTÝ Ç‡áSf®wåRX¦ vÆXfêcăð÷eÜFY|L\ Ãþr¸kÑÑîMU5¡ï;&ëä A‚i‡“ܳ;”¹Ý>Ћ!,¯ ¹{IȹBù~góooËù&¯œ+Ž³§³Ó;í‘,Œ¿±l"X¨E¤Ó§…ñ‚”ÇævMÑÕ{[U´Ú6ù‘.íº#Ñè`O¤ÁA+ï>€Ñ¢`…ùNv‹’„Á^…Ëî¸ß›Ö:ƒ¢ãáäeGš8 «‰´ˆŸ0üHñ†“;…ׯ¿rÖ(£ä€G) ƒ‚BÒ¯šè7d²ê‘Z÷æ>lØtº²*s‰’hèôŒW@<“P@ðžn ^Χœ.Ý1Ï©Œ.ê¼ÛY,{*¦`EŒÂ¢nð0©‰cLªG"VÛ`¤»³…GY²¸|i°¦ë¯KÞ  _XS5·Dpëy%é©Âž¶m³÷œóÙá „ZBÞç¹Ð݈= ­SE«£ÓËÓ^c€ëä€sm¬ÞœÒ4û¼–^§aLCâÌäKãXÇgâ²ð@,z­¿·Î‹À½ƒÜutE€ë¹ƒÂØÙ+n¿½1mM<ÀE‰P*9cD¸G¯jxq I’mª 64&±²µ9½PÎË{—¦N(·uaó„BÊwpX>œ5¥ï.›Ú”‡Æ%AÛ…‹Wp›Ââã‹ŸÖ½¯nþöåÕRÅà·Ï.U ¼=p4õ<{NœÙ¥¿¶9Þ_p åK¢Ÿs (ç.!8ìC|V°ýQ*eËž6C£± æÛ‚R´öùGŸä>ò~ëè¹+«ÂÊøÃD‹$~¹í«± Ûšz0üf»p}q&TªžÏaceÐlþúé8oB>âšprÖ•¤öQd¤’¬¯ åÜÈÙ@Å{'OÁv)6DÐ… ìýúéÇ%) ËgÌxÎTsîs\ÛèYË47̈0¹dÄs‘áΞa…ŒO\Àä ÔÐPû†¶-•É=ôØ¡-Èd"ŠÔKíANP¯"ëB7ñçõRã-ÉZí€ykR&I¼bׂ¤]B:ÏgªT°¡1È<Ñ‚@Œ =ÚöXsSA2MŸÎâR ‹PÔØý¡*i}婢úîÚ 6›ÞX¦â<?²Ý1Æ–€yÃÌëdfYÐ'Ã%d²Àyùs¸æÜ‚½0qʾŽášÃ"‚}?äD‡ã[UÐ!F'ñÊõ`*ótlKÅÂÀÂÓ‚%Ñ¢µa<7YÁæã¶ó_g‰ÓAê‚;¿C·£z@E@+·¼2‘Éd´™ÛòÛaá~T$²$ZÊÚèq¿ÞoùLzÌÒ6NÓ¶“ŒË(lXORó耋­ÒÊàç—"´«iÓé3&£”ˆ’ìXŒË‚-ŸgÊ 'TW£Pnœ~˜ìÍ <*t}RÕÌ»Ï3¶Ÿâ¼"|^’h&‘t•ˆ7k½”œ«è_pz{o»òê)b-Ò1;ý¿—I•éÙòãΚ'J zTñóµG¸\{¤‚³üØ °AÆãP§G©)Z¹'*¤=CÁ Hf8•A=⪣‚ŸicúCè=ãíLáÑú%Ž2(^X}CÐÝÝs á”ø=¢8æ\^áë4{¨ìÐw(™99Ï!f£„ÇLóIîÈ4Î펵-ìvëfµ®‚…tçž>÷LüPþF|Av'#ù™à'çõ4í‘)M{àé’Ost5¼rK˜žŒ¹²)ýÞºÄ:Á´®L¨ ð8‘é¹¢1HD¬†áUd e‘Jžk–€x³=3ô(ã@ÜVåRd„^5¢œÓ±Û±œóûÑ Í.†“_$NrßÕ#ÇõÑpÒú‹Í\Õó3×|Wæ_ßv½Ý›Í¢äkÆŸJÏV>A¾˜pšˆ Ð+ 0Ö¾‘â<ý?6š§‹šè3hÇX%g+Û?„Æ—®dÇGeë’RF†ÅîÆÖf<ÔãÎfTJ¥«w¨6*8/ÖP¡ÙIp¯›z=VÉ”tGbÆU<•·^l‘%Vì)ˆÂÃòDP ¶1¾'9œ‰ jrSQŒ@Ь£VTâÓá¨ø9"ÐÉPyìþî›”Ên–§Ô³€¢)Ýò%f¡Â%³ÀžtV#„Ügá_‰­¶6úæuèn»¦íçŸLu[nZCc ÀZ\8+oí¦,8ÓE‘PÀøe¿}ƒõÊ¡©8ÿº*¿4° aTTŠÔ:uÿ°œ„¿ iÛ“ fLÃlÝÆW~=ç·cMõ’îzðúõñ@o,ƒB—K72Ûºwíæé¶á·´ÿM½ä÷8[H k‚ˆ+®tÿ-þðöú‡ÿÒ¡“‹ endstream endobj 317 0 obj << /Type /ObjStm /N 100 /First 872 /Length 2361 /Filter /FlateDecode >> stream xÚ½ZYo9~ׯàãäaÙ¼ŠÇ" ±'°3Ävv?´¥¶G;²:+µÙýõûÝŽÕ:¬–Ô`ƒìf5Y¬úê"eUJX…¶AXm… NØ„nÎ ë ºJ8å…5FaT%á#ZD4hñŸ¼¹ ´i´¤@‹6ñ8^hÑ…I˜Ü(¡^[“„&¯¡õŠ„·hSxt’Yk„ŽN¬:aNI& Z匄6z’0Úb3šwá”0&b¼5Î: ¥@˜Õc£šßa¢ÃæÀŸI2pyÇà–·¯ O.ï4Žx³X‚?0,ì2qä¹ÌÈ2óN{-“1;xA ´<•÷ÄO†ˆ/Y˜ ;å~ x Ñ9¥™Ö §™c‚Œ #‹»¼ÏŠ `“;bbï„óÌ &w!bÌå˜H˜/1)f"Xè‡u±ö2²#彨—Þ9V“‡2}~ƒNÀÂ6B­ z²‘uÆ{…È(%lkyÍì€ G"ðY˘ÝñçPŸO`úò;غOy(Š :˜+è0Zpɪf¨Y@)XHðˆS,3…df™†ð±Sø8Æ+ˆ#dâÀlHÐ4#8¤LDdÀ8(9ê< kÇ£“…§´ˆ”qoÐÉz16 µ8è ¦à0#+AƒY¥¸c q™³èYYü$khôòå¨øôß/•(^Íçu3*®V7M~þÇtþç¨x]/&ÕⳂ1êëâ]ñ¾¸ÀƒêzT|¬Æø ”I]‹’X.É„ %éÈ€î•xùRW¢x[ªEq)~O›J^Õ7UÔ/Ä?Žð7#䤲¿1¼„öd„Ö¹ÄlžEñÛ¿~ÇÒ§¬AlÕ‰ùj6»ÞKë[ZÝ›–¢‘èëо©çMæ÷ öd¬zøê @làzÚø«à# =>Dv1í\„¡Ç}|bÃÀ??aáââ_U¦(>\¾ŧêk#®» úPÞU£âüUóf ‡cùsVò^-ÆÕ2{Öüêçj2-_×_EV%’&–1A!ʦ`oö@žU¿ÄâìÊ™%¶/n¯O…ó±‘d±0 Z“ÌDEÉ^4Ï Ø ‘åêf‰§õ\Zy\7xÑJB³)ÈØ}ƒ£˜¢ .æÄmâÕ˜þxíÐÀëZr¦u?Z Zw¶×áÜí~pÃãwY‡¦,äó€·îϺQÃ\IÅÚ´I²ˆn$!ªRR«cžUŽ€U8VÛ´š¥5õ‚J?ggx$ìQÇwv÷¢S‘b·]¢=É%®IÌaàv—!Ô‹¹‰äÌ°­Q&¤¸½h ©‚Nû¼?šutºÏeœª+§·u•Î´ê–?ÎëZ߶¡mãVo¬rN™P’„¹€  ÷hÂ!«wÝ`r/ÚjIÈ7ÖVЦÎ<ÁyO¤óbä%:xÃ\ØHÍU|‘çò­v;YyßT÷Ò Ç‚QÀŠJ,ï\b†dÁJ ŒŠ(àkxŠ*!åØÏBØ Nõw@ÚǬÁIM©­A.u?ZÄ8®¯úÐ: ‚ õ£% ¸ø^´\k$Û_ÔæR‡~<È7Á²÷yÌN\wŸÝ¬h=>v£`7@®ÃûYÚŽ‰tn™@ƒfOÎX e´c0º…Wpeú  Ûɲi;^zÄVoè„x¹7«î–‚ Oþ¼L‡Â¶Vé(­RÚÌtÑyHlO¢XßÉs¹ÃšÈNÝ°O[öá” ¯ÁØ·é‚wmKƒÂÎÉQ¢Ö„8\´Ò»”!í†õíônµ¨-wlŠKM”ˆH|'µáSÊ$UÜc\­e  „R e·!ìU?_vïâ‚aêcëîжq@¨ƒN€t×ð÷?C;ðà¶Ì™¦l®Ój^‚ô‘nFkÏmxÅß~½ù7c‡'|ÏÝíïïM¾‡h8«³Ï TéÅAAo»˜p¶‹i+à­<¢—†ÏÅ¡2¾ŸP žÆ‹`‘¶ô‰˜~@ì›æÆÏý¥G½åµ–H¹vòò,WuSþ ¶#Œ{–Ï;P>j“$_1œ—¸mø€ÎáÖ~Oµçd¶`–Ô™0‹æÀW,mÑb[dž±-„ã™…ðÆG:çøšGò}óHÊXy› äýWïÊæÊ wáƒDû#¨È9ú}F6c¿Á‚ˆýex¾†±ßxøÍô\𰠶Ƞ Y—jJóu®$¾ïs2F÷øò­ °L.‡5é‡r8Ñw’„ (‚á•Pʨ±zŒ’`,ŽÔG:ä1Í ¹N2Ú[¡“1Á›“Áö{ÀtÄeYÚ¾óJ#ý²½hQ IçÐ"½K‰CµBëÒTy†»HÇÖt7[Ï8:u÷šÏ?Ñ—;¥¶}y<×—oܽ],ª’QxY6•øáòïF)¯’1Ú(oì ¬TOöŽ}š63Œ|oå—?¦ã¥øuÕ|Y5/ò~'«qµàQ“ï Úµj~õ¢»qÖÈÖù¥¿”÷<†ÞërYe­¿—_n/§ó»›²Y1‡YŸã]5û«j¦ãrTü4×L—oåײØSÚSþíu=›š÷§¯ÍÛ«B…H^ ½wp\n ¶ó>¬ým•âŸÓù«ùrú4|9½½­€ÍgᨸŸÎWK‘|ñŸR£YuÛí\1©¡ÁårZÜ-Ê¿ª¢¯šªOãÕýí¬úZ4ÓÙ¤*îËñ¢ž7‹ 4ø¤¯b2ÅËé²@פº-X»š³Yùíå«ù]¹XÝÏÊUSÔwõ¼ú;ãù–_Êq5XÒ ¯€¼ýé¸EžÇå¡•²;IG¸¿5Z"8òÏo-zoze‹ûÎ3.üS‘p–y<~{ò1NÛc|ŒÓaÃÇðoU˜%×þ*‚©2`>è¡`›=­Ô’T°2$@ÐÝŸ…]”Íøçzq—Âp)¡öе¡'f¢Ë1ï;1³‘Y¤! ‹#-ÔùLiˆ‹9PþpÇÆ=L KÛFude&¬Eõÿ^:¡å endstream endobj 422 0 obj << /Length 1268 /Filter /FlateDecode >> stream xÚÕXMsÛ6½ûWhÚ 5AÒÓv&ÍØÓÉ4=$ÊIö– SŠtIÊŠûë»À‚)1®-Ë™æ`X,Þ>ì.>V¿MÏÞ\ÆÁ(!‰är4]ŽBIdŒ"É•l4]ŒfÞt¥Æ>§ÂSãëéû7—‰ìLàŒ$"QT½¢,@¥>*I¢VGÍë ñŠ%~ëÆÀ¥¼¬jìÌ‹¼ªu½Q¹¬7Y­ï2­Jìëü±éö<­õ˜‡Þý˜…žjfÖªLçµ.DZ6# 0§ Z d»R};¡w—–éZ„YêÈçI`>W„!.rY’qâ¹r%~%Ž½Líhe^qa«J°Ñ"¹‰Ig™Ðé¡ ðD;XëÊ2Õ¹]”Á*ðû* 2ö—.Ê kWWõuoܸ‚¨¬ÓºB×!¿Ü:„ö]áx‡ÞBå'‘”ÏÊŸRåsU!檷Ê:*¯LKz-Qxà ÐKó6ZjÎ¥Æ̺¢zœs@"ÎŽ&=Oó†=Ò-Õ] &sçTxpLÝ„"‡ðaÈCœ_ê/.÷˜8K4–;!…Ôɲ¬~[;‚´CÐÄ zÜÅé×±rcK-+lþìã7‡|¨®hH͘ù.ŠÍM¦>çz¡—FÀÌßPì[LàÿîØÚšÝÂò ê”–*Ó·°5Ïï‹Z5B«jÌ\?b'g õmd*¿­WȾañºÐË=,ô¼Á¢“Ct<ér™ô̸øü4ö§‘É®µõ-ÎøÅAêªn'û|ª·÷ "RmÖÏòÃø©¶º6ËƦؘÍcšæ\h±mršè:Õ¥.a¦îaï=‹qy£óEËvÖxÓŽ]w½2óûcOílç7™Ÿ#Öõ€÷^Žiw £Ý“„%1‰BûPYŒ¨àæ2+¿øb·é$Uok9Q™²÷Çó ìÒ< A’Æ ¥žÏ†røH’@P€Ê`„Ž‚ Hz2Hч> stream xÚ­koܸñ»ÅâP »8//‘TrW ÜMÑÆîõÃå€Ê»\¯íÊ'ií8úÛ;|I”L¿mÀH g†Ãysßžýð`1ËQ.¨˜nf”æˆæÙL ‚° ³Óõì·ù‡7;y¿X©2ì¿Ÿþu Lá… <ËðÇCÕ}Æ~ÿõÂädæÀ¨9°a'T!œûËwÀŠÆÀT •ç3ì€Îu÷®ÞwMÑvm µÌâ=ôq a†r©D«=‹jI°BÞ—D!¡äéBáy½XRJçå~ÁñürA²¹n»ò|‘Í‹N»om¹»¨üûgJY{عAçÓM ¸±ùªß[}XmÝ[ágºmÝztÅYmH:ºÇfNÍËγäÁuÑ–ÀÓ˜ä!`è¶a¿8Ú/ã 1Én—ñH:"C”’¼9ì­Î;å„¡,çc­¸Ú–vW ³ÕV¯¾V9±ÌØI#Ã]qVVewíf깨۶<«ül»**½öˆ"¹Á°Ø¯ÝŠFw‡fïg{ ESìt§ävÚƒ]•Ý¶_Pz˜¶œE³¶ûÂã 馩›-–LåóO–b¹?‡uR:™KáiÄ6ÇõµðZ_ìæ äÖO|Ƙ6öÌ`6Üu½ïÜ×iyQ•–yåˆ–Ы¢+4(KX (ìqyºÙ=RnäÎ3 G/Ã.a°*ö™{Ög]Qîµ_TxR›ºªj#…«öUBï— ¬’*kæ˜JÇë_œãÛ]6`'~\ºglÆi®ëÃY¥ÿµ/×åfs<’¡wÑ߃CÌ(,½(W_ÞÕz“XfQçŸÁÝZoûñÀæ?u< fí6'C€$,ÍQƃ{X,9ÅxnÜ6%ðòlj°uK@­¤"ŠJ„0¡í gSEò”˜ŸD˜õ„©€ó¶Hˆ7É{ÂBˆ/G˜÷„UŒgyOü6ÃÃŽ%Šæئ„#Î''muá×¢ñÊyª2„ÃÏ(}¹Í»ÃŒ‚ÕþbšÄÔ\p,~é#3¨3ÂwÚ0q©xlÀ¿.†°(ªò[á<'cÌFó ö½vÃ(^šá¡µ‹1>ÿ=e7Ý=‡Â†¹?»Ù̇²é¶Çn Í^»W;í^­ß6/kížõ^Óí°ˆ[_f)—mJ€J=„.ïÒÏ®Ý󢪻ÎÛ8:4º5/Ì×gÿÁ¦|>öÞ^¨Áj69ÒãoÖ•Kp¸…‰ÂfåOÞuÒÂ¥vöƒˆwn„bFæHâó_šówSEÛÖ«²p”©êsÃd;ßÝƆøZgc.Þ{} ‘¤Ò—ºú;¤>펼â!¢$t, †omæ%à47 qd|œ*(Sòáù#‡x`HM[GÃ'kYéav”¹ùÜ `<°˜·zUš.1±i¿²˜âmôÆ%"n¾Š”¬zmæ伨€Ò~”:y,»âÚa9Ó¾#a>©ÆqPÀ˲44!0Š”êË`,!'°u>œ0QóóÃòBäsm‰8›`¿/à}û2ªÌ‰Rc–!OfQñbÞca5a¹ËÖ˜IžëCç׃΂"§f™•n:k0fP±Æ£¼ð‰*¤¼ž“Míƒ ÄšœOêˆq& âh\r·\ÕÕag2B’Çå€võZW­ûhl¼Õ~õIŸp‰ˆ÷Ô&Ss:0q™a7~¾\Ã9 zŸê½-Sà[±qêjéaúÕy%øª ø`¥lûÓ›]Ýx×v´)*ŸÑZ3Ü••w‡uÐkfH º/›njTßX¼ÕÌ„K¦Ý£Wmk‚YTS…œ“e1ÔWfQ•õœ ¨ ˜rUö8ÔŒÎRÈ”§PæP¯»æˆ {ne‹@JÒðSB9É’b”Rõ(Àg!ç¯ÌC¸‡ ƒCÝÔKäS‚…²¡5Qúƒ8œÑ¾† æÀfL÷C@¦| 1•<KžæÄÈŒd³Ä AIØl ^fw‰V¨E\Œ„;f"çŠGñubL‘ªQ}™©l}6EÀ½?=úãÈ(ž *¼Œ`0 5[íŽ~ûÏÖð,e0ueAwV"ðRÍNŽþyôÖ¶'ü{T˜#È¢ïÙàZ‚@°on]%XË$b€5GŠ±»9d2a ì6b-Sðn9”ÂËã"AQÀ2Cœë]4m]r'ÍY6(Ñ?²€tj`’Ñqªñqb BMâ1]&0AùAyJ3FF48‹^32|‹W!êê¥t‘A1-¹‘G‚fÏÑÅ•.òÇéâ!)´¶òÙ‘™8%=QšêŸRÈñ„É(A,8¢+ˆûú¦T TpôR€£y–L`/RÝoŸä¦Þèsåv³=Pðk#CpQ‚ —S#ˆ@'—ÐU‚+𢶼~@4¤q4¼WwùíÖ’!IGÈlØa!m†—8mf“´ÙLAcàåÞæìú½1–8»€¡K¤ @ñŤ4Ú5'#1œ@öºÑ£z’r’!Ofôå”c¬‡¤÷ËÕSr¥Ç¦^7Ôƒ¤Ô£p¨A£d·8ƒð¥øÚÅ*nw¾Æ¬ÉIÏéJ—ç[ÛÙ 2em›ÁÒg¦2tä(Euj÷<óß¡øôH +; §Ê.C^,]^,C^,òžº½ôqu›èUßÜú˜Cf¶OrSÄRÚgˆ÷„À;NMØ{Ðé© _±¿³W1£Ñ5ºèÌ·û®Äãå'ïï¥ ¤ðé±’šü<`¿3¸WõÁü~ÃÌüÏ=züåM¿ëïÇ¿Å×’Nß©`6¾]î\ÔeÛö×Ië¢+FŸã]ßr D deN’o¥üÉQÜ×ÿ…½ßÌ endstream endobj 439 0 obj << /Length 838 /Filter /FlateDecode >> stream xÚµVMoÔ0½ï¯ˆzʊƵØI*@JE‘z ì­íÁ$Þ]KŽ³$ÞB9ðÛÇΪi*‡nö̼ùðøe6$Âðˆd11LP–¦QÕ,¾/§fàÁv0‘à' ‹ÎÚÅxFSB&b¾_-N΋4*QÉ)Vëˆ0Š(£QÆK§ÑªŽ®ãsµÙwr™P†cvºLR‚ã+©…UKÊâ»%aq°ö¶“fc·^j×~í¤VÕšänIAkXô}[)ˆÒšc¯Ñí&Ä©„–ËÛÕç“s‚ùƒú’’A]e”¤áŒùú.Œ²JhÕ+³qNpÞ„T²`¿ÚãL ƒr{+:›ìw^RVvC=BW>ãÚežº ¿¹´g­ g˜´PP¤ð¨Õvh¶|x\JQÆsp 7˜¤3*óbÄÈÊö>žï>혠‚r­²{i¬Säq³×Ví´’Ýß\FDõä®y¬ŒkBåºàÂs$U )Ô ÆTÖÞôc+è[ëÙ1$QßúëöAL=ô2¡Êà‚'Õm%B¾|Ì—‡ƒup9²FÀÎ"ßyC¯š K`­¸÷‚mCˆj«ää`9ƒÂTïÍã:Â{iƒb¢ºCv}ÐîD' ] q‚×/ÙéLýÈ_‹§Î3TB¹7ÊÈÓÀ<¥‰ô[OÒ«¤ik©©_'~ݘæ3\µ{ǧùí—¯¿úÍ+¿\~z$œ'9>àH0€~¦tp.3(ýpIÞå_Ž®Ñ-Ĺ~w~{‚€«v­\n@äi>>Í×®/Aeoçý³Öµh”¾ŸøíZãÊ„ôµ°bbn€ñB’BÛ­+mnv=zñ_bj±ÿ›ZÏ]øï'<ë÷M#º{w=Ê=³„Ö§c4ƒ×Ž®Û&дÓ/ÂÛAzc(†÷ŒL»õ"Ôöÿ"â̵=‹–Ðèp†¡á¡ãgòN SÉ0:d¯ê½Ð½¿Ê|Ls]ºãfã˜|áD`s)kˆ´)hScðRü ig¢$9MÉ2!Ð'¦¸,R¸b„ ¦ü Ѳ(‹ƒTäç3ÿ‡f!ç|#øN ß0”:äâãjñPfe endstream endobj 428 0 obj << /Type /XObject /Subtype /Form /FormType 1 /PTEX.FileName (./gnmOverview-qvplot.pdf) /PTEX.PageNumber 1 /PTEX.InfoDict 442 0 R /BBox [0 0 432 432] /Resources << /ProcSet [ /PDF /Text ] /Font << /F1 443 0 R/F2 444 0 R/F3 445 0 R>> /ExtGState << >>/ColorSpace << /sRGB 446 0 R >>>> /Length 584 /Filter /FlateDecode >> stream xœ¥UMoâ@ ½çWøÒâz>2×VÝ•v¥–H{¨z@lÊ¢’¨û÷ד I@ˆ‚z ÄóæÇÏŽ-`6ðž<Á;¤IƒU¨5(-ÑJÞc*aŸÃoØ…MÉ]ýüã !Áø¹xøÅ«6…ÉË+üIÌù·IDØ?!jŠ.¶ìÄf ‹S<šÆ Œ¸çóÕ€wæ€KËŽlGsÀöþü>‹ñ­ê3ñÕ«]r÷]²bÙ‰è-È!Ð6¨–ma"¦m’Ǭ=ú®³á6W^Ï•+1pÕõ\¥ÍñõÛ„Pµa¹:SiB![¹Nq®7\’@ãz<š#œ³äiÀ;sÀ£Ã>ò?ŽmÐl§h ¯š.¸¡>§Íe¾óÁuäŸÍËE¾4SÓòyõfºRüç"½÷¾Æk VÅäë“ä·¶NCí+Ï¥—/“ç¼\6ÅGu³Ïwëæ/ToÜÊb]T»ÙÇÁ¤jrXÖuµ*xoµûeµfwË2Ÿ¾B6¿²hɇÂ.tšÜc™¯Âq7įžõ%ÂTµgäuSl—M>VéK=kÜ·‰¦6#±eLeÛk”‡Š=6›×uÙ/ݶ•LDÕ!Û'ÖñNê^Bb¿Ò8ä4½Ê ÃÙ¾Wü[”µ •E/n§KcQpÎ,Ÿâo§‡FÅ—©FcÏÐ㸤‚ÂýøPÊ…æGhA¤†"­GïÆD(¾‹~„HÉÝñh†²hFCD’A×ñ”üòæ†ý endstream endobj 448 0 obj << /Alternate /DeviceRGB /N 3 /Length 2596 /Filter /FlateDecode >> stream xœ–wTSهϽ7½P’Š”ÐkhRH ½H‘.*1 JÀ"6DTpDQ‘¦2(à€£C‘±"Š…Q±ëDÔqp–Id­ß¼yïÍ›ß÷~kŸ½ÏÝgï}ÖºüƒÂLX € ¡Xáçň‹g` ðlàp³³BøF™|ØŒl™ø½º ùû*Ó?ŒÁÿŸ”¹Y"1P˜ŒçòøÙ\É8=Wœ%·Oɘ¶4MÎ0JÎ"Y‚2V“sò,[|ö™e9ó2„<ËsÎâeðäÜ'ã9¾Œ‘`çø¹2¾&cƒtI†@Æoä±|N6(’Ü.æsSdl-c’(2‚-ãyàHÉ_ðÒ/XÌÏËÅÎÌZ.$§ˆ&\S†“‹áÏÏMç‹ÅÌ07#â1Ø™YárfÏüYym²";Ø8980m-m¾(Ô]ü›’÷v–^„îDøÃöW~™ °¦eµÙú‡mi]ëP»ý‡Í`/Š²¾u}qº|^RÄâ,g+«ÜÜ\KŸk)/èïúŸC_|ÏR¾Ýïåaxó“8’t1C^7nfz¦DÄÈÎâpù 柇øþuü$¾ˆ/”ED˦L L–µ[Ȉ™B†@øŸšøÃþ¤Ù¹–‰ÚøЖX¥!@~(* {d+Ðï} ÆGùÍ‹Ñ™˜ûÏ‚þ}W¸LþÈ$ŽcGD2¸QÎìšüZ4 E@ê@èÀ¶À¸àA(ˆq`1à‚D €µ ”‚­`'¨u 4ƒ6ptcà48.Ë`ÜR0ž€)ð Ì@„…ÈR‡t CȲ…XäCP”%CBH@ë R¨ª†ê¡fè[è(tº C· Qhúz#0 ¦ÁZ°l³`O8Ž„ÁÉð28.‚·À•p|î„O×àX ?§€:¢‹0ÂFB‘x$ !«¤i@Ú¤¹ŠH‘§È[EE1PL” Ê…⢖¡V¡6£ªQP¨>ÔUÔ(j õMFk¢ÍÑÎèt,:‹.FW ›Ðè³èô8úƒ¡cŒ1ŽL&³³³ÓŽ9…ÆŒa¦±X¬:ÖëŠ År°bl1¶ {{{;Ž}ƒ#âtp¶8_\¡8áú"ãEy‹.,ÖXœ¾øøÅ%œ%Gщ1‰-‰ï9¡œÎôÒ€¥µK§¸lî.îžoo’ïÊ/çO$¹&•'=JvMÞž<™âžR‘òTÀT ž§ú§Ö¥¾N MÛŸö)=&½=—‘˜qTH¦ û2µ3ó2‡³Ì³Š³¤Ëœ—í\6% 5eCÙ‹²»Å4ÙÏÔ€ÄD²^2šã–S“ó&7:÷Hžrž0o`¹ÙòMË'ò}ó¿^ZÁ]Ñ[ [°¶`t¥çÊúUЪ¥«zWë¯.Z=¾Æo͵„µik(´.,/|¹.f]O‘VÑš¢±õ~ë[‹ŠEÅ76¸l¨ÛˆÚ(Ø8¸iMKx%K­K+Jßoæn¾ø•ÍW•_}Ú’´e°Ì¡lÏVÌVáÖëÛÜ·(W.Ï/Û²½scGÉŽ—;—ì¼PaWQ·‹°K²KZ\Ù]ePµµê}uJõHWM{­fí¦Ú×»y»¯ìñØÓV§UWZ÷n¯`ïÍz¿úΣ†Š}˜}9û6F7öÍúº¹I£©´éÃ~á~éˆ}ÍŽÍÍ-š-e­p«¤uò`ÂÁËßxÓÝÆl«o§·—‡$‡›øíõÃA‡{°Ž´}gø]mµ£¤ê\Þ9Õ•Ò%íŽë>x´·Ç¥§ã{Ëï÷Ó=Vs\åx٠‰¢ŸN柜>•uêééäÓc½Kz=s­/¼oðlÐÙóç|Ïé÷ì?yÞõü± ÎŽ^d]ìºäp©sÀ~ ãû:;‡‡º/;]îž7|âŠû•ÓW½¯ž»píÒÈü‘áëQ×oÞH¸!½É»ùèVú­ç·snÏÜYs}·äžÒ½Šûš÷~4ý±]ê =>ê=:ð`Áƒ;cܱ'?eÿô~¼è!ùaÅ„ÎDó#ÛGÇ&}'/?^øxüIÖ“™§Å?+ÿ\ûÌäÙw¿xü20;5þ\ôüÓ¯›_¨¿ØÿÒîeïtØôýW¯f^—¼Qsà-ëmÿ»˜w3¹ï±ï+?˜~èùôñOŸ~÷„óû endstream endobj 455 0 obj << /Length 2527 /Filter /FlateDecode >> stream xÚYYÜ6~ϯø©q+"uØŽì.`×3X?ÄyPKìiÂj©W‡ÇûÛ÷+©£[=CªX¬*ëdÿrûÃOo…_e^ËøêvÅ^œWI˜y´p[^ý¾yÕ¨ý^ZÕ}w½d´iö<êºW­êú×ÜþóJˆÈóãôj+„—Eo~Óõú˜÷Š7Üô¥Ç³7mÛ´<ý“‡Oy5X´µüÈÿù¯?ÿÂ( ñí’ºíסê U<çžýû íônÞܼ¼Þ `l|Ï7×Û0Æo€p3‰!FþRþ'pø…![â?Ãô|O¤"ŠËŸ–>¤ðI˜,¿“Ý«ñ@2’YMì¢4“Ò~ /ÊäÈMñÜ^Ü‚4MÒhâ†"í§„ÕLgó¡ƒïãöfä–È0³³%2ŽyèI?à½`¨­}'¿·–ŸxºqÈÀË„`DýµîNªítS3õSÞæG¿àϽ³òS£»Î!íó£®xÞç•÷ ;ëb4ǘù®º‘¶ÎI2Óx€oò´˜<˜‡¼sœ»¾Íu­Jëägt‡Z—  ÷—2½S.‡¼bäR}Òy]¨üxQâÇ–®e˜:Ì»V©³À²¨lŽ‹[µ.ÿò¯,Q‘d^®ˆòÛpÜ9ý¡ 7’÷¸¥Îí–´ó§·i0 D¼Dj 5¸DæoúCÞó,çÁ¨+¯-ðSÓ§JWþîìÊÎÐuÑ´§2~B×ÍÄG$’²è1™ÕaµÒ—í)ÏNG|ǶÀÉ®)„^‹h£¦­ö„þì„!é2‹pÒ…ÀžsœuKWb—YÑÔ.Gm§Šæˆ»¼d$ÞÉ9W癆ûÜ7W¥)îe¤@šPJM£€­–¤z‘Ri<”ûÜ!¨ÿVCø—W ¦Éh÷º?Øåˆé;]“%:M©ìŽ^è¾/ cIpÁñîû8CQÒáRs£[Qî”Q8!; EˆK…e‰—dãqß½FŽ5›×$vˆÈÆÐ-6gÇÈÃDq.6Ö%#ö #RXi /Ä휴V=˜Uj²°jE`‰SÇQæy¹v…Ö#˜_¾—¡À,V¨ˆØ“a87„ÙYïº8LÇ6÷d¡‡šfq-td]ßQ˜÷7tŸÍÁ:BÀˆîjÝ¥b8)l¶îÚbìêïM<¥¥¯ê+€çÈñâÞ®ê+MFßÊëòi„ƒÌ‹‚ø±{€™f©C !²;{$x ™>Ç- Xz˜ cégéM=ÇnRÔ¥`[™!¯g7q§úW£Í]ÊHÍéⶡÿ»ÆãñålÔwWéŒD‚ëOG»jvIÒg»K¦QæÅS¬!óè™­5³ áøκæ«Ò“ÓS¤³'bû¥èŽgEÞ)»f!¹÷MU5¤ùûîŪš#¤Â(š”6mþÌñu®m:Œ“Նߓ.>R¹¿²ôìwï8ä3+4Ze“ëß,Ê^·]ÿìùZ1öãõ6 ¼‰5ï¾´Ñ|”÷Jß\¿aÑ «U'|‹Íä1åÒ”>–ÍØ?ÌüãÝÉJßqVZ>¡ß0UåòX±ú„þB¦ÙH‘[ ùÏ[XÌD‘º‘~#Åó6yp¢HB¬sPHt'è%q…ðÑC“,¥š¡¢¿Bvp,ÓtÆ+¬UUúÔŒ Ò\¡Œ7o‡Öæ)l2Wù9?ž*eiƒ ¥»Yù^óšÉXºQ¦°;OØ­õê&^èñc ’^xV‹¾%úl!êZ66Ê 5ºŽŠÁ¸[•·¼óÙ¡¶ìiÕ@ɬɵp´GµÛ3)7ôc[*aNYP†ô7Û¯1!äO¤Ë”Z+C/ÈÆdºGbçÒ–¸‘cØZ¤Ò-Q BžÙC^qùØ µ?5çô½{X.Aa6Õ$_+`"ONÕ–‰–°§’eÑ¥8ÚXwÃñ˜·+ÄQë…£’p!‡¦ä(@¤M¸0¼Lc“‘·Zî¥naV8dmLÀÎÖA„Š|dqP•É¾úB—†kÁX³šˆcø)´bUçñœŒõ/M½3ñÒ—"S/rq6ò"K^Α%P³lÖíèòæ?¯WÙZÌ­@×aiÞ 9E¾7Þ&M>¦\_áêÙÏM‰µæDÏŠ¦o¢E®°ÚÚ7KÀôSs^u#±ª-9Û§Ž’MÑ*Ë’ÔÐØñ¬™'$¶­Œ&þk©uLQäÏé+`ß}.ì16;îi~×4%ÏÀÚž›£P¸0y.ômÁ$/KM´MäÀ÷"&„\Ò<¦cã ªÊw k•>š¨îJ}Ûa„”]„ îè/Vb×èF °‘Çù¶l´ŒlÂ5-‚‰äÜィʷÀ¾ë‘qøfkãÁe#!án±?:Ã}Ó~äLˆ}íulFs¸´ÇD˜Àc·rŒ¦‹(·<Õíj@‰—eáìžâfš^E¡¸Fk–ºÒ=´ß1°©ONtëÅP妭’“q™Õ÷üÐi»‰ŸÃhæ ;W–ÍèG˜ßsrÔuiX¿Õ˜ Ç«-«!ILŸ²PÃê»çÅÙÆ®›>ì{Ss™<éFÊõþè“€83¡0º4µ)+6LÍÛÇ^gSë&‚µÆ×9öˆ°ÏׯÒO£§\eäñˆˆúƒ´ð :œ Çr+ùØÏÐwáÈzŒ";W°±øX,ü–¾K`ìó’ËÉ#€ÌTO¥Sr#ª¥ÈÅw‹qG~‚Ø' Bª²¨3W$&ØbM.žÅþ,é9ñÔtfP©o,Q\üJ4=øžL`ášÎ-W*#÷¨l¬8c¹1êªèÙn<êd¿Œ`¹Áúò¶r5:}—yO÷#P¶Í‘aï¯ÚT‰Í;*~Pñ3ÜøÜç;ón`»øŸíصk¡t¿å!Yõ¶3:zøÒ2ØÄ1D cójaØIã\?¼¹ýáÿ$Øl; endstream endobj 465 0 obj << /Length 2993 /Filter /FlateDecode >> stream xÚ­YK“ä¶ ¾ûWÌ!‡žò´,’%9¯²coœT\©Š·’TÅ9°[ìiÕêÑÖcg'‡üö$Q=ìÛq_HQà‡Gùö“ÏÞˆXßQ¡¥¾{{ºË²H‹ä.Ó"Šµ¸{[Þýk÷Þô•ŸïÿýöÏw{G…w{!¢"MiÃPö~¯Dï’GÅcÂcÊ£æ1ã1ç±àñŸŽ›H#]\1ûâ~ŸÊtGq§Þ<Ï…7—iøÛtÝ#”7O·s”(fa4 óå n0òˆÇYx]yt2_€u]Èl+ kæëfï¤þ\®…Öqh}3÷%½¹ò“è„ù*LpC\O-Ã{`yYO½yìÍ2 Ì×ëæ°JcíÍ… “{&ОŠ4¨¥D¥AaÞHþ\ÄáõÔ[ÏÒ°fþ$îßšôuâ¾þºü±Âðmú槜ΟûŒ²1½5hæO‹¾·ûÎæùFzãvÞ~-Ãë/æ Ìgorå¡ò~B Šœlï÷2ËwßÙãXuî!ÛeQJ«†žÕ¥îFZjºÒÖ´\ ´té»K7Ø’VO]OËãÙ–ÖJ3š˜æéîé\Ï´zìšK_ vð˜ÁgO÷"Þu{|Ûµ¶™Y;ÚÞ,B滃Ÿ¬mCZÎ2½;öÝ0쵆êô\µ´|B@«ë‡è~¯óÜ©ßt}i{šŽŽéî{€ü‘—Îxdœ9-À‰’8ß=ÍÌ }-ªÏ™qÁÐ04¦®gŽ>Q8S›@‰dÌØ3æÕ‰{;Tåwÿ*`ö$‰ŠLä8¸Ð®;Œ¦ji>Œ¦'Éðéý=8‘©'gx&{ÂÄi'ӛƂ9xÇLgÙp¨êªµfþÎöÍç#ìY²½€5Éù9±m.ãó·äjøü›=mó=ÜÙgzú/ {ñ@““iªšßý–†'[‚ESßòtÅÍŽƒékû TEÈ•Xºß¡¢¶ÍŠG©V¡|Ò¿ÂÌf»ÂÉÐ,ÒFùBåÇë†ÆOi¨Z0v{´îüvªG=ùX0áÁç!¤šOÿo¥óÛaU+¼±ó Eß³:®“Ìx d›ÚÖù0’iœo»#j h ÑÏý…Îý¿0`J˜¸åñKXWi2#VªË? ª ‘°áÅmèN´p™z@ó Qàý½€«HËè(†iuÆ{Ç¢£Å…bi _Á»˜YoÚ²kœ!ÝŽpðCpðíÙ”ô¢®Æ±f¢|æÂ÷€ô¢c1CÛ÷w4“ÈäóˆrŸƒcˆÌ‰ëj‘D@¸ËŠ­–éUŽÁ¦0ùîÑÒc;5Ëaõ†ãêi¼BÅĜ҅Ä9fZTíÓ\j&9¿8ši˜?;g[_Ì7X *±âòF÷Y% 5Öý¥O`|{•$,w³‘ ` …§ñ(˜uÜÊ>ÎS[ö¶Ä'B÷ZE©¸j¼=‡N-òH?ÒÅ`aÝ’£7 ’F¾ën… ŸbÌñTï!ÒíªÒeY°jh+˜¼´45Ø —аÿ Én&Í{ÈÀ0q \ßÀõoÈ*éÕœ]ªh³¯ƒ±2‘mõâÒX—ű>̲7ª~S=N³=5§vPXˆ$R‰–ì)åÍÕ`v6ˆK|†Cƒû£€¦Ká\0ÎÀ¸ÇØw|ã ’ø™¼üf׫ü\E™-¬Ž&ø 5@„ºœçÆÆÞ^ Z€Þº3ê2ÁhhØú‰ÛÅ&¨ ²Ùnxn]]r9kÃó­9.¨˜Íµñ  Ç:Q¯ÃBo;MËÉ@ò®ç.£* ð«—IÝ•ç`ðN€[—k±£?æ%Õdžæ”à>Ñ\ôÃäh\ öq³G¬¦×®3ḡ·×ï¹]$ýÃP¡:Yþ£»æ/\S“¾¢~Ô*ÄÖˆ°”zíåÔ4ÏþMÞµ½êå˜PmYÑšrÝ.µ×‘½„{@ÝeùUò÷/›K¿Ý9ûw¤•9ç~$ß¡^¬Ðß™x¨\:ùªËks6æ‚·­ÚŽkS’ŽÚ‘²K·m.9*Ën.6è’%óÿ›Kû´Ò\Õa\öáÞÁ½Áã8Wu*Š‹«^èÜ1Ècr‚li§ÆüGÎæþèK-eã%„Î}E0èŽb’Eqæ÷w\™P˜”ôöÔõÌ»âv¸;> stream xÚ¥Mœ6ô¾¿b•Â( 16ˆšJi²iR5‡fWí¡éÏ`ð<½Ï~6Ã̲I£j¥åñüü¾¿˜xEà/^Å Òt•’8J[UÝÕ竈ӄ¤–`Ú£Ø]sˆgï:¾z­®~…?:–áŒçwWÏÞälUD§|u·]Å18ç«„@ÌVwõêàܱiJþ|²˜¯T·/9ªñj‹ÏAŒ²¾ýí5¾•}€nÜý÷¿Ü °Uƒ£Á kÙ‰~”ª/[Dmä¾UáNÕ¢]ÿy÷ó³7…qOÄNÆa Úgl2‘$E³uB‚›¿Ënߊñt1ŽŠ4µÉ*¤IDÓÂ]‰bwéƒZƒÇ°Rí¡sæ•ã¨*Yjéíµú8¾g¶qñ¹Þ5ÂxŽ¥qP"@‚Q¬iüµŽÓ@ ÆVsêøYÝ”±M¹¦Ä‘"f#DTûAíÕ(jÄ[wNŒZÙïQ¡}”NöC‡Jãðñ^Š$ؘÔ9hwª`®¦«1O§\q×w MIÄâh-;¬>•;WÌUéË£•«Ywt@o™¬r' ¡ú¢ˆÏ‰ÅÿÊVôú¬ºÎE#wp°j¢ÑÑTâ'ÍC§zH¡óÑtªóS¥¿²Q3!{Ðe¾\Ô&5³óû÷-(G|–øp èp¦ àÓÑ"ç-æ‚ÙQê!íeË^C÷¨NäáK™äÿAܬЖ z{håFî’ž§ê"*˜ðbCfð[DclÑZ \V¥m3òÔôð|f€©ô4 ÞmñıàÞŠßlÅ›5V”åvÒã–ÏâÆ8 Š˜ûäBË£,õçØ¥æF ÇÆ/ˆYGFæy?\”D e^ÔõT ýEM´jž|wá7¬‡î|æNWÉçó2š¨±ËÓ©Ë[c–Îk?Ï£$/¼Ò å“ ¸»¹¡J#X'BØ* Ÿ<¦Á.ú å«©tU¾›Ùé<‚îÆŒDEB¡ŒÁÁ±«AðÒ’Âf1(ØÿÒ·à3~EDÿŒôÅ’Tè¨$9 ٢ЂdRÊK£4çsyO—äåf‘;‰KÅŸ‰«Åå$ÿVq颸øBܰĉEP-syÀn)BIÄ/BT-³K‹dÎîùBÊA.W‹ì<~_çÄîT´€a\Ú}‡MCÊ`uc[8@¾3fxèç)™g¥K§vë¬.Û·¢l¡{/tž ÖÌãä·êÂp0¨—;hÆÐ ì‹©q z^»=ÐŽcsÀr[úŸÁ±‘UƒŒ¼!NÀåBa«Œ³Sg_Yëâ‚ùµÎ€¸¾§™6l0OÔx’F¶õ“u ;APç<”ð!‡ÎónL{ÿö˜AØŽ|ƾ:€||‚ofd*jªNV¥K}Oünonq'[˜r/!2ê°7Ñ„¥-Ä°³¤ÚÍ PóyiÏp`wQ›Žž$ÁÑÑ× O{ûÀQŽŽ»vG¸õ°Wí½±À~øqûPŽÚI’½»×8z›NÞè˜y¹;wrq§µCÜ~‚hKôÙ@˜‰5KÖƒt[¤x0qF?ŽÎü‚ûÞÃÊ „E$›¶H=ˆRÛ°/ŒÉ æ$õ”Ϙ1|ÉÆ€òkÝëVí(½$¹Æ³Iýo‘yø5 @ò˜„……tfýÝÔŽæ«IœEE> Ãåí¸ðäÔ±Nû‚̧ eÜ‹jªuü(ö«…ÿ0¯ýöÿ0»öJŽÓϘ5l]Bâ‚;û9ÀÌwdÿÉÝ9ôv[\L –EœÎ+ü, ^Åή¹÷ÁÆÿ•:˜æc0ÿàcŠÓS|L™áßa7s}åz¢p}×¼mY~gúõRª×âèúÿìÒ<3Éj¢Ç“U˜0Ø@CjçóÕÍÝÕ¿yÞµŽ endstream endobj 460 0 obj << /Type /XObject /Subtype /Form /FormType 1 /PTEX.FileName (./gnmOverview-residSVDplot.pdf) /PTEX.PageNumber 1 /PTEX.InfoDict 475 0 R /BBox [0 0 432 432] /Resources << /ProcSet [ /PDF /Text ] /Font << /F1 476 0 R/F2 477 0 R/F3 478 0 R>> /ExtGState << >>/ColorSpace << /sRGB 479 0 R >>>> /Length 918 /Filter /FlateDecode >> stream xœ­WKoÛH ¾ëWðh<’ó¼6m6 tmtmnœühíýû%5²¥d»ÞhÚC$3Ô'>æãC×€ðßš÷ð |6ÖAdã°# (gã wðìšWÇ¿ÿx WóÆk- ¯ó«¿ä¿ÑÃ÷æãg°°j®åï¡A}nš×‹æÕ;s‹{½ —äI[~p¦Ö.³I ‹-L6SX<ˆzqhÞ.þÎÞÄ(74˜ÆÃ)±ÉŒ¾Îº%ã­ÀÙ_a=&%v$“\ Ü–Ø%ÓkàhJ ‚Á\Ö ko–*àrb)Y!ZÅÁ¡ †" ±:1nƒñ,pq¢Ây””s(Ö+R‡>¨×HAÙ3„óâ<;ãjàµâP€*hƒIb÷rѤšØ¥Þ%åÈ„Pc=jÁj r i¥`YbÏN[Þx¸ô ›c¬j„A{¢ÆÀ5ÍJøCq¢Æº´v/pJ« Ö™L­\Óëœ3AoÑ„ç}ÔIA 6Öôù¨çΘMªp^¼Ž öÔ”ŒLSVÎK½×°Ž¨mÑÕ±NÜI³’>ïª8t¼`FeOUŸo /Uµ ‘.Ê­Òy'6·¬óU#RÚŒXO¨ ·‚u²•HÅe)ØšÔ%§KÛ¨ÃæßpÙÝ~m5‹mXeñÛ'Ùc'n`þL]$¡0­Ž‘óã[èÄ^Olu<ôØëY&ÿÐzú¡;gïz½&¯ þöø“à·²¶¾£.·d:E¹98e¤2£tÙÓìžÏå"¥›Áôr0QT>ô`ûr,Ëâ"iì±#ìrj÷†{¸|(ÝXR]$.­VSýLÚ‰‹Ôkѳ>~Rwb¯'Ù•Èžõ8ÐË^0x}'öúΓ~àÝ0ðA´³Aêà”©ò…ÓºÖ}§›º3&<á¿û2X¨OeÑo™}ÒD>¡ÊðƒoÛ8[²Sêþ$r$“ƒÈ]úŸŠ§‡çmzI)rÇL÷Œ™­ìä¥ œpÛ¶ôqrµß~]ÖÇýö÷òz\¯æÞÀr·‚›?ßÂýT& LöXÍVëíÝî¸Þï–›égX\ŸtѨDeR’ÉTf|YÝìa»_ÝmF4©–­0é—³¾çvwÿiR^v£ïú4qnØfehd_’QÞ4\¦0“êŸ<žÃüßï@†<°H?(Ës윔†ìZJ;n=×B|ßü›mc endstream endobj 481 0 obj << /Alternate /DeviceRGB /N 3 /Length 2596 /Filter /FlateDecode >> stream xœ–wTSهϽ7½P’Š”ÐkhRH ½H‘.*1 JÀ"6DTpDQ‘¦2(à€£C‘±"Š…Q±ëDÔqp–Id­ß¼yïÍ›ß÷~kŸ½ÏÝgï}ÖºüƒÂLX € ¡Xáçň‹g` ðlàp³³BøF™|ØŒl™ø½º ùû*Ó?ŒÁÿŸ”¹Y"1P˜ŒçòøÙ\É8=Wœ%·Oɘ¶4MÎ0JÎ"Y‚2V“sò,[|ö™e9ó2„<ËsÎâeðäÜ'ã9¾Œ‘`çø¹2¾&cƒtI†@Æoä±|N6(’Ü.æsSdl-c’(2‚-ãyàHÉ_ðÒ/XÌÏËÅÎÌZ.$§ˆ&\S†“‹áÏÏMç‹ÅÌ07#â1Ø™YárfÏüYym²";Ø8980m-m¾(Ô]ü›’÷v–^„îDøÃöW~™ °¦eµÙú‡mi]ëP»ý‡Í`/Š²¾u}qº|^RÄâ,g+«ÜÜ\KŸk)/èïúŸC_|ÏR¾Ýïåaxó“8’t1C^7nfz¦DÄÈÎâpù 柇øþuü$¾ˆ/”ED˦L L–µ[Ȉ™B†@øŸšøÃþ¤Ù¹–‰ÚøЖX¥!@~(* {d+Ðï} ÆGùÍ‹Ñ™˜ûÏ‚þ}W¸LþÈ$ŽcGD2¸QÎìšüZ4 E@ê@èÀ¶À¸àA(ˆq`1à‚D €µ ”‚­`'¨u 4ƒ6ptcà48.Ë`ÜR0ž€)ð Ì@„…ÈR‡t CȲ…XäCP”%CBH@ë R¨ª†ê¡fè[è(tº C· Qhúz#0 ¦ÁZ°l³`O8Ž„ÁÉð28.‚·À•p|î„O×àX ?§€:¢‹0ÂFB‘x$ !«¤i@Ú¤¹ŠH‘§È[EE1PL” Ê…⢖¡V¡6£ªQP¨>ÔUÔ(j õMFk¢ÍÑÎèt,:‹.FW ›Ðè³èô8úƒ¡cŒ1ŽL&³³³ÓŽ9…ÆŒa¦±X¬:ÖëŠ År°bl1¶ {{{;Ž}ƒ#âtp¶8_\¡8áú"ãEy‹.,ÖXœ¾øøÅ%œ%Gщ1‰-‰ï9¡œÎôÒ€¥µK§¸lî.îžoo’ïÊ/çO$¹&•'=JvMÞž<™âžR‘òTÀT ž§ú§Ö¥¾N MÛŸö)=&½=—‘˜qTH¦ û2µ3ó2‡³Ì³Š³¤Ëœ—í\6% 5eCÙ‹²»Å4ÙÏÔ€ÄD²^2šã–S“ó&7:÷Hžrž0o`¹ÙòMË'ò}ó¿^ZÁ]Ñ[ [°¶`t¥çÊúUЪ¥«zWë¯.Z=¾Æo͵„µik(´.,/|¹.f]O‘VÑš¢±õ~ë[‹ŠEÅ76¸l¨ÛˆÚ(Ø8¸iMKx%K­K+Jßoæn¾ø•ÍW•_}Ú’´e°Ì¡lÏVÌVáÖëÛÜ·(W.Ï/Û²½scGÉŽ—;—ì¼PaWQ·‹°K²KZ\Ù]ePµµê}uJõHWM{­fí¦Ú×»y»¯ìñØÓV§UWZ÷n¯`ïÍz¿úΣ†Š}˜}9û6F7öÍúº¹I£©´éÃ~á~éˆ}ÍŽÍÍ-š-e­p«¤uò`ÂÁËßxÓÝÆl«o§·—‡$‡›øíõÃA‡{°Ž´}gø]mµ£¤ê\Þ9Õ•Ò%íŽë>x´·Ç¥§ã{Ëï÷Ó=Vs\åx٠‰¢ŸN柜>•uêééäÓc½Kz=s­/¼oðlÐÙóç|Ïé÷ì?yÞõü± ÎŽ^d]ìºäp©sÀ~ ãû:;‡‡º/;]îž7|âŠû•ÓW½¯ž»píÒÈü‘áëQ×oÞH¸!½É»ùèVú­ç·snÏÜYs}·äžÒ½Šûš÷~4ý±]ê =>ê=:ð`Áƒ;cܱ'?eÿô~¼è!ùaÅ„ÎDó#ÛGÇ&}'/?^øxüIÖ“™§Å?+ÿ\ûÌäÙw¿xü20;5þ\ôüÓ¯›_¨¿ØÿÒîeïtØôýW¯f^—¼Qsà-ëmÿ»˜w3¹ï±ï+?˜~èùôñOŸ~÷„óû endstream endobj 484 0 obj << /Length 1142 /Filter /FlateDecode >> stream xÚ­WYoã6~ϯðõw#†‡HQF»ÀæÂn‹ßÒ`d:! ÃѱÁö¡¿½C‘¶%U¶‚¶~¡†œùæ$g|¹:»¸%XÌb *f«ÍŒ $b6‹AXÙj=»ŸÉMmTj*“?-V¿Îð, Åœ»óoMžÛ£€S>¯jUÖA³u”©u©jSäB“¢™2ù?Åܯ',œðu‘k»q+qÇà¢ÀE"ÇõÑA~»"Y±Ö©è;½“NâJ¥érÌÚ§<ûs¼)ʬI•ÃýÅ-IÑäµûüË-w7wîãƒ[~û< ›´¶hÀw¾ç€ â©ÊLú£§d[˜ª*òóÖ6ŠQLIßÀµªûfe:¯UúY«´~¶Ø­d°¥ œöI¾*ôfc•s^D(Žy_5øKÙIô¶ná˜sl½½ì|_5‘!ÉxÈAk€¡Ì"G|ŽÁãV˜p¨Ä0 _w´Üt¾ok=)_'…˜Äb¯Ãâ`Å2¤cé†|@ Ö‹ ’Q›[@P—ú°c²­2¥^‰c"å^-Å,Â*ûœÍœ"†Ãn_¨_<ý´¨S@#,—Ó,.Ccª!i\r¶·‚ÅÁcÇñ¿²ûzÚ¨›i—ç0B8 ûi¶µDIH™•""ûDÄõµŠ"ÁÇM?\HÔ*å¯:M#€s‚Ë Á\»":vÇlŒ%¦‡r… ËÎ}¡Œºr ¤„û+ÞìE¯X%CBdíeˆ0ãÝst_‰kýݨ<ÑËE Æp›y„!®cõÿU«^+ÿ>>› zi¨_ºÃýk º2ëF¥Žs½Yúºþ¹gÇïÕ3\OJؼ, ñêˆ*) 3óRWŽ¶ÚªReÞ2¿G°[ë®áœp÷ÀÙ=•¯ýá>)Ò&Ë' ŸFäø€\lÀÐZj³MM Ðzð}Aø\;AcŸ^•ØÎ8¶DYƒ Lµ[sèTиÿÔkG«Ê­Æó}z»kóÓ‚Cãv[úÅŸYGbä‚°o³(Ê{m¶,^¿–Åc娟·¾š¶ýàn;òÕY«í6ýaÛz¾ožþ«j2,K\t÷œ€=Üw¶A<>îzsúØ7è¿Ø$†qºóUÓ5*.lavÓŠE¹'xIøÃiß&–$>vIJÁvÐózpø¾_ Óé·>=®ppø¾Ÿè …ºV¤¯¬z)ÛWsèÏô¨GoPBO*Ùû0PÒõb¢ŠŒƒI_ Œé­«¿7 v¼~ß1ñ|̇~ÂÉšÄ Íà“0y3˜•Ó#0t"¨Îßé þ]ëýÎ=ypH0i Ê(Üqµ]ÎKƒÛ#1?=’ñɉŒO dS×­U þ¡Úw‹ B(cXrh”0äØ è0o ¿CEDp$0LL@NÚI"Â4fþOÁîéQ¿eRc19€…Y ÌD;¬9k’DZBoëïòf¤ÿSèBJqTú\Ñv´>»Yý j;¶° endstream endobj 488 0 obj << /Length 1750 /Filter /FlateDecode >> stream xÚ½XIoÛF¾çW9Qˆ5™}1š‰c£- ° ôä@‹”M”"qÇ—þö¾YH‘òX±´ºh†|ó¶yË÷øáúÕÛ ‚åÌ #©œ]¯fJ!IøLI‚°$³ëbö9û´«í,ðÕùÕÉ|!¨Èl úôÛÕ}Y×ó•ðìÕºª‹ðèQ[”ÛÜ–ó¯×Ì„B‚ñÙ‚d„j$•„óùBsŒ3ŒÇÌh3_(­D¶€LC¥g‡³GäWëM^mËÂV QÚŸ•áì‚ î`ãHÞ^h6òÛ¢?AšQÚûÚ–Û&·Õ}›‘•õÃ\Š d3*3{W‚`x\W[-ó:ì—mc·yg»( $nf8ð_ˆP=IÞ) q5°ižd#{’eÞÍn¼Æ$kol^5à-ÿt×UÍmBŽ!H`Ñ3¹-íÙac‰’!Œ¥¢‹î+{d ÎÊ‹oΟycóÛ¨½Ëm¸tÆ bÑ“[÷G©ÀYgÁù¶»r»m·]X¯ÚmX ´û;ðÛûÊG:¬òºkÃêf ]ov¶,N.À ŠAŠÅ\ú5DÜœÿÿeþÇ®qñyyF\>Ô1H7Õòﳶ\%^½þŒ¾B4¿va|â}€§aûf¾ p gÛrŽ¼ ÛöþÏm{ÓEFD_ùWYÝÞ9“tNÂD€˜˜ÔüT“ !&‰˜´lëg™ÔÓýÀ¤>¿¦%’`ƒ4ÓÓT?ïlµvÕ˼² «sc!4‡sãÛé Š*ƒÝ¾ Á•¾VBÝ#ê4>ÀˆB•#X§T?ÂðÞ!Å\h¥†Œp éþB†gûó +AÙ˜¡!P«_Èðcx•S…±Plo²bT ñB†ç‘!q¥…kŒÙØdh$•2G^ 5‘8Œ4Ô؆Ó5`+û„IE'ÐH^\ñØ3›]èÛŒøhŠ)Íø`1˜SMyæÉ&ïYúÎŒµ”{ß`L”¡Çnï$è£Ò@jeFAa8ôþ²:}•Øñ‘¶Dq¢‰Nô¨ƒ›¼ªš¥kARÅö‹Ðœê]ܶ«D@4€®b«»)mžEäéЫ.ð+ÊMÙecÃv·i›6ù Ä®Î·n/}tí9Ðw¾gÃÊw¼ÉÁ¡ã¹ÞkdVE½hû°q Œß6­›è CÒL¦j•PR͵ ÿ ©Ø-ýC$+ ñŽtïœ=žöÎÉu+0(_—À'l ¸»¥ @ až]ßÅãÏbƒËUH”ƒ{8pWýè"RRP¨ÝÕ‘ÿM¯rC'n˳j +ÊÎ…ÙID&Œ#¦ðè·*?€Ðm|6ßµë^#CÂP"®ià¢$¿‹^h‰—g.ê©ïl‹ë¬éP&BÙÁ…ÊD$2Ð=ÿÎÃ@·¸‰Ä¥µß­‹È"Pû+†7=e`åëq¡Ó®m ¾_'T» ë ¢BÀ²WHãóª±8ÆVjn¿µ äT´«hKï“5“jSCô±|xç#<_Úªm\4]¸·}a¦Ps4 ¿IV·åëM]žÜÿ÷ç—UÄ¢·#%šE¼8")âRM«TÝFlN˜˜"|6vo˜›¼'%xŠó$blæJ¤áÎçS¶ŒA)'.¥ÀDú.%Õ ({¡„%…¬¤>–'ÐyoRò ÒŠ‘8žçFݱ¸eRô÷—ŠIqD¦Œ›2bt,¸¥.ˆ#Éè¢<¥IqLñçiO¦Ú³tx¹94¡ý!;a&W}šÁ©í\¸N°Ówèi³œB¦ºö°,]q²‡ÔƒÅ“YÎÉ8Ëq(…°¸‰ݦ\V_0¦¾º{þ[Õ pÝòQveR³±pµŸé·s¢³ÛÝÚ£'" 2¿ ÉûC¥ztEFC7å,óóí^óÙ‚SŒ SÓ"v¬RïŒFdçkË8yÛv_õ_'FZï¿r€™yý[™×C­Ÿ¸F"ºÿÞQä}÷̇ɾ®[w÷©¯)0œC—á29¥_žÑhÓd°mÖÎpÀÎýîÉ?˜Ÿ_…Å›al˜ì‡øsÇãÓ8eÄ=6ýj« ÷Ãd®Ý´U×µMd|1z=öd?úÎ÷÷¦²àÇþÏá@p¹k.ý¼mz.v›h®Ë—0B);&G× ›þRL>Bß~öÄ5Ü绣]g€sOSãíž„”;*ü_áÒ‘’©b?'À ‘7 ^ÓÁ}§YAiré4YD ‚aÎH\Q?–½:¿~õ/µ• endstream endobj 496 0 obj << /Length 1591 /Filter /FlateDecode >> stream xÚ­XKsÛ6¾çWè(M"âåi:ÓÄvãÎt¦ûÖô€PÅŠTH*©/ýí]p!‹”`'¶ìƒ .ö‰Ýo±Ôû›Wo/UK¬âjr³œ0­ˆµl¢#T±ÉÍbò÷ô3•ôªê|“ûM/l6gÊP:½¾¸~?ûçæ £$î9cÄJ‰R‚E…¸…’t:§„J©À`/Ãa’e@á‡òóžq>âTÈI £Ve=cZAp“QjS~1P~‰Ê#ÂpO…½là;èˤ`¯$ÊÄûçÇëuQ.‚ã²­¾qGC©3¢D3ʵ’û(˜±\rú° (.ÖW4le:˜Ú–!-”¼™Í%šŽ«¢j;\½Ã é#ýoÇy»?Yž“ÙQÞÀ7Á2tÊf„eúÀ©g¸ðþ÷?ô®C,ÓG¹ã ‚Rjà?WÚ ¥1OI=LJóü¿è]I$%FX›etPØT+%øËžþå JNg' ¤(C—åß}Y' ÖÏÖÜpÃÌ>ÅÔ™ÜuF5éžâF„}f#xB7ƒ7‡¹ Ú2j²!2%3®b(,íé¡<ÑüÑ\Éí& uYk‡˜RCNì‹`‚Ÿ€iþ¦ÕÞK åvˆiÝWS/åÿù þ?†i)™d†ï!)e„Á4?ÓüqLï HI˜ F—9\ÖR–µz2øi æ€:Ñi†©åàú>”R¥™ˆ¡¤ ô_îAÝû’jpˆš1£ÅA;ár€,çþ[áªÜŸÍæJÀä¥Ã¹ cU¡ãÿË»¦­+ô#_óöëÖgH ²BĆŽ]ùäÛb±u%r.–g ½·—àÎ~J¤ýL|(ª ÌP!gàÞÇz]ßúÊ×Ûv6çDò¡¾÷ ,Hkzâ|ÉRVt_½ãñy×"0Ê¡4%, Qy¯–€ƒŒN»•ÇESÏ ’ïá…M]µ@j^—Ûu…ëåŒÑ©Ë»ºi‘°r3N§ßfLN=ÊÝkkÝ:®Jô"SÙB=%¦uƒ›߬ܦEá¶NË ­¨v­×uõ€½jÛÅ1ž'²X„1â(B„ãÏëmÀSX~ñø )K¿£ÞáÓÅM¨óbS¹ëŠ½£¸w¤ü{Ñ­pµz ívøq!°”=Nû(¥’Ø}i„´¿AÝÊuÑ6ÂE‚>) áXL~YßÆŠrøa“aÅÞ4Þà`Ú@Y 7oZ€6—M†|J.¦y"K8è°¾KYµð•’%âµTX=¶‘5´÷:eÏÀy ÌeIs´Ÿø÷šò¤9}ó‰ædÒ;0פ4 "4ÚûiuyZ„ý×Yªwp ý—Êqý\†‚æ&T1¼üëÖ›Ò‡*4ºÇ"ª¢U H˜û^UçùvãL\yݹn›ê\º¦V;‘…ëšl}¨rm§Ë¦^#é÷º^¬]…äp 1«-~‡‡í¢Å§C†¼®º¢0æÁù;ÜëÜ—ÒÇýÒµmñ™RÞ·[ìÄØàx• ŠqkFüÐ[Œ ÖlZ/‘†45q#ø[7ÅmQE©oÁaøŠ&*««Ä·Do­¿M¡ÿÑáAD¶ÁA€ŠS•8$tLÜjÛ:/zÝ©Ú·GàTl×ß?èo°•êo0Ðf¦¿¯µ^aW£%8’Ö#)Ý:o$úÒ‡üF©•GŽZzÔÔwgÔ„„µ+ª(\¸Û¿pž™œ^|ÝÆ&ÿ\é«®¼‹‘GÑÜU±$˜êëutT.‡ë¥Š]yÙƒFÆë*,vÖ¾¦Œ§ÞÉ›«õ·›üáÉÁʽ†sp+af.)î ç½>K¨šs*I¦28 ò8wýŠSÓ§}àË/s|ÞVëPª—ÿŠ„ÿð Àõë8uí«~Dïqôæˆ!€‰Š-á—±~†íç³éŠÔ¥ƒaün4ÛnêÐb˜Á²÷ÃÞaìÛ£Ÿé¸_UEW¸²hC½§fÖmUõPV 5Ý|»‰SyU‚É–Š"@Åð/%|^W©2;ðþ (%¤B) ¸”)G)ÞÿJõêâæÕÿ1XJÈ endstream endobj 501 0 obj << /Length 1032 /Filter /FlateDecode >> stream xÚ½WKÛ6¾ï¯ð­63¾¹h{ènMÅÖ·¤Ŧm¶äXR½ô·g(R‘äÈ®ÝÕ…¤8o^ܽ~ª&–XÅÔd±žHE”å­€P“Åjòvúív÷³¿¿Mèd@¬”qc“ïßQI×Åq_ï²Ù\29ý1OG÷!Îþ‰CqôŸÇù÷qX¹²òyVùbøÿÑg› 7²¼ú‚÷ e¨%–ÁÚïõ®úµØÿ*(ý]g{¿û{`¡ðeY|f˪¡…ÅrY1ÙîÏ*«ê²ÁÍ[D\£TòbáÖk¿ô.¯ÊèM ‚(ª‡ÀÞ7yåŽKw¨¢¥À•¤É¬³˜ )aŒÕ’S"¤ [”ć‚Å€2a"­ÛR1Öˆ 2åP¦£¤Di£hKɈfÄy™z(ÓŒQáF0ÓáJ§X(òÄW½`Èt˜<:¹Ï¨Zð–Ó<î†Ëk‰EòKE‰‘0ʲN‘Ô†²«©Eú¬"N(ÓŠuá.Íu®Cs¥¯;¯pÖ¥€‡RÙ³”©6vÖ®OÈ5ÚØ Èb°Ç4Q"¤¢M°ùæ”P»Ò¯êl— X#‹Ð§9—¯Ÿ ïu€ ðRˆ+±ÑQÀbfðþÍ™ah?6'{Ÿ»°äÓ—­«¶î÷|Ç— }½[ÅåûDûÞUUKZ%qØ[3Äò—ØKFÚmPSl\Œ{.¢µý~•iC´hqêžâ©:±I‹… Ѹe•o]h@ƒ ßÒ§Ž´¡X'´ÛdÂ:X–‹ËïÒ%‰eFƒßŒõëÏuž‡­F ¦Í±š×‡¸ò„FeIÈhè³î³6}¶ôq?¹O‰>ü”Y^|Ì‚Ë17B‰}u’)3‡˜5Ù…• K’)ËXó¹ûeq÷ Ò3çÉ endstream endobj 511 0 obj << /Length 2106 /Filter /FlateDecode >> stream xÚíYßܶ~¿¿Bº[{iþ(Éh Ôv$ˆ6¾·8@µZî®PID?mþÚõ]_õÛ]"“=Ò÷•¹­Š¶4ôëºØ×fûóõ÷vR±ÜÏ}c¦&ñœ¾¯;óžZÿñ»êTµÔ~BŸƒé‡ª-†ÊÎû_UÅéO8Myú‘0Œ‰Ù„7—zøÖ6v6 Í”œåRD;!Xžh2v©åv3Áä#¯äÿÊ«JpHpë¹Ü¦¯Ìã÷wÖgnçƒãæßk\, Ç’óŠ·;©aŠ’,Ñ"@d**½¨Ì¼¨Ì™ˆó-èMøFÃ7Ia>‹…Èpþ³×™š@'X’’²ï ´¦ g,¶Ê¢7¡Ït¾ëjjßwÚ&›¢òó÷¨k»›>Tí‰zÎf0=™ÖØ‹×bȬ|šIRA ÉÌsv½ãB­Ÿ°¦Ý$ƒc8ÆôÊ_Šæ&” <Ë,ÿ÷­Û¡ˆTeȾ#öþ‚9³ð­·e¤»¯`ÚÝÊîüºÓ¤¼Þf Ëc¾©êúÒ]1úMÞBãÒû†[PgHaï$•ñ§/®ýÁ÷õ碮ÊP=ɧ±ÑPÒ$çÖbª®8£b&²1†‡b(Hë±³ µ^ÖÕ ÖÅž~Ô³ c»)$°@ã[S`žc“*Q®°ü0xÍ!›!幎½¥J+ØÏÆ÷‘ÍØ¢‚A(i)u½èý™~àáйTS›÷0ä¬sÎ,”ݘ®4íPœüÀ­— ¬ûýC±·—Žd+¯îÔÙËMOí²³}¿[«€óJ)ÓÜUküx áD¨éà3;›â0Ê6g8úÌÙÖ¾û‘'pZ›?a?+{™1)&µ ª¦Ý|ÜÒª£&U<é6 (CÊó{ð8²¶‚_Éx’OÍ‚Aå)ñ 4¡;ÀRQ1OY¬æ‘0i“ 2¬:l;k‰‘d“-ï©ÃíK¼…·ÁÔ×W¿š§ˆgN £¸¥]ïõî‹ò_;( mghC '´ìBsi/íÐS'•(·Á´)úPÄ£o]cugóMkjÌÕp¯æùÊ– Öœéع¿ÓìÍÀzc.ƒG»°ÀË:¨ùç}ñ¨Ç©T <µ¾ú3 OÔ÷'ú v(êG–¹»-ºù*å¾j…ο -ù§‘zv3à ¼ ¨ýqì#¡§E‰'„Ç.¸žCG›âŽ{C6‚½~´ Á}ÑW%õL˜0ËMé×uŠ;„O1Poå˜ìnA¿ âa¹þÒÐBöè/ Á³É…k<¨3»F~’„eù—œH}t÷ÖÝ/ƒ®8RL‹tÜ=nln&”«˜Lº"ÈxÀ¯ÃùÕß®¯Þ_ Ø  V HéH¥9K’,*›«Ÿ~æÑ!ë\×'ÚDJ1GêèíÕ?®^à“ÅÜüQgB?æŠ$ËÒükbvÅEÍ€3E¡5%Ð(;ø$Ÿý5~.àƒ©¨ø=à3ªú?|þ øÖµ‰4žâçùêu'ƒº-bwï‘¡½5¾z¾ ä(¦%½±áÒw0CQùlÛEÁ¿):¸ ¦«~uÕœî³#izpå›ÈO§é=ê‘ÒôlBÕ¿ô÷w¤Ù]JÂî@Lü>ŸÚfLåx¸ãÑž§úÅ ’BmOV‰•;ÊÞ„kèŒÎ€HfîyŽÒÄ{òcï”O׈Ó`@[ÁÞ¢©j¿Ú7ôêd¸uîâ5'uO—æïÕþñOçáõnaE]…H/ ùñÒ¶îÈGÍMÞg+@†ó gþomò+ÛšuÞ:5a5¼[þnÞÏg«ês¤ô—zš¯Ü ×ûàN|ò=ùá͹,Ÿüà¥5ÇcUVÀU{rPp(3PrgªÑìïÚIí ‘czþý— ‡‚, ÓùâYꔊ•L@£HÃ9S> stream xÚíZ[oã6~ϯPb×ÆÄ,ï’Í>´ÝA[ @/ö¡ÓÚ¢m.d)ÕeÒ‹ýí=IYräÄž$@ŠÎ‚Zåª2!ÆúÄu’ ™Õ۲ͽ´°—5üF¦“"d`Äø)ñï“¢eóÂùˆ³¶”é¥Q¸LŒ“6ám Û¼ÔÍm §ÎhÛê÷X]¹ŸÅÚÏè! Ô=N†ð¡É(ÕN0®×‹îòá.ÈÝ*/B¹Ç¦ÃËwêI¸(b‚œ—ÍoV^—áY—+1è(ž&ˆÑdÌàºQM ‡Œ!§·õþ”àݹ_7UyÓíC£кÌ=t¦Øf:÷g &FûaŒ Ù_R3,¢Ò[”Dì`‰³ˆ1PñÔëXfSAÇ 0f0ìÊ/Lø`XŠfâˆØ KÂéCBWNOáD œL´ÑÇ=eîkZãL]¾Ücš2þúôû+®ñ¶éc$`Ä!ˆô>Üö!Ä!6XÄ$²O²O°‘%„u)QÞÅæ~`÷…pñïë‹ß/læÁƒ8D Ó97d­v¿þ†£ ú@e$DÝv#wcÛ\”G¿\ütñ•ý@©ßKb1¢±|ÔŠ’8ýrlíØn(aµ7à&`KŸH)ÅOTŠŽ”¢S9‘# Ó˜ôâÄd>$ÒåCY儉,|†MY_J1¶Å\¼{è{àÜÏçÏÀž é{žÂžlâg[Âä>f=mãàÌ:U Ìê?z¯&rÛ1=ZbA(Cð™©0ÀÒ¡){ ÉÙsEúç#(à[îùúa/+êß }²‡÷ÿ' û›R¯×fel9ÈaÄ”ô¬åßàÌ•¾iŽ¤ËDd Q –áþÀú«Ó?ú;ç endstream endobj 531 0 obj << /Length 1223 /Filter /FlateDecode >> stream xÚÍWMoã6½çWø°]sù)ŠA[`Û$è-P¤.zhz mÙ& K©$;0Pô·wÈ¡ÙQ² 6›CDR3Oo†3ôwÓ³WŒ¦#CLÊÓÑt9’L•e#2BS6š.F$F„0büçôÇÑD¤”dF&Œ£Z\Ôùò†*ZÕnåÊ÷ã‰â*YäMëJÛºê~¥h-¿ÁÇ?ø øø*šUwå­G‹Û[¿›0`˜ßïSÙæõ<¿mýzà'Ò”0ø–#z”°L«È_H¢¥ŒéëðgGTÉÓĻϛ^ÿv‰œ(8 )ùgõ_pÔœôXãg|Ž3gZvy2ˆ7Èþ¢žpÅ“»Ü­Ömƒ“eUûKÚuÕDƒM5f*Ù¹r…sW¶ŽÚ£±…­mßwë8³¤Ú¶ø¢Z>‚v¶\<Àr%>mé1÷"çU]mo‘ÌÁh@±Y€¢I>™vþŸ-¶¶Í¸l|Îò¢ò/ôXG3Ã÷-Šï³ówTÒ'iSÍ-~®f®pí¾kŒ§Â.$B½Ã ¼™ï<'YwœÂ˜»À—A‚acøA¯™C,:\}üé×Ë1˜¨Ã؃‚•Ò(ÓM©zL²¡Ì倕‚UhGP)@ù«mœ2.4£‚A‰ZJw°è+ ¥Ú𫌠)”é ÂÑ‹gï]ï¿­” L©d¯!¥ËêÇ+H™Rc8͆¤|¤bÄôü¸Ãq —®M<¸Z$uÞl‹P@`²pdúyžQb¤¼€sC™ØMA‰Ì:›p¤®)|0Š=ΖuµÁQW´„ío«O0¼˜EÓï‹!"L¢iö&U‘±"Áà‡Ü¶kú£ÎŒþ€GYfcN“Pä€ÛA£¼irè8 ¥/|ĉ‚Øü$')ÇAu ¯`]ü ±ß]$7÷k±|„q—Å‚šƒóf;_ã*í˜eÉÊËìç¡ì{ÓÖë_y7† ‚€áKZ´ õ¶ó³eªO5Vho<¸šKCCÈßû~$eìG²SpV{”¸TJþm¤ä‡ÐCq°óÝ"¯÷8kÜÆùÏGÔê}fCZj[„÷-ðedö šSßMµè,76î6‹h°ëmán(å¨C}i§*‹8ò¶mû&Ä;¿µ54*œõz6íz6íZépðüúA)ƒwþìþÖÌ®¥áê×|®Ê? û¶wûü¯£€»3òå7;<„ –K ±ß!Î\Ym UwÜìÉŽ»ÊÇà™ üSéZÀrÛ‘ëmY†ˆzäuñÒçà÷s£!ñoÈù¢*óbÅ!Xi*ý=”Ícáìrzö/µÔ¿ endstream endobj 419 0 obj << /Type /ObjStm /N 100 /First 892 /Length 2416 /Filter /FlateDecode >> stream xÚíZKs#·¾óWàh‚h¨ä¥>ªœbÇSVÖ¢Ò;£P2¼²Þ0ª!ÚK['㹤lÍFÎ;eSŽx@¯,Ý=+2>u¼Š¬ƒdñÀò`¹`Þ*òePüä#à±WÄø0E4ôŒÆIfÄV9c1(“r–¹ã!ËQ^ËÊ9àõìð j^9¶‚£r¥{Rø³à¬\Æt}ÀÀ&`¢kð07Ì°S PFM”™`–>â…Æ(íðO†‰P³¨M&É?ü‘ŒnLÅC:Û aP;‚ž!‚Nq0è"Dm$5@h±&±p ç(#CÅœd¬„6YT—X“¼ ¨‚%è9EÈJ›¬‚³Òmœ(<¼Œ•QèöÐÀ%¦“!' f,p4YG‰BǺŒYe,Lt°†¬(ÚbC*²,wF›`09 €EHÈIÐ1ã%#Y%˜`|c‡aÂÉ*C}‰½t'•Š`Q):ô²ò­3¤§;gB›,ŠÄºç¢$T&Ñ1f’a´†¡`6FÔ¯rîP_NÆp…œa Y9cN,6o`㨊x‚áá 6nBÉNE ù  ™%(Z ¦ä5Í :[ÿ"–Éìf‚¾ƒf¤£¶Çƒâ  †[˜KDÌsìµEˆ±tð´#!™GFüɬµ5–DgDá „´ÌͯÐñ"q-qÕ2‹-²¶DfâGÙ$?ÅÇŸR\Vö&¸¼NpñF?×]Y„‹î´Vß\ü•ŒEP'$”l:ƒÔ¦¿ñ·Ïƒé¿|TïÇ݇ß½‰úq6}˜MÏÊÜú³^=–_Æʵc5Ru¶E b+˜4m½ì I#Ü~ÉÃ=2lo¬÷…½±\xgdc9oø?×ý~<"g‘o<Ãð‹“ö!þÅ~{_ävA¢D¹­‚ë溼×Ùƒ° ë^õJn“Nm™é9kÙt9S&Œ(j8è×]€˜fš¬3Ëi’Õ6b™„Œœ!Áb<ÿÉ4=(—X;9½BÒ]ÜÛr_vƒÛÁÝl\ë°êaë>à@JØŽèK¡é•4%…u·Ù×âú©@ÜíT úUmT Ô–­_„¿ø¾9Ï “ùsòÍg¹‹ùf²ÿKùæa¡Öb3ëá™.€G…e’)[~“vñ•`û¾iú1/æ[ìè‘ù>â ƒàÒéq8D¡Lå8©¸œ0$M¤ÆNûä6y{7® ó4ªÐÿŽèËiÃ8.ÊIÎö`±­E0å“À ;’5Áópm ÆNp«¦ˆ¯ÜVmËYøy«s×¥ð²!¢¬_l¿w’S9sPTJ~-*%Ú'*-è,¹]ϪWB÷RÂú¬Î½ç˜ÖçcÜÞŽ—Ú>f2Y“q§2K浕Eí«ÄÖ”˜ýéKn7×¹Mcr›ÆdwÌ3}uc3B´ÃHÎ"µíDd,OÈèàã€xÖœsz ²­•Ð‰¹Ø닳ØèœGp:ÈYí#ì0¸(hWâe™·÷È¥¶­G²-JÙýðÚzùØ›^üæ¸ÈäË.¹ä­+$¿‡²±ëþ™#9ù(}‘oó rß Ûµ Ë÷ìƒI¾° $ùÀ>/¹-C[ƶLG%(rå{g§[9Bª'_-‚ñâ›áSsS'{ă7¤zr‡á r=ÉùND²m—Ÿ€8c4qÜ‘¢2²n¹§E¹«SbÅŽµ{-÷=n³;æP¤ ìŸpÀÛ˱Р€¬œÙaA|Ö”cL¡£B œ¯K9íVh[n·lKÛËmÛoŽµMa«¶ÌXÌå®WŽŒ–.Vl¢û} ÖnY0ítË‚)¬žm ζg[‚³©-çILG½eAð*ÙC²œ¨¬)F¤d0o–ÏNát~ßÑÆ¥' ¾˜wøïš™ëÅ‹7Éý./óö„<° &ÿ $…òèä@ä ×Ëe:ÒÙBŠ¥É—+p:Æ®‹Ü­0å6^ Ô”åÊàî@„”ˆŽ˜Pm>Ô~%Ó\¼ ·²Öm Ù­ça”¤¥öòÓqó¬$ß4EY'hÅ1¬ÊH¾ïtBoLJÀû-y bUÎœÀª^7ÿ¥Å$ endstream endobj 540 0 obj << /Length 1645 /Filter /FlateDecode >> stream xÚÅXKÓH¾Ï¯Èií¤é‡Ý¶‘ f'+v‰`‡¬8'q&-;Ø΄‘Vüö­êêvœ` NîGU×W_=º“§ó‹‡Ó„R–j©Góõ(ÒL§jkÁ¸£ùjô&x<žD2 VÕ¡«* =´‰ï M ÐÖ i,º þñ´…´uŇ:MmÑ—# ]:Ö¾R»Z¶7ƒC–ß™¬\æÆ­8DÂÒT%éÉ­éü|™guS•„i¹1“æÃ>«óÕ#‡9fyþêšs¦ÒTr%ü”‡qÄu¨†’ ¯AÊö5'¥©ˆbî§B¤‘P!÷ê½(>èeÃqMÃÝŸêô[] 9OB5è$ƒ<‹Ÿf v¬$’ßR2.4[¡üc¶Ý9Øx !×”àÁ²*×c°o>|=Ú½¶¢ïÞ|" àu„)Š+õ¾ÈGõ Mœ¥ Šv ú¤¬ÚM^£bd»]a– »P­é»2ÙmUÒùanäu…GÓm­ ¡±qß[ƒpïÆ"`´´¸§ïîdnqåM¿*Ì{å­S0`ÅÀsMÚ|HÌ7Î6=(q´®«-ÚG±Ëvpþ@X¡òAB†xƒ(Øíà:¾^q9K"[yǦfi¬d3MY§#Å´pɦ8‡½š9ë6 PâHa´7i?ægÈ¥ }¶+©‚EÖä…)sÚ°M…†0$AdÙÍ~^áJ?¯pnó i å±Õã¶Õ£µza«Ç-Ûêq¡ßêiÅ”Ëb¿Êý6—Ç ]›2Câ b z‘SÛ§4„Å Û³Šwvœ”Öq•þûÀ9PH•Y:uOI(DàëŒ$)q@Ÿã=‹z½{7áçí–FÓn:SM›•-z|q=¿øp! IùHø£„Œá«GËíÅ›w|´‚=HcxÝ%£ƒ•ÜŽD 7™P0.F¯.þ¾xjÿÙR£®ºØöt³(’pgI,Ï/»丰ÿ€±0WKÀ°zÙë £Ýdå{—¥uVºè½¸ ðî5À)qà.u¼7åªÌ73Nî¾ut˜ìoÒÙÆ.Æ.+xá2ϳI¨Y¢Rç™òlþ§T endstream endobj 548 0 obj << /Length 1745 /Filter /FlateDecode >> stream xÚíXKoÛ8¾çWèh£Ë7©`[ MãÚmÓ{h{Plº [­$' PìoßáC²äÈiŒÍ"{èÅ"©á7ß<8ëåÅѳ‰fQ‚Iet±ˆ„D2a‘’aI¢‹yôitm²¯ËºÇTàÑeZ™¹kÿ¬—& ®Ç–ýÄÌ7³´Î¡…}•ÎꢬÐ8f0¹XfsUÌM†éMÐ0͘ˆÑo¥©Z½i5þrñúÙ„0Ña®9âŠD1U(ÑÊ3ÿŒ©ô²ZG%Šig$b,Š â ÷r%³Ñ,ó²‰ìÀ‚ˆâQGøùöÌ"ܨ%<¨e=µÖ¥]½¤%·E3`WGèÇQ¾j+ÅJ5¤†ŒãHŠ®ª'ƒŽEBƒM£ÿÆ4¶ê@ÓȽl;Ø4öpQcÿiÔ¶Œï±Œ4þ? šx¸|iZÅHa°Š /e¶ØÍ#†$‘­dKCý#Ì“¢œZê˜ËN•8:»8ú~d4ŽHÄ(1)&5³Õѧ/8šÃË×WëèÚ‰®"ƶFæÑÇ£G/ÝåÑ¥ßBQ Ï_À"Š´J~i»mÀŽk8Í[CYßëÛnÖ<0):”¥pMm"bÒ‰Áˈì¤j9h£ÐâÒgÀÐôQÒ™øéÓ@ýNŸ>ìžéÓËžÙC"é¥ÏqËl+3®ƒ4‹)G¢I¡‹¦»³_ž­Í­žm–®û=ÛgŒi]wÚ5ßy^Œ©]WvÜÕNÀGÜ^ ÒM÷b ØQ™UÆÌ?c ü¸0B4bèÎ!ZYïG'~ϱ~]¯ì¶éédêþö˜øçÿ8ùólÚ[˜žOû “óé¤/ñ×»ó7ý•Éi»g€¡•RJŒ^•fá8½{ÿvú4 ÃxbM óEºÊò?~,I7U•¥ë 1Oë´÷~V¬E¹Êꛧwi¿2åeQ™ÞÖÉÉÛg·+{A8 ‰…»‡D¡è7á:MóüxHw‚¥·Éû¬%0P)Jú'*¿ŽFÜ¥ %¡œf±Èf™Y×þ0ECÅ€òÚ‹œ÷!Ž…w›;9lU;Ùq†æ‰I6Ž²1F Ú±R A)ârgwp~PêüîŽåî…ZR§d?¥Ørâ oI`‰y—"S,ì×p³iy@lr 4Eçl¢ …¬rû~º侂®¬G=6rk TDëÒ î BI;Ó’ÑÄÛ¢4âb-{³û2ä÷·M d,[.tË]QæÂâgœ Þ•½zå} ª;¸Mº”˜Ò’†ó$ù 땹‚³<3ÇãX2ŒGø[::ºïMZVöû†;æË,®¾oÒÒÌýÊ][ÏM•Í7ijƶp#¡äÀÅÜ/²þBV*‹hg…äêü=Ôˆ²¤‘ñUÄC ÿ¬·:ì÷—+{m§e–Ö&‚~°.ê°’gáÓL®³z¹ƒô--Ó•©M¹OÕ0þÝ¥1뀶úV8ƒíÄùêÍ@L(Oìà¨xH¸>^¦ jºÞ´D뢡·kª™#È2L]ÞøûD(hißMï'r4”XœÀéRØ€ SÝöÉ3—W€7·²ÖÆ¥ ŒêðÊ50o£á”ï­àÒ} ¼ñ³ÒØo‚ÎGVvÁ¦ªÑò/ìrÅP\ø‡"²)î;jfòÜ~Ðä<±†¬-ÊàfÆ9j{ŒPù†Å¨XÛ6ÂŽ6ën¬üZ/º0ÏlÊa›YX í·úÔƒA î?ØîBY–Š‡Ï®5<ÁwVy¹jÔ{uÀj³Z¥e w ¢µ³¡·öÚV÷“¦åùÝd>r“¹sú ÍNø“NÄN™ÚÌáîMò²À`jàÀ4« V™„{xšþjPb¸Æ™Ô:4ŽÐ- ÌtÓÕ@ïDáí ίhÞQ(ÿ„µ7îÝ 3ÃHµ—E8ŸgU­à–ó¬]í„ÑYYÚ"l‡µ\¥ù&ˆ½/mD_ü¬nÝÙG·Ûñ¶Év^‚™bJ´FiûÆY­­Öb°‰h:zló¹íÒ[<%eãAh‚,¸%Ëç&ÆbÐçx èûü- Ô¤dJjËcø÷1È0ü1Üéá[DÎÈ6ÈZŠŽÍŠªADÿ‡¢¡þ´€„² ×[ÀDÊ¡ÚN¡$IÉ!WliÒá«> stream xÚÅX_oÛ6ï§È€sÐX•HI”Šn@–Õ[·¦’}hû X´CT– QNšaØgßï$KŽÒºA€½˜äéx¿ãýãÑ¿\ üø õÒXÄ‹ƒ(öâT¨8ðü88¸È>L~­õâ£ù§ïþzszt8D4™Á|Ä ×E“Dœúž”JÂ*T°|{|8•AäÓ$VŽòéâÿ`^EûÊ|_D~ô@ùËöP®çÐþq8 pÁööV"ðý‰ïq,Sø{'"õT¤xñ’ = âb¼'¢è“XŠt‹è‡2f\á‹o!î{FÙC £d‹(’(b©¼$|4Ä°CTB†r‹(#![D0¹H 1êCŒ1•AØúQ@¬ËÇBŒ;Ä ITïŒ*è )œ:…¨zg”ªçÇ8õÛc‰Ø˜…{Åêt:S•øÕص®­©J’±Îêl¥]ÓrQñd™m¬5s-²•)niÞdŸ5“›ŠÆKM£ï)™€a"Kd¯È£0*HÐŘP¨`r£Íòª±´¸Ê…?¹> ¢ &;âl`G$ sq´›ÕP•¼9c±¹¶óÚ\¶ K<×óÆÅ.¤'pO,‹iFUêÊê‚tdÚ*»mµ km³ÊÉá´‚qQEuˆ†#ƒùCƒ)Þ=$ägN‡ºZ{MvYhLýe†!ŽÇÖâêNÚß½nÁ ©†îØûºÝçÞD¨Ä !®vð½Pˆ(–P@’ÐH)¸{ÄH¸LÓÔ“J µüÌ¥ñ—P†þý–FäÖ$†üÊ1Dqê"ÆÖ´šWu­íº*sS.‰d›¬Ì³:gvL~Ëb) €ÚU"Ë8 K=$t;¶‹ÕÆ64»ìô(GË`æPŠT &¯A¬L¢IYáºnæW4ëR1åÆU&ü”­×Ö·€ŠUƒS I« œÅ› J¤ô}Ñ@¨D¤„¢½ ¨ÛÅÌ{>á]B3w¬7¦(H~­¨6h è©R¸­Å°2u%‡Aòa2ð ¬ú ˆW;ÓÂ<ÂëºúZ†ƒìºK¹±,Û©wgzašÆ©ïîÓ\íÜo»uÚ8¥[HÞÆêÞ+íe‡ç «§.gÝ— ‚Ž;Ú¸• HMK…l:`ÜI㸽SƒP¦ tTw¯‹§®4Ü…Œ¾’*E8yÏ=#içÒ¸bá /…âƒÉ5Ò³bƒ5\ ÌuU»_/oû\%-\YAÞóÂ|&’nhÌ &páZ(ÐÊÓ±‹ý>v3¥›‘ƃL ú™öѧ6jç¨ t\iËãšËò4¶I8Ǭg­wt±}Ä’YÎ@Ÿ;h®*ë®C¿g)§ÎRHí,… *À0!K!Z )ÎR8é[ŠY°¤™ë5³òìßc$h7A¿ô!FB¡®Pj‡²rnùñ0Š&̲Ôew’º*4mp×ßÙé)+Žg*ó­ÆΆ"Ä@2´áÂÚH¨¶ aÒ‚„G³³ÓÇ’¶9q [8¤XÄ‘, Ÿv, ŸêþÕi CŽÜ`Ͳd­œu¶ßÔ} >Pò@‡‘ †e2Ѽ©j汚]ŽK 2AÒN\º¨ÁtI|ÿÁl¡¹ "°/C“ÜÔmç‡K—-0Ð!ÖØgãb[  J@—Ž†íô@Ú6™(êB[ký†OÄ~ø•ÆE[³Z|‚¡‡¨l1ÚX§Ð:,`c[c÷¸Lùžÿ<ÈgÐJWšgeÕЄKlÛ´‡mç ¤â‚$WŠ°£ %û´ÏtÚÝÓ¤Üd˪Dï8î±rr½sÉ‘sE^ª’–‚ =•^k—Ã]n¡)îÞHã-=î×ÙHãÆO ƒ‚Þû»ïQ¬ÍÐÈø~$„ÒÛ¸·'‚C¨ëu­›Œ_5b‚íè6í„(í½§Â áEþbãË5c3]êÆ=äF}ƒT„ÐØjI&1ï4]¡ŒºjŒ¯¨AÅøþçÕ°»ÃºÎü½œÒ˜Y¯Ü¬tmæXO«/ªzešÛ§Ý%0’™,KøÞòÚz?ö6gyÇCYËrån£“Ù)þå?-ùµöŒwýöêt@èóÙE[ž÷ïÎþRf'˜‘##—Rßx"Ýý¿æ§á:Ì‘gM6ø¾5ÔÑ×Яu}éZ•ÞÖÙñ›óWßðÔñMüIV/Æd°P¿M1TûÿñÍNŠ õ8á4©—$ )-Ýß4O^]<ù±lë endstream endobj 569 0 obj << /Length 1799 /Filter /FlateDecode >> stream xÚíYKsÛ6¾ûWðVi!xô4Iœ¨Ó6š¤ŽgzHr $Èf+‰I9ñL§¿½‹)R&9qr"‹}|‹]‹§gGÆË(F±¤2:[DJ!Ix¤$AX’èl½<ËÍâ-xòòÕ‹ÉÃáHP1C{ ƒ$ôÉ*]^ûöcÿ9O6E‘&ë@1Oʤ5?ËÖ‹,_¥åõÃỳß#A±^ê•É§YaZKÆO^¼~n¥º#ŠQL |ŠIPö$3‹E:Kͺ,ŽÁI¬ÚÜŸüú|2„ \'§“ ­:ãÓɘzöD “ö: °0Ɇ£8VzŒr*êÐB—;k'½<ýÃÊ&ã«Àºß¹Y– éWhd5â ƒx¹.æ[±æûåš!©e{ù§e[d°ª(Q B–¹uÿºäPB¯Öà´JY¡Ëb)ePž#¦˜"uj¡âO›Ò+—ª ?”PxK:”±Ú*¬kÝ qå†À©¾Ày¨‚ªWA"„6U’´7x¦VÏÌ„òÌG’a<€]ÓqWä¾2I^dëåé¨x¿Ir3?ö#û–žš"o’eH X´/Kþh¬Y#Ya—ll»Åg-¨RYŒ§Ž›©QXãhßbÒÅQ#Ê+ŸC<Çlá¿åVĈÁÕtKò4)M ;}cÙ¹¶£I˜ÍÍe–—fî{Ókÿõ\Ö¾37¹o¼^¦ÿø–)ï%ò ërŠ1 ù—j>˜)lÊ-çirž­-œŽO(RÂNm Â:P(Vª¢1³ ”‹Ä ½²@¿s8¡H¨T;5ÆÚÉ)™ÿ½)¼ù¶wy™g—<øsHؽ>srs¾ó/¨b¦CˆÕÙ!‹ý:„ØB~j–­.slÇdé¦Ia–é: ¯2Èx¾ ¦ÒUÍü§¡ƒ¢bYýÀ|¿˜ey .³0d¬æ¢< ¼£Ñ;ŒuZXˆ)¤9J.ÒÙ…òεc&=¿(wÃ,iº†!O˜Ü@ü&_ù ¿_-½¹4ë9ìx?ž­ýp¶Ò\4Åߌ‹=¤†—ÕÀ»Lg›e’·B]6Ìô;Áò^×¾°Bpa½t ¨BÔ„³ûf]Í€ɬL³0àT ì\c‹í¤åÅÎ|Û¥Aö[ŒY:+}?¸Ô‰ $‹l¹ô±íÏp‚b¢Q1¢!3I¤(«âŠVÙFG¢‰iWŒ(ˆÇÜÓY›$žŽ@èˆâQƒøq—ôÌáÌëpnŠÅîh²eEj嶜8âº%ïc+ ¢¬‡Õ–Š«:t4Aš²tÙ¦„ÈmM£=¦±Û›Fï™iìKLã=¬niÚ×±ŒßÝ~ä÷Ìiâîö£¸KÓZ ÁQ î5ÀVñðS2[v÷‘=IÊ †kÕáצ óJanð©ú¤V­ ×QƒèALÖ@½WžnÉ#¤3ýjÚø±›W,[¼¶°o©(àIZ°=?;zdÇpD"ª`'—QHú³ÕÑ›w8šÃ$üaéèƒ#]E*F”Ú¾Œ^ýyôÔÞÑ[Ωy 0ÀZ²×? tÓ*¾%^}þF÷ÑAê í  ËE3Ö͈߇!¾ôGýÖqã Wi3$&uJ‘šw8P ¡ÅAy”×ØyMäů7ôû&6¦¤sà]$¶š×ýMlò©‰íGbû¼Ä6ëa·2ÛqG‰fÄb cÚUí¨¦UE,µ·V¦êÛ6Óƒ™»@ÃØÔø/ÜxiéË2,ÔªÜl¸æÚÒŒ­2q<ØéúÜOûK44²<=O}¡Á2Ïæ5‰«¡4hõàÜ–Â`ž-‡ÅÀõzVfyáÉê…&)*)‹Š¿-㤶¼Ø[@±Å—Ö=?7ÅfYÕ58èMCÑa[ƒƒN]|58h™Pð58h´kpÞ!¸]›$Â%ì¤_BiÝ~…'’õÊßOÆ?ðŸÿŒˆÿ>ðÿXÑ°/­ûjѦðo Í÷àà:*¯–J5JÊ7ªgû&Ðz” :“Ò}ê(UÚì}4rŸ~&Ú1´ö„µÖë×NAù$Y.»x¿YM6˶‚ßÅ—ÕCØA/;ýŽ¼ùwáº_ø¨¹ðcºù·ïOh¢}ŹcÞy¥¥ÂΣ£ÊŸ%:Áw>¼¹\ÖË)loÚ/Ýy|Ò\õŠwûcŸö˜È˜ A,m ?;  Î-„ü+¦åø›­ÚÎÌeY;™ û¡Û:‚NLø¿¦5ܯùïï®D…úÕ°‘X#%vÂZAN’]°‰¯탭CëÚ ·'‚!F4FZ‡Ó5“Õ‘þ§÷‰ä endstream endobj 582 0 obj << /Length 2150 /Filter /FlateDecode >> stream xÚ½XÝs㶿¿BÍd¦Rc! ¤Ûô¡Í¹s^šII;¹<Ð"$!G“>’:×Nÿöìb—_2ëÜ݃_DXì.öã· üéúÅ—W‘LV™È•¬®÷«H¦"ËìÊ&‘I´º.V?®¿nÜþŒåë¿û·×›m¬âõŒ¯x\¸²Ëiø}þGŸˆ>_ÐGHbõúêϯ7?]ÿu¥¤YjVÛ(Y“@%¬µq¶Qš •é9E¯’`ÄÇ©‚âƒ6áHÿ ‡‰‚¸¹´„¤‘*cŸYµ O<è“Êä™õÑOꣳ${^}ÌSþ²&åóê?e£¤zf}’§ô‰lÛçÕÇ>å/#ÓˆãGʼnPi @†Ü{ŸW;w¹Ù&ZÊ5˜SdRQŠÈ¹êߺ¼iëŠÙý¶}wÊW\ÒÌS[¿s­/NyÉgÝÃcaÛàÜ/¯R=ÁJÜœ « m~"U¯»£oi´Ë[×Ϲ†‡y?èî7‘\×ô§uoª÷ô½wþpìÚ ü—­ëŠwíë†×nÈk·5Êùí&~¬A]½‘Rû]Çbvu/¾W²tmÛóÉÇdo”\¿ßDñÚ5ù7`8ìêÂ|ŽvÉ‚ÿqMŽšX®óª ©R¬>ŒH}ôÊÁ0(ßCãòÎ5=aÎ[ë“Sý¦D‡HÜl5˜ûúèZ¦ßõÜnxÂmÀÕïñ'/O öŒÑ¾.ËWïÛKŽy )ä•%Cü‘bóèò$êwl6ÏëîY,}çö¾ë|u .÷¾;Òè.oò[&ié?F ~÷¾i»Q$o ‚9êªíšÜWx,œèjú¢Ç–<ùyH]ÊY) ô³åPÀ!&b ³À‰ŽQeaFE4ŒRH•Y“@rñ¥ îGÑ’:Š÷ÊžÖ“ÄJ™a¯”I%ýÞŒŸüùµ£ ÷O?Jf¥Šµú”“¨8ÖƦ yp>lˆeN$Y!BS "4&}Eÿ¯7)D÷MÉ´†fþÀ_J„Šþ!aõ¥KS®c®¥  ŒoÐ]aˆ_ôէñW…ùßt¹g¶9MqZ‘ŒdÝv€yS°œ¦©Yx¿ º`X†[—IعLqïÂD©Œdo(8ÄR±Áä´bVè DøîûÊïQ纹%…ç¬6vÊ ÕB§ø„ýod¤—ÄØP™¬¥kÅn„µï¿yõõ«««XoçÊv!D¶™I (±qDeƒ¢:1ƒ¢Ù}L,’L=VsÆ:†OG-ICdÉÊÀ˜ŠÄ?½» •(ËÔïiø²ñoCUFªPpö/uY@4wÓtL¤;µ,"§ÿ­¿½+=T4å˜AwlœÛ†Jš?Д¯ó]çAZÈ_ÿÄÎt¼–Éú°/Â*M`Àã70¹k\Gù¡eLǤUÌü¨"Hƒ)ËM†--”yÛŽ ÃÌMO…Š3·¦Ðð å^ã'MHº¾BD Iþƒ5‚á”ü9qê}? Ò¹ï•Ð&úßxVÚó¾\Ï\Jî¹½žò‚·–-§äÛŠ@¡šá¡Yà”˜hqqÖ::†&¡/ôœ¤:ž^ySáúÂت{;¨ÓU§×)ÙAcë*LÆì€óÏo, ~0Ò¬&Ô_-ÉÏD$õ‚ýçr3™Î/I´"M’©À/–¦¢c”gä% “ÉûùcäÍbg*ŽâöQbÑhÍD·%‘œ9áí‚T-l¤§ªM:¯™L@èhŠÐñ¢ù£Ð<öûÇn{j*úr ¨•ˆ6jv×”Åì’Š4Ö®d¼¤ä™SlJ:2mãíZWAU×ÚÀÕ)´›à`Esõ` ­Þù<lX€Æœ·õÀðH=S@ëéª6±QÀ\j߉.¤5.þÖU-sÆý…L .ˆ¦Oc8ýÄ1Ø(0j)$ŒˆÇ"‡°g\‰@Ò¯D¦‰„¶Ñ‡Gfü!‘Ï#sµ5öì*>xæW‚F”ã[Ô`êÐÙ쇻S×#éÄ{a‚J7²*,¬4ô¦Ú~°…_ç^Óx^¯ÃYël¨ª†/d8‡2œÈï Ú“×3*Ùf(tg·¯HCDCÃòrß—l/E¬4*ò.'æ-v³Èýÿ¯ ¶çÈá‚V¡)I/h\à ®N4š~w$¢þ@9æQk€gàn×›ºm·ý ŒK@zÖÍÎZ¸[„Ö¿ûPRgW}˜ ^-iöÒp è3{A)‹gÖ?,˜KEBšdŒzjìcªO—BÇ hV8¸¸.ŠVv*™Xò‹‚>×À§ÝÓú,–Ÿ4Ø<Æ|–µ‰ÔÌá=!£GàÎ>x~¸…\_u¦D¦ìܯ÷óç‡(ÜÌ}1>r8š†V·nº|è´(“£1ý%òëî«ÎÈFßD£oFvc0=Ù´5sÊC¨ži7Ú{¡éœYûW-íú¼mânȃúÖ‡ÛcÏ3ðt³íú˨ìK…Ljÿîêªh‰4ÜÚŒ|äd˜²Ì¿àieŒ°ñÇ„]¸Å*°ë.7díé6 }f…uo~ÎøP?£NV¡®Jðp¶¥w#|pRñëQ:{=X­sÆn4”‰3k3-tú…ßóË϶ô=T·áÙ©qïf¯µè¾ß…¤Ÿ¾Ú†Ù`˜Ù[î©ìÉKªŽHÄH €É%ý€Mj$FÜ+¼T ìùýø³ÅOÈó3f¸Ïo}ù0#¹«=Äzõ$wÂòɦð`ÇÓM@øÉ2e<êoxÖâ)ÆNñ,42‘$†¥Ã«ö‹—×/~³Á„ endstream endobj 591 0 obj << /Length 1797 /Filter /FlateDecode >> stream xÚÅXKsÛ6¾çWhr¢’Áûá6=Ô‰gÜ™öÐ:í¡î–(‹ E*$e'—þö.ðŠ;Óê"<»ß>°ØåÏ^Ÿ,gIåìb=IÃfJ„%™]¬f&çeÞæi‘7yy=ÿë⧞-AF·ÿë¾,íÖBP‘4mZ·‹ýÎÍò6«Ó6¯Ê¡{nÓ¼¼{ì¿€»tÜßVef×_Ÿi<Ѩ(A{õ~p2—U¶¾ÄïË|•¯í¸³¡úÃÔ>…Së|™geÛ8>ÕÚëP‚YÓžÄôþy_´VØ»O;û‡¬´Wþ|_Ÿ¬à ]CÙj¿$nã«ÎÐNªáˆpŠ^`„…ä c>_£„]¡šhL4ïN-‚còˆÙ#ó/!VÆhÁ„êCtÁ¤æ2‚ø16Ç¥SÄ°0Œ2æc‰MÂa< Š‹M,ìˆFÔPˆ„Ž¦ZŸû‰pq1.«í"¹lç ðS²I7¸Ê²ÒšÌﵕûÏËU6'"ùä—P7Ø‚Mò]‘gµça£5 H›¦ZæÝå르·ƒ ¬#*ì˜&Ë ì:(†ç¡!×s‚“8Ô@Ë@R¹² ¼J4_0Ì“?æ'™Û^¦¥Û^Ve[§Më–³t¹q£Q7ïtúv“zÚN9Ö£†AQÍÁû·™c¢Àó½ªv;³Û7ÖpYq²ºj»ôðƒdW®Òzå¡Õuv\÷ue1 ‹-k27Ü¥uºÍZç ˜§þ]fsO^Ä ˆÇ y]gí©7S3Ib}¸±vàŽ%9"bŠ†yŽçÛ´øœbœüöð|ܧMngDÀÕìf¿§µ‹ƒÑ£S $?D d-÷s„w§gzÜ6‚b%Ÿœ}©Û°I’ ¦È HB´âÔj-4W"%æ?€~œ2ÑK€Œ$°fÃàPBé€s-5eOàU´I—iÃô 2³x„¦TbÃÔ0‡˜b`0þd^cÈÄ”ȲƒÊ†ƒG$¦LH¨Zho†68’a ‡ÃÿÊÓºxïÚ·ôýÂ÷-åÖr8«³ná÷×ãzqSÙ«f—^ìtn~1úÂÑDPÀAB({Psöœ&¥gÏíÄŠí´ú&l;¨ÏWqn‘ïW B1Û¦Û¼øìx½q»*‡ˆ-½—&Û>ƒÞ÷ë¾|ò?ýÀ;Ï'<úÅ.8¼ÁÅ7vk—/?œú/~1²çw\÷¼o˜6ÍÚ~àa!²{›æq66Í=£6Lax¡>ù±N9œèÙ¸2rì[ 8RÐW¶mGP ½šÔD±I»Ì¹ÔŒí;WcÌRÀÈ·A14ÅœC‡ÊG ¼2i™3BëÑXpiÌW·°GP *Tšá±sà1€¢T Îñ@ô‡”Iþ^„ 5FîBX¡Ã‹!%”ÕŒtes‡iKùìÝųêÅE» endstream endobj 602 0 obj << /Length 2581 /Filter /FlateDecode >> stream xÚ­Ërܸñî¯P¹rÅ30^ÉM6U›¬í$.ï!V9©ÚÝÅÁH,Ï2ɱ¬òíéFƒ$Ha$¯W:ˆx4ú…~bþzñìåkÁÍYÎr#ÍÙÅîLš„å<;K`܈³‹íÙÏ+ÛõÕ¡èíùF§2Y½u¾‘‚óÕ§cÑU0ûõâŸgÉs&Er¶‚åIB'ß÷ý/<᯾Üà‡Á?±>ß$€¥µûêªjêï>7½Åuf÷¶ìaEŸo¯8ãôGÆé4O…ÉdÎü (+Øê,ÍD&FJyÎM.Ì87™Rå3ÊæwP–#eiD"s!ŒNóqž¤\f"2™ÅHYÉLéd¤ Å¥š8ɸâiâ(K¡™Ö‹Ûvöð¡h½AÄ@¾Õ H„Ê´ÊŸúƵ6šótÆ$'ÔJ& LýdÖ2Ù%Ï x¢t yù:gĶD_J2c‰ÈˆRÊœcß¾±µm‹}õ_»OO+¶Ûª¯Î_}Æ–VEUÓÈz¡Ó¿09K³è8ü¿p¡bl¤Læjþ;O³öÄ êf_•Å}ªºNÌ´€:zóûwÿ@}øóÍÖî;¢œ© –mDš‚#Ã7c&͉¹ì™duáÄÎä ;ì{Ùó0J !™‘ú¾ä3úIÓl!8t‚#Pð}µ”`2«¦%(§‰Èþ­m9š7üƒî…¾¼#¨.ç&Ùßâ 4÷)îàD*WeS÷U}eë¹ô‹}q¹·(V*¼XòpÓVõ ×~Ð6ÈÂ-M<°àýñPÓ¢×9Œb:—gzÒç:O:„7û£5µ%ÊN‡HµiíÄ8l×áf7‰@-3L'‹ˆvê!T†·È yfbuqí·:Pæq±ÒÍèzöG¿Y6mk;`f P´Ô7ôµEyM£‡¢x>:ú6»ÞÖ4ܹË-{~ yͱGcJ€-ç°zUÐç`‹îèT§"øÒ-p×·¶¾ê¯çÛE×5eU n¬W—¶¿µ¶žÖ‰àŠz2w ÛyÆ*P *š1Qðm_ÔåÄeDBÎ'e­iî,#Ô—ªcî$d²\ëùñרî´´û½F¶¨}0* , *.&ó•Æ›ovfÀhUFæ«a:÷T¿Z•KÏ€•ü,-ˆ6oƒØHÃëÂÞQî]ÓF&Ä©` pVÂL. ÷U„ÐIÈ“ð˜ÔŠ)‘=Ðà& …vià~°0÷“ì÷1Ær&¸Š“9c¹K!g÷i%,ZЋ=HÄI@NGÉaÅ’+£ä2žGÈÍPA~JþŸÔ[Ï”Ða†ÔL¤˜C†v`ÿ‰!c\J¬ç˜Q>‹~›À”ÿz¡s¸¼ ÊAeõ`'ÇÚëF¯*»ÃR£gªú·¾Xš™‹H¢š'5Ø xn{Éc’™ΘUkЖŠ -ãHÁ•Ð?ElÚf´ =N æ7©R(r@'”8¸­Ë§&]Ý^W5î»:œÑrtϘ“R=Ú-¦D‡‘Æd«}U¤…ݱvIn“-•0L¿ÞÕôÌÕîϹšÂ$!) îKÄ×»›ž¹Û‚¤L²I—û@hŠ£ ‡)m¡n\Ú‚Õ1mÁ¢¯l`5ZMJpÃü·6Q]KŒÙ¡à[»ŽXûŒ*´ô‰øZ…dqqˆ]ªß¡uç*Ñ,M³yÚ¦ ŒK€48Y‰« :ˆøÐÚÙ-ŸÅ …ýdÙOiÏ€ëYòdÈKb!oé!|ñ*/+é Sታ ðœiäa©Ëò‡„Ö,™Z>†§¢ÊQ×ÇÃ¥õd\xúèeÊR)ç×8Õêë‡ ”£>ßFÚ˜ôe_åë#èhÆJ©¨ç%¡ýt,|=…ù¬œj /U¸u(ú¶ú2ß+¬¶€ÍÓÚµÍáÞ©Eÿ=ï‡r"M~kʼRu½ûB©tEB ¾µc˜XÖðäûn7ÞV]yì:×^ Eyμ·åìÒ€|Y`k ·ü¥8ÜP)|à ¦Æv$ƒw$¡±­\kHè³o®è¤KFÔӆź™â4‚Mr¸`¬¼ñ·v h¥:]ý:g€¼µTí¡wì[÷ü‰ë>©Õ±ó+ŽiƒM-mxÞoÊ‘uc|vË”GI˜¾Y M/j€·Ú‚½Wý]¬™ñ¹8|W)†î¦î êi6ùToóïóÌ=Á(•¯ºë{¥9IŒk¾#[ã¡Sq–ɱ>¹½¶E±H™CDSÔ¶è OÇöD¥/>º šÖÉ1pôá\ÌŸ‹ jjLq}8Xì-`i$9UG }…·IÚÇÝ« Âx Ä«CrWÙýÖ/ûVwK³Oбvx=I˜IÄ¢­tmmjVw¶h;ä(3«×UÛõ´ìì ¾]YìÝ0Auv~u Ëþ‘ĬÊÖ’ùÁ¸ ½ÚŽ¯)fÕãþÁuû¸7µöç К‚ºÆÃ^3F’UYx†[{Ó6Ûc9qåIÖÅþ®sF ³x3»¹¼Ý¹»ð†»ù.bDˆfÒ;î_襌ƒuÖnçEï‚ëTÈœz·´òç }Ý5ÿÁíÓÂ=öO­{ƒU/ñyôô²CÜÁS¢"R=âÊ®CòÝñp°íßÚæf¶|(j0µuŒð (“”ÝäOëA'74øž>ÏŸOÍðÌã•zt?M…Sî€{#0è;\ö´æ#]îÚFÐx ih! ­yKá«Aé¸8Z2Œ=Ò„H!ƒ_VµƒÆåÀ„a30a7÷®G6ò¾îóªÐ'‚–JXš?´2f¦F ƒZ$>¶fc°Ätt[¹w  ®TÚ!^ÞÑ"*—Ÿôþõ `H2X÷i_œxzαÄ3ß𠊨‡Ë“N“>KÓíÜWMΡ£)¹zåý‡cå”öB†)øê¼ • &Õâý4aÉô–¿ ­±º)Úâ`Á·ºSµÕ©Úl{x(ÈÌ=ÿj·#u…¾}UP¨ûå>â8[ÄÑ‹EŒð®º+Õþnæ­WÔK†Oz<ä>=Çœ>¡ºÏÛ÷}Ñ.Ôp(I ézag^Ý#7Í—p\Ìéo€y§Æfêc3Ý ?ˆÅ8¢„~Ba‚£ì“à^c%"+»[°üÐ6(ìçµø—§8:ÿqœ»G°|øAí_Ǻvî;ϘÐ軧?âÑ¢cJ2„fÿNMµshÿcpž?÷ÞkOé&À5u’|CXÈ®#~¶‘> stream xÚ½XIo7¾çW>IhÄp_‚4@›Øh (R£9$9Ð% 2‹:K çÐßÞÇ!g‘2¶ƒZ_†"ßûÞö½ÿzóìÅÁra‘T.nv !‘4l¡$AX’ÅÍvñqy]l]ºZ *–ôexÞ'.Ý¢jcS· ;ÿÄgË°ú)<êÒÙ:sy}²{ݤõ',°~&ûdõùæ÷Áˆ†5!È°¼wU²EáÆÛ]xžì¹¯§‡°‘Ø|ãVkF^^­Ö”`¼ü£ôæ__õÆÖSÖÜ’ ÜÆ îIeí ee¥Q–(e ÕL³ƒ*"ü± „ÅMˆµf"¬_Å#êÖDz/®4eÆ›ÓH¨hnŸ¬@ük’ïÁ8V\XT6‹«ÒUæ*ü°ñ™äáy³Ò°y›FYÅ.<ÿZB–J¢—û•èo»:jKQXø(RŒIÇ~§9FPE¼V–E–ÛT¾füúö>-š| žx%(¤ž#Ãùi2nM‹ÂD_%øšìóäÆtRìƒ #¡ ‚Ódcë³Ë(œäµ+í¦NŠ<ƒ¢pr%jWÕ-^8¼KêCÑÔáÀé} ¿YnŠ¼ªK úª \÷0’nÇ•Šµ6áÞ;ÀÍXL$cäAÜ^è·öFüÁ8"5tdy/eö>ß%Õ!+ÂNëLØ:ñd,SÜÖ6ÉÃ%É,Ä&c/ øî¤O¢Ÿ6ßú^ô†Ò&óú°YV›ªjŠk³üàKÓ¡$M§vA0„ÓŸ4U|¿ôfCÃàQÃÃÑì·vom™ºûß\²?€gß÷—"ˆQÖIomm}ø¼‡dsÖK8·Õ`µÝµ+ŠcvJ»À­¡°Þùâ÷’‘9 Xfïò¢¾?ºê´[Ü^yl–¯£µ°cÂÃsi5Ñ>‘0®ü•® cUzºÈlš†¥÷¹ê:¼ç“_®¯ß­coga´<Ò1‚­’4Bëª"¶·o—3e[·[A@,”v4ïêòX½œHàÐ& QAƒ¯OÂ=T9·õlCz?+µ(2?šb¯Öá¹Ï3¿KÕcc¬KÎSS¬“óžOa‚«®„h‡;?aÆí\:ŸÏgSà]žÔ‰M“¶&¬½oò¼m“6fµ-ëµçfÿ+ñ â ¤BèÉ«YÛùç×Ú¿©Yø¶ÈÝÓ#¬Až\¸ÈÖ¡±`q´% 30˜|³‘ê¼è.^.Û×äð†Øßëæ š6È!bóe‰0¤¸û:>}¨™rãG4‹^³J…ÓkæZHÿ{"ú?¢Yöš9ÓF˜ åJ‹¹˜U8ZC4 o%j ¡™!s5ë^³`ŒBXÌ\)Hõì8›>„Ã|a‚˜å†ÎŬq ÃÐbC˜R\ÎÕLzÍD Á¼÷Cœ5¨Â3kCÓ^3V†REGµ¡f”NðêÙ‹Æ õL6)/ñ9MÚÝ›×襵uÞuÿ h="¾›ðà’!LKþXÖ;N¯`\P6ª!¡IÍ<ÅCv™Aø¨p¤„)öXzVL²Ò×`1Ê®”†ª™ˆÙPR3"¥)ÖœAíÌS<¢WZ¨Tc~œR1OñÀ®\b€©S"•šWt¨6˜œQ3ÊaLBCÍÑËz½DE´¦fa"¨šU]€×-‘Böø8ÂÆœÎ,Àšp p>. ª‰`³ôÊA/Ä–(IFxµ†’³ôª!q04Ñg‰£t^ô`¨VÊ©‘rP̘:ÓùÑÌ™A1% b:âidŒÄÙ[2%I _ùÔ ­U0Æ[Byvyóì_?'È2 endstream endobj 615 0 obj << /Length 2056 /Filter /FlateDecode >> stream xÚÍXYãÆ~ß_!1@Á#¦ïÃ@¬í8°qü2;-±G¢M‘ IídòëS}‘Í µãø)za³ºª«ºŽ¯ŠúæþÝï¿Wt£K-ˆØÜ?m¸(…¦)p‰ÞÜW›‡âÛ®ýˆ©l{°ÛᨨÛÑöŸ¶„¦io[…e׆çxŠìf®çËX§î)Ñ_€Üõ!¼·]6M=¾„׃‰üûxΡ;_®cRc¢æ§®i:gËóðõöñþop#”ÝmvX•šêp™?nwxæ|6äññá?ÚWáñq”o“G à°÷e÷eX׃–»@pKÏç-ÀHd&YjÍ7;‚J¡x°#¨e¡Â©pb› ¥Àˆá4žñ0bPW cÁ˜b"P*¤¦yyôJ’d’ +Š%ÃA %4ÑbU’f’”*ΰ@SÍ(Åë’ì.zs’S¡›Hi.^Z+¢$Ï$Áq )Ä'£šk©ÉªN‘Y‹àJ°’p†ÖjURf:UaE&!‚€äª¤Ê$ r`ï)@'áëÖêLkB çz6¼-1¢+‰¼ã%¡Ë$ŠÉ\ÙfüL2çÛÿÉœ¼¹+csvkŽ8‡|¹™Ì)TðƒJ‹PIªõídN’Š"&%…’òãv2'ã xÆs¨¨”X¾™ÌDÅ rº8§Ê’¿™Ì„h5—;Cœ}&™“q\a,3IáÃP¹o&3F ¤g#UŒ0t;™SîrNÓ3PÀµ N÷Œé»ðF¹bA)%ŒÌç ™vã²8‡E$$ÂÁEÒ¬¹ X÷0ÎqaÈ$Jè,ÊœvQ1]xŠH¤‰È<¥•¤è†Ö9¤ÓXfP)”Ã:Öà<›¨Ë ¦Ù ® !ƒ.{_^¦Ï§úpŠ-µïLÕÄFkŽ½-ö¹#ÇýVÁξ‰jÙº¿=™ª»ÚÔÈÛØ“¿³m۲ì5v8rÙR£®ÛcêÝý«Yáp2íq3Âs¨mé=ƒY©[ºçÇ¥hBwŒY®6¢Ô’*ß‘bŸ)™fñ—•Z%Al“1ÍËr¼ìö£©ÛÏŒ—kM#VRhkCæð©rUéuþGo«Ú‡ÅQö¦oìËçeW±í~ñáÚŒnÿÅš>ÒŽ¶íÆ—‹u|_øæý»j­-§Rg 3"b0 7†™ëmt¸?…ªAÅÅôælÁl÷Š‹D ˜ÐG *~ @IDÎ6RÇÁ6O®ê•*ŽYzÅýñdÆ´²AÔÕx,¢X™³Â‰Éy2L´É„Çó©k"O¬ÙV' æù ÂæuH%ö cßÿý‡-û§-ä_ñþ$¦˜?þô[ð¢·p€mG>pܦ·Á-QºÀ¶rs’9ê?¯õ7, –d·ö°èößøð×xREþ<Ó|™a€Û1ø8kn‚IoÚ_À%š€ý&„Zc˜6q øâGkWÒ#hD„§Â´ÿ2€.GÑó Ž›gÇEÂA5㩘ôƒ¦ ö€ñá °áÃïò¨IȯL$¸ÂQb1¶0 Η$þ6Ì7¥€CÚ¡{s¸î¢À6Æ_M«ÍRìŒIÈ3_ˆÊ’¤û #„@Ê¢ "¶ï·Œ;LÖ ‰Ã‡ÎäáÿV%ÔobY6/§Ãë³®9}ëYq:ÌëD.¼îÇÈ$%2^Id2eŽ¾t´)aÂ1&²â@¨‡ð ¨/Ð.‚‹µ‚ájéa×:ítÚê.5“îz<•si¼ $a£cËÒÏ|ðÝñM}iº1k_Ã:âL „ƒˆ‡yZ˜Þ†Å±7˜“L^«z¸4æÅs±Øò`Öxv1ívU Æà®Änú˜ï÷4oÑÙ¹QþµuûŸí!Ùa‘»y£ å^÷]˜ÊX1)ÖPÌ™Gy .'Áœ@ó!Í÷Àø@rn€#±}$>G6ȹ¡®|¼gi³CdHJMxM®ñ¤|tƒ$x…J ^™t˜ðØÇØ9ñsÙGž«ÿ_ÉŸÜ­ÝyáY4¹*Kúо_ª‡®¹žÛôOÓèÝ}ìúÚiJ9C76­í®à²àÅí®ð tYÍ ·1]Ö¾@àéü:ö׃ïMŽðÔwg·âÞ-@p%·#qýº8iHF¯ÖÇaþÚÁ§^ñ½»Ž¯õÌ®9$ŽèBâ2>bý7{Œ ÐPšúßi4Ú× ´°4ÀeÃbe<µÉ…]›&¾kúîXÞãë5À>õrœ9®3XªÈ<r´6@R÷e53±2?ò’Ò0>&TL[^QÊKFÈ&ãÍ€jñ/I‰ÔÔÿ°f™†ïÇl²%xÅ2¸"%Ÿ1m¡R•XéÜ´_£Óé’¹C4R µkÆÁ`©&×ÖkºHɨ\: ³U}î¿æ·ô±’ÓÉò5°Rpœ«ûjÍ(Urþ?ߟüæû‹ßtò+ï¿ÈÍpúé>š<óž|÷§ûwÿ“½à endstream endobj 627 0 obj << /Length 2492 /Filter /FlateDecode >> stream xÚ½]sÛ¸ñ=¿BõôAš q@d®é4Nï:éÌMgzžö!É$Qæ(RGRñ¹ýí]` ” YqÒøÁ—‹ÝÅîb¿ìÛ»ßÿ”g“‚’ÉÉÝz"$‘E6™KJRI'w«Éûi§ëe9K˜H§î¹T5.TÕ—m­z=cbúiFÅ´¬n«ûröñîïßÿD‹•)És>I2IŠ”!“)£ˆ\È7'”Ï')â¼vÔ(0 BS>a”pîH¥ ö$@>ã»õ0FXN#0v%Þ—ìý <þâT‰@)r1a™3îµI3Äë‚æYn­ úÎ' 誈HiŒ š ð®&È¢.Oå$)H:—ˆùÊÅ2Ž ¬Ÿ³=a9™çs•’B|î¡Ü™ò<ÀÊÏϬ£Z)ÀyŸ¥ÅkŽ®ž:ËIJCÁ®ÂÛˆ«½¸Ú“x_²÷+ཱ0A$—ÆH&ßçiü–䩘$àA”æàQs˜ÐgÍÁŸ¼%bཚ%TÎñ!í#MÓsl˜ 9•c>³s°¦·ú.z·ž1 ”.?÷hì]y´ˆCClÏ\ÌøC$EÀ•DÈb’pA á0UˆÑ©R2ð{3K GÝF¤‚k;'Ôߢ»áˆ!ׂ'W&Išeˆù°)[—"ûÏ•MµßÖ¾4ëcä06¥_qˉ˜çAÕ«0cÂãÜF‰ð€ˆ—³Òu©ZŸ´u½*w%üª{,܇Uù!MY­{ÝÔC*M!;CÆÅÔ»Y§œ%Y&a϶©»¾UPXÀ¦1e¾X™ÅB益7ëùTwk½ÁVø¶n›­ßä1¶Íª¬^ÂZä¶,1°n£ª —û®<°A]¤¡gQI8Ô N Ð@ùÑY&ã…G[©^9Ne‹‡^npyAª(“’ˆâ(í¸ƒÓ]Ù.AÑêÞ¾çÆ9,¼*Õ!`%…0ñ|j‘yáFÉ%¨èÀ±ß*—=ê´°†5,<3€€ð,ùtÝ´ìË!ŸÌáT«Ë^—~kœ”V{æØ¿oÑÎn§ê\ƒAÒAeèh|HEúoã3åjU¶‹K§û¶~é< â:OÔG‹9ÿÁ8e1ý›Z€@Æ x-ŠÈQ‚æxÓ!¸Ûßß—=·ÙƒOO™âw#aa(!&½z‚À„ƒ¢Oà å{:HÕÀ;êÞ½êúW÷½^!Há»W®2•¶Kuá™wm³kZs÷”ñïy6íÍ›3çLè~ÛÛ;m€Ö0icu+„è ]7[­ŒîxšMõvW•[pFëð]užDìú°£‚ÉêÑè0â€ØR8ÔõŒB×°Õ&àúÚo1šJa·pè>òú}½…¬ ²âàü"íJw6Uu :§¦jÊÇÊünhã3v~9¢±¼1—yþ ùl÷žŸ•á—gÖG"ærÈØ~F·1€?%ø„øö_|$ŸßáCCä4îÐÌŸ÷Uoç%~ÿdïâ£ÁA®#ò|gÒµ€|½vš7è¯ñq°#ƒ``pôg…¤šÊ±žÀu‰÷IªÒö.DDùç¾®í5±ÚìUÛ'û;1´’ÊX¦#äÉ­[H§ÛžùãF¾âψ¾s¨¿6uqkÊ¡iàN§®Õ¹ó¥DI6DÊÏLC1bŸÝШ* ¸…Œ ”u”ìáû0X8aÛr×B”¬]ä=”Ce´k4”F7ªi}èJÆáã…|Fó迨‡Ó1^ ÜlÔÂý€uð¶Pa¦¢¸ª•»¨Ø¶C¤NŠAÁƪÒõ‘j»Z–u}ÖO•ªÓ®tÅÂfÙ7m-?ç$—Gb©ˆŠÀ¿sëzF^-ÉÎ ½Žh G?éÓÞ2¿Þ[.v|ÐÍGR]ï ž.ûºA·åßÎ2ƒ8 è‘b‹ˆÒNxû‘š[é‚B«'кdP$,%Pˆ_eõñ½8cuñm­~qxuÐسcûºÏµºgvÙêâJ«ó ½/ƒ¶†eP^¥°QºÒqh9t çÓ—$Ÿ_ÓÃäAúꎦÐ>åtÊ›C4 2kY…™8õw›}wfº=8Ž®Ø¿Ø¸?öQdÑÊ2ͯ—G9øº©*ìºHá™ø ,?Ç°ˆÜéå¯o›rm®@Piº*îæ=ùh6Ýœ«Q—°õˆâò”Ú¹ÝoÆ;·ªoõïf¯%ûÞŠûщR/Gå$;Góö2ÍäYDÁ†ÄvÝ6£KÛè…¥óY  ƒüÆ!ïL“PªÊOeeËt¬”ÿh êòß+½=e÷Y¨W¡Úy8\=FÇ%å¡&v»o~1•]þ¸^›‹s3ȤcÔ·Ív%©ñuóNo‘žÀd7AqJu¤¾ÛkÕwÚÙüÿ5(Îið_>€}+%ÇZ,Ý18)ÿaêv™Ã}Ý·v ! ›hBζ„žHÿG¹6Wâhi„¶ÐXùá›m†-÷T›‘œYìkýÛÞíÂh‡tÚrÝ´>í–ªr9Ü|Ú6ýIpÈãÃ]è3¦3°ÑR¸ÚVf.ܺ”ûú « WX¹â»{î#¶˜¶ ™!šµéßÍŒˆñéí£û|Þ\Lçÿ—z×6KåX˜Ó¹pÝ“ð<8¬eÙE[)8$å—G¼ò0¾=?â=äÊÛx¢<ÌMÜsIòq 8It8n½Xòc†ÔýÆσ;ݪEå[¹¶Ù•í!õÄfïòær£ê{§· ‹šr?t‡Ö1MÇ<â‘® Æ½{½ÜC–Ÿñtjæ¿)›ê5~´ç‚§5§[#»‚]2‚..›'ÐìYóÈL^žÀçù0²ê/ Ò9Qu‹ÂþhpœÕâiA«÷fè;êÑ1)ìl¡fزòˆÊ’€®OÃòéÒÝ®R#ÇØ/+½*U= ád _eÿP–îÛP ·HœVFpß@B3…ý×€S{-Åô»ZN;Ѓ•âK(Þ¥{Ç3ÂÒì’a3ÒÎ K°p~ÉÂÐ4´0,L½•è‰•,ÈÄ,7ÊŒ¤Q1G;¸ÜwfôfΨ»–ª=§íÂÞGdú3‡mN8÷þ-!Ê‘3m”ÜËW»©4z'‚cUú?ïبáüËšúäªË£qW[þ¶×­aŒêî &G:J¥¶žÆ'­|_ßÛ ÔîÕ¾ô±mi²s§ìü¨ìLð‚6a¥mœtÿ­rýV"@´ÀcqÛ¾øñîÅÿ,pÄ‚ endstream endobj 635 0 obj << /Length 864 /Filter /FlateDecode >> stream xÚµVMÛ6½ûW‹=ØMÄá÷mÝl6M-b’´kmjÀ¶Yq’òÛ;iIÞq€ Ö9oÞ›Ò øÁÌ¡”ÏŒ1PgwëÉû àŸ £yØŒÀ¸pñbí²ëjò'?‡­<:ÍG^¯æ“‹¯2d¥Íæ÷p¤2#•@¥³ù"{=½Y¾ûP—³\˜º'³\!LÿXUÍ.,U÷aÜ-›hTla²ŸI3-êeÙ| åìíü÷‹²#JÔ €4‹ìØÞª`u$Ìrþ`SÞÈïëjfÍ?‘ý]¹)ëbµü·Œ*n—«å¦,êð¶®åêXy½-êU9C3rESˆ(Fbrô ÐÉ,WR€6AÕ¯³Üp¼“Ï/‹¦^~ +?çaüÔ·»7`àv¹åü½l•œoër±¼kªzŒ¾„!d³= Ã>äñq+‡ó# 2‘™MP‚3à`þËX sb w¤øÕ_×ÇrwûE'vÏWŽì‘£Ë>–åvpÁç‡-ÇÝûº9¶h91îžýtÖ¾½Ž¯øD¾ýFW'Ù÷ÿ#{]}Ü벫ïe—§#5«r_®vÙy[àu»¼ú—±ENx½«Vß:ÞNn‹]§ΞVëmµ)7ÍÙ Ô—á,!Úòý$Ù}°ï}GÏax°*;¦ü€óq#¢áó ’öRš€ÈA(æ°NêT_E”Îj©Q(E•7©ØžF Íw†6ºG¶FKkS¨ëåµU`¿sV§-—²†<¥PÏz”t ÁûÅËŽãR)ÔMBc%ñýz@ií‘’\Ï{ðUÇv=ÊI~È»T‹(VsÊ8ø€Â¶H–U'ûãE´â¤¡vàé@&ùoH*-ìÁ5 ÇMÓ_§šÌý`“ÅPñ šÿ¬S¶ÍP¬<9–¬u²òòS ‘#×à X¯\ ¦6âö…£îD4Ó7x+õÀ¦É¯]’Í 0cµV}W{­¹´É:Ú)"2}ùfÁ•j7ÀÐh%Gj=7N±ù¡k¼&’=jëø&ÙØ4ÀˆO‘µº‡IkŒ— S°¿˜e§¦¨ÅyL|·H$Á ü¥€m,±(]&Ïæ“ÿ¦O‘ endstream endobj 624 0 obj << /Type /XObject /Subtype /Form /FormType 1 /PTEX.FileName (./fig-Effect_plots.pdf) /PTEX.PageNumber 1 /PTEX.InfoDict 639 0 R /BBox [0 0 432 216] /Resources << /ProcSet [ /PDF /Text ] /Font << /F2 640 0 R/F3 641 0 R>> /ExtGState << >>>> /Length 4095 >> stream q Q q 28.34 31.10 180.06 156.56 re W n BT 0.000 0.000 0.000 rg /F2 1 Tf 7.00 0.00 -0.00 7.00 44.97 51.31 Tm (A) Tj ET BT /F2 1 Tf 7.00 0.00 -0.00 7.00 67.68 72.28 Tm (B) Tj ET BT /F2 1 Tf 7.00 0.00 -0.00 7.00 93.33 85.85 Tm (C) Tj ET BT /F2 1 Tf 7.00 0.00 -0.00 7.00 94.07 69.87 Tm (D) Tj ET BT /F2 1 Tf 7.00 0.00 -0.00 7.00 99.71 83.19 Tm (E) Tj ET BT /F2 1 Tf 7.00 0.00 -0.00 7.00 106.20 78.95 Tm (F) Tj ET BT /F2 1 Tf 7.00 0.00 -0.00 7.00 113.86 73.01 Tm (G) Tj ET BT /F2 1 Tf 7.00 0.00 -0.00 7.00 133.51 60.60 Tm (H) Tj ET BT /F2 1 Tf 7.00 0.00 -0.00 7.00 159.70 56.30 Tm (I) Tj ET Q q 0.000 0.000 0.000 RG 0.75 w [] 0 d 1 J 1 j 10.00 M 51.68 31.10 m 185.06 31.10 l S 51.68 31.10 m 51.68 27.65 l S 85.02 31.10 m 85.02 27.65 l S 118.37 31.10 m 118.37 27.65 l S 151.71 31.10 m 151.71 27.65 l S 185.06 31.10 m 185.06 27.65 l S BT 0.000 0.000 0.000 rg /F2 1 Tf 7.00 0.00 -0.00 7.00 47.69 18.66 Tm (-4) Tj ET BT /F2 1 Tf 7.00 0.00 -0.00 7.00 81.03 18.66 Tm (-2) Tj ET BT /F2 1 Tf 7.00 0.00 -0.00 7.00 116.42 18.66 Tm (0) Tj ET BT /F2 1 Tf 7.00 0.00 -0.00 7.00 149.77 18.66 Tm (2) Tj ET BT /F2 1 Tf 7.00 0.00 -0.00 7.00 183.11 18.66 Tm (4) Tj ET 28.34 51.40 m 28.34 167.37 l S 28.34 51.40 m 24.88 51.40 l S 28.34 80.39 m 24.88 80.39 l S 28.34 109.38 m 24.88 109.38 l S 28.34 138.37 m 24.88 138.37 l S 28.34 167.37 m 24.88 167.37 l S BT /F2 1 Tf 0.00 7.00 -7.00 0.00 20.04 47.41 Tm (-4) Tj ET BT /F2 1 Tf 0.00 7.00 -7.00 0.00 20.04 76.40 Tm (-2) Tj ET BT /F2 1 Tf 0.00 7.00 -7.00 0.00 20.04 107.44 Tm (0) Tj ET BT /F2 1 Tf 0.00 7.00 -7.00 0.00 20.04 136.43 Tm (2) Tj ET BT /F2 1 Tf 0.00 7.00 -7.00 0.00 20.04 165.42 Tm (4) Tj ET 28.34 31.10 m 208.40 31.10 l 208.40 187.66 l 28.34 187.66 l 28.34 31.10 l S Q q 0.00 0.00 216.00 216.00 re W n BT 0.000 0.000 0.000 rg /F3 1 Tf 9.00 0.00 -0.00 9.00 93.91 198.60 Tm [(Site Eff) 10 (ects)] TJ ET BT /F2 1 Tf 7.00 0.00 -0.00 7.00 97.36 4.84 Tm (Component 1) Tj ET BT /F2 1 Tf 0.00 7.00 -7.00 0.00 6.22 88.37 Tm (Component 2) Tj ET Q q 244.34 31.10 180.06 156.56 re W n Q q 244.34 31.10 180.06 156.56 re W n BT 0.000 0.000 0.000 rg /F2 1 Tf 7.00 0.00 -0.00 7.00 347.42 123.78 Tm (1) Tj ET BT /F2 1 Tf 7.00 0.00 -0.00 7.00 360.67 127.41 Tm (2) Tj ET BT /F2 1 Tf 7.00 0.00 -0.00 7.00 360.77 126.05 Tm (3) Tj ET BT /F2 1 Tf 7.00 0.00 -0.00 7.00 347.84 119.97 Tm (4) Tj ET BT /F2 1 Tf 7.00 0.00 -0.00 7.00 347.93 116.64 Tm (5) Tj ET BT /F2 1 Tf 7.00 0.00 -0.00 7.00 358.95 113.00 Tm (6) Tj ET BT /F2 1 Tf 7.00 0.00 -0.00 7.00 347.20 111.99 Tm (7) Tj ET BT /F2 1 Tf 7.00 0.00 -0.00 7.00 348.58 106.53 Tm (8) Tj ET BT /F2 1 Tf 7.00 0.00 -0.00 7.00 350.60 106.68 Tm (9) Tj ET BT /F2 1 Tf 7.00 0.00 -0.00 7.00 347.28 103.77 Tm (X) Tj ET Q q 0.000 0.000 0.000 RG 0.75 w [] 0 d 1 J 1 j 10.00 M 267.68 31.10 m 401.06 31.10 l S 267.68 31.10 m 267.68 27.65 l S 301.02 31.10 m 301.02 27.65 l S 334.37 31.10 m 334.37 27.65 l S 367.71 31.10 m 367.71 27.65 l S 401.06 31.10 m 401.06 27.65 l S BT 0.000 0.000 0.000 rg /F2 1 Tf 7.00 0.00 -0.00 7.00 263.69 18.66 Tm (-4) Tj ET BT /F2 1 Tf 7.00 0.00 -0.00 7.00 297.03 18.66 Tm (-2) Tj ET BT /F2 1 Tf 7.00 0.00 -0.00 7.00 332.42 18.66 Tm (0) Tj ET BT /F2 1 Tf 7.00 0.00 -0.00 7.00 365.77 18.66 Tm (2) Tj ET BT /F2 1 Tf 7.00 0.00 -0.00 7.00 399.11 18.66 Tm (4) Tj ET 244.34 51.40 m 244.34 167.37 l S 244.34 51.40 m 240.88 51.40 l S 244.34 80.39 m 240.88 80.39 l S 244.34 109.38 m 240.88 109.38 l S 244.34 138.37 m 240.88 138.37 l S 244.34 167.37 m 240.88 167.37 l S BT /F2 1 Tf 0.00 7.00 -7.00 0.00 236.04 47.41 Tm (-4) Tj ET BT /F2 1 Tf 0.00 7.00 -7.00 0.00 236.04 76.40 Tm (-2) Tj ET BT /F2 1 Tf 0.00 7.00 -7.00 0.00 236.04 107.44 Tm (0) Tj ET BT /F2 1 Tf 0.00 7.00 -7.00 0.00 236.04 136.43 Tm (2) Tj ET BT /F2 1 Tf 0.00 7.00 -7.00 0.00 236.04 165.42 Tm (4) Tj ET 244.34 31.10 m 424.40 31.10 l 424.40 187.66 l 244.34 187.66 l 244.34 31.10 l S Q q 216.00 0.00 216.00 216.00 re W n BT 0.000 0.000 0.000 rg /F3 1 Tf 9.00 0.00 -0.00 9.00 303.42 198.60 Tm [(V) 60 (ariety Eff) 10 (ects)] TJ ET BT /F2 1 Tf 7.00 0.00 -0.00 7.00 313.36 4.84 Tm (Component 1) Tj ET BT /F2 1 Tf 0.00 7.00 -7.00 0.00 222.22 88.37 Tm (Component 2) Tj ET Q endstream endobj 648 0 obj << /Length 1153 /Filter /FlateDecode >> stream xÚíWKoã6¾ûWB21")ÑRú’mS¤· ìÝh‹±XÈ’*ÉŽÝ__²»kw›C 4Î|ó 9äÐóÉícJ½ eŒ0oþì% ±Œz3†QÄ°7Ͻ߂y!:1 )‰ƒ5ï[¹ŒhÀ[«]'ª^ò²Ü›™¾°3_[‰['Y™ñ/|ÑJQå§(‰ð4ÄA–¥JÆh&4 Ûzm¬û‹Mƒ…lʺ7Ê%¯¬ÒZ5mo–"¿QØÏuk É^Ñÿ}þ«J<%!g^ˆ1Ê’Ää}oÌ0MFv34‹S/2wŠ2›™SŸSXLÐ,"ÎþÉZŒWœ!Š™3àU> IÛ)IÔúJÑKXmPást’ %8þ<zÌ'N¨³ÏÎAÆ(K±³øîLN1~<›PBØ‘1ˆFj‘c”¦°ÚJIìjÿ¨¶\å¼àm)öïêÒ ¿Íw â·"÷oŒÂ_”áC­@Îl;ƒúÇv!«„{ëýŽVn–…~…*ÅV”H†Ð7P7#§“ù­Þ¬=˜³1+›â·ª±ÚÉ¥ËÎFlEs@R©ßŒˆT—˜T—¨\¤³+åútÂ؈GÁö×›®9œô‘©ÿ`Ï+ ôy<ì·‘sÞsÿU¾8}W¯›ºR·c`zÁ’œÏq%öb§‹Ç¤HQb‘A婧)Êì"özÒ×wÎÑ܇Oþ¿uGÛjqèü?ú§ûÅ>_>¯ë†çÍG.ÑÉÏó œìÈÃŽÕ9M<ÊÂ3o¹žü9AŒÄQ¢çG¢žr^Vqû´N½ŸêÉ{õsS¡E 'í…àeñ(ÉLZrµFw\z9F+ÐéRAç9ÅjoÍÊE¯“ºíˆ‚S ‰â¡mÁ•OF}‚+ ëPL#ë2Ühêgó-j0~1ƒÞÁ™b]sãûêºÖs­µoEÉ{‘[”Ú†¨„ƒ¨r;M’š™O–*/»ÚI¥ab¡7ÝŒª‚¨:ž=WÌ/…Ðøª ºÖ©½tZ2S ;¥.£V@;×”abo¾Úªô×Mé×u®[¸ÆvHC”•¨DËKù×%KY >v†¤3|˜¦Q`Ý:!Èþy&¹¦–Uí1›éw|ut’¥Ã¶ƒ’7ê=°“êù"à‰¢U¥\Y—Ù¯œÛº7›¦Mw®^`ÔÙ²¨Žãe¬d8Ì4Ê‚n³, žIebæêÄ穪yµË6Ü¢ÞôÆŸ»` Í h¸ZÞvFPú[ØéO„Ð"ä;Ù))6ºá…1Xyj¬ÚzÓX{K<^€¦ $¬µ2^åþjJÛ:ó:ˆÎ> ¾vò¯ü_trö?íäÉuòx²ðK`Êý,Þñ:ö¼ÓFÞRW]o†ÀÙ'ˆ]CÝå~ìý…Qã¡BÞ&*»**Fôª¥üÂdÝ! ( Ü•ªo®dwâžkw´,6½›Ø¡úßëÀ‹+ÁOÞŠb,VJõ7£&`»‡ìß0¡1‰ endstream endobj 643 0 obj << /Type /XObject /Subtype /Form /FormType 1 /PTEX.FileName (./gnmOverview-Biplot1.pdf) /PTEX.PageNumber 1 /PTEX.InfoDict 651 0 R /BBox [0 0 432 432] /Resources << /ProcSet [ /PDF /Text ] /Font << /F2 652 0 R/F3 653 0 R>> /ExtGState << >>/ColorSpace << /sRGB 654 0 R >>>> /Length 758 /Filter /FlateDecode >> stream xœ½VMo1½ï¯˜#Heb¿{ i’6R+¥ 5R”MˆD Ôªÿ¾³ö.ëЄbUê…aöÙã7ÏÏ–p๺„g0…§PkPšÐPhVSø‹j8®ŽÖ_χp»®$ !@dÏõí¢::#p|’°I¤GŠUàQ0¨øµ9ôŽû0~¨NÇqì¿€•EaìPR“'4<³tH>‚O ÀÌ‚â/%PÛþPv6®™4ª>-Úzfæ-•}V¶L˜Ò¥ˆàó°‘h-ÁHe,a¨y#ô©ÏŸ2p.y¨t¤6(,HoÒòðz$ Î2B¢N`*‡ÈëŽTƒUØ°Ô)ÎìTë°åÅ:­EJ=0%3K šûÈòKÝ·`/’âʤ;W +Þxô&‚}ÉÌ ¯Ù²‚\‡Ý BÍa]ë/W-˜ ®1¯“Ñ ¦çèä ÿë ü¬®o@À]%á‚?•Œã®\ì`rÇ9(ÏféšðF;éñþ£”­›±}}MØ壵ˆm¾ »¼b 泧0Ëçål«ëò¹sÿ¹ø¶ŸÕèhWŸ5¯ƒ³‘ÊÁ»ÈÕû¾‹‚CØšÆ[0¯b£qyÑÛ§cÒk4Š©N‘²½‹Tï¤ùÿ6ê²Ò¨úõ6Ý„]žH!‰m¾ ³<k6|vù¦œ6ŸU—/<[í #A³v-8‡6ìiõ^°dÚ„ÚÓë½h’½}»ÙûÁ,}ow{/X±eõZ»›Ui‘ØIû¾ ÓÁrÜÐÿ2l_E:d+ªF™zG™: ­â»”«½Žk½î gOË Ü÷Ù¡·\Á÷ÉêqÚgaAïÜM6“þ Œ/µo¾©x6âúoJËùÓr1]làÕ#÷ Feä@zמ Ý0”™ò¸uYŸ(Ê趒õl3]¿‡ãÁ?ßJš¡ÙàŒK­øÁ=`Ö'«>³Á?fÓͬžLÂ;¸Ú¶á²ú båSÚ endstream endobj 656 0 obj << /Alternate /DeviceRGB /N 3 /Length 2596 /Filter /FlateDecode >> stream xœ–wTSهϽ7½P’Š”ÐkhRH ½H‘.*1 JÀ"6DTpDQ‘¦2(à€£C‘±"Š…Q±ëDÔqp–Id­ß¼yïÍ›ß÷~kŸ½ÏÝgï}ÖºüƒÂLX € ¡Xáçň‹g` ðlàp³³BøF™|ØŒl™ø½º ùû*Ó?ŒÁÿŸ”¹Y"1P˜ŒçòøÙ\É8=Wœ%·Oɘ¶4MÎ0JÎ"Y‚2V“sò,[|ö™e9ó2„<ËsÎâeðäÜ'ã9¾Œ‘`çø¹2¾&cƒtI†@Æoä±|N6(’Ü.æsSdl-c’(2‚-ãyàHÉ_ðÒ/XÌÏËÅÎÌZ.$§ˆ&\S†“‹áÏÏMç‹ÅÌ07#â1Ø™YárfÏüYym²";Ø8980m-m¾(Ô]ü›’÷v–^„îDøÃöW~™ °¦eµÙú‡mi]ëP»ý‡Í`/Š²¾u}qº|^RÄâ,g+«ÜÜ\KŸk)/èïúŸC_|ÏR¾Ýïåaxó“8’t1C^7nfz¦DÄÈÎâpù 柇øþuü$¾ˆ/”ED˦L L–µ[Ȉ™B†@øŸšøÃþ¤Ù¹–‰ÚøЖX¥!@~(* {d+Ðï} ÆGùÍ‹Ñ™˜ûÏ‚þ}W¸LþÈ$ŽcGD2¸QÎìšüZ4 E@ê@èÀ¶À¸àA(ˆq`1à‚D €µ ”‚­`'¨u 4ƒ6ptcà48.Ë`ÜR0ž€)ð Ì@„…ÈR‡t CȲ…XäCP”%CBH@ë R¨ª†ê¡fè[è(tº C· Qhúz#0 ¦ÁZ°l³`O8Ž„ÁÉð28.‚·À•p|î„O×àX ?§€:¢‹0ÂFB‘x$ !«¤i@Ú¤¹ŠH‘§È[EE1PL” Ê…⢖¡V¡6£ªQP¨>ÔUÔ(j õMFk¢ÍÑÎèt,:‹.FW ›Ðè³èô8úƒ¡cŒ1ŽL&³³³ÓŽ9…ÆŒa¦±X¬:ÖëŠ År°bl1¶ {{{;Ž}ƒ#âtp¶8_\¡8áú"ãEy‹.,ÖXœ¾øøÅ%œ%Gщ1‰-‰ï9¡œÎôÒ€¥µK§¸lî.îžoo’ïÊ/çO$¹&•'=JvMÞž<™âžR‘òTÀT ž§ú§Ö¥¾N MÛŸö)=&½=—‘˜qTH¦ û2µ3ó2‡³Ì³Š³¤Ëœ—í\6% 5eCÙ‹²»Å4ÙÏÔ€ÄD²^2šã–S“ó&7:÷Hžrž0o`¹ÙòMË'ò}ó¿^ZÁ]Ñ[ [°¶`t¥çÊúUЪ¥«zWë¯.Z=¾Æo͵„µik(´.,/|¹.f]O‘VÑš¢±õ~ë[‹ŠEÅ76¸l¨ÛˆÚ(Ø8¸iMKx%K­K+Jßoæn¾ø•ÍW•_}Ú’´e°Ì¡lÏVÌVáÖëÛÜ·(W.Ï/Û²½scGÉŽ—;—ì¼PaWQ·‹°K²KZ\Ù]ePµµê}uJõHWM{­fí¦Ú×»y»¯ìñØÓV§UWZ÷n¯`ïÍz¿úΣ†Š}˜}9û6F7öÍúº¹I£©´éÃ~á~éˆ}ÍŽÍÍ-š-e­p«¤uò`ÂÁËßxÓÝÆl«o§·—‡$‡›øíõÃA‡{°Ž´}gø]mµ£¤ê\Þ9Õ•Ò%íŽë>x´·Ç¥§ã{Ëï÷Ó=Vs\åx٠‰¢ŸN柜>•uêééäÓc½Kz=s­/¼oðlÐÙóç|Ïé÷ì?yÞõü± ÎŽ^d]ìºäp©sÀ~ ãû:;‡‡º/;]îž7|âŠû•ÓW½¯ž»píÒÈü‘áëQ×oÞH¸!½É»ùèVú­ç·snÏÜYs}·äžÒ½Šûš÷~4ý±]ê =>ê=:ð`Áƒ;cܱ'?eÿô~¼è!ùaÅ„ÎDó#ÛGÇ&}'/?^øxüIÖ“™§Å?+ÿ\ûÌäÙw¿xü20;5þ\ôüÓ¯›_¨¿ØÿÒîeïtØôýW¯f^—¼Qsà-ëmÿ»˜w3¹ï±ï+?˜~èùôñOŸ~÷„óû endstream endobj 544 0 obj << /Type /ObjStm /N 100 /First 903 /Length 2579 /Filter /FlateDecode >> stream xÚåZMs7½óWàƒn `Ë•*[ŠíTí&)ÛµqV¥MŽ®%ÒáGÊûï÷õp(’¢$S=9ìAšá xht¿~øˆ¾g¢÷†bÀ5Î×h ®øk2É¡\ “q›œ Wo G¼Ï¦HéÅ ¹¤Š!J¨)ˆ!fT³!P6:´EzC†D[lPÕEÀ(œqN´L4ÌŽ{1Ša°Qœá€+&ÑJ²aÑ … 'm]Øp¡Ð`÷íD ÆSÖÞDã½â1>ˆëÅäŒE_%ã% BôÌ'ÔnJžÄ&¸Œ “‡a´ÑLðLŠDÁÔœM(­«!l3&|A0TdÂ㌺XM\󢞜´ÿz“Êhí™Ð–MfÑJKÆãŒP° ×1aIFÔH(hDŠ6QŒdÆ?çã€q%‡'71 f‡-ˆ§‰aqbRpZN !.›$0‰nÊN+,¸Áè‹IE¿Üì`uÂfr¹'äqƒ¡8VfôR(™¢Þd“cÑÏ‹É 6t g|.L¦8tW˜M!¯¯ÔçRÓ]S‚Ö ,Ak†û¢qࣔäñ†³ jÜ ž«Â6pQçÑt…c Å«Ûzí£‚S7•Æ•ƒ zõS§fuT—PFÔSŽ‰:0¹ûÁØzǸCËDµ>|AMg1HDAß6±¡ñ  $pïçÏ{Õûÿ~®MõKÿ²îU'“ñ¼Ïgp ‡ø|Û«ÞÖ³Éb:¨gM°5þYGý—“/æLËÀ,Ãs ›Ëy5MQ…zײø‹ñx‚ÏšhÇ“&Ú—×Ø^¥½¦æzÞûþû dÍ÷½êÝâã¼ùýÑøS¯z9™ëiƒ€Î«7ÕÕ ~qçŠy07gÞ±Å3‰mâ1 waN6ø„R/Ìóç¦zgª×“÷Sšï£ymÿÕ?3Àp)YxÉG)6ãÚÀØ‚¡·J`ì‚ÕP"!àOLÉÂ¥îòæb2™£¹Úú#Z„œ ›á¨>GŒLÔµÉI‡C“‹•o€¢fˆrjÎb“ÒÞšêÃoÿ“Z¥¾”Ø:ÄÞxquu~oYiÊrtÖ#æ,›È:ü&u$Á­²¯¼ ÞW wö´üê8yJÚ_ oFW—?ûØ»ÕõÉeD¢Âê—éd𮆕À §¯Lõ¾þ2¿š;¤‘wH#È!¤±Ñûö³êFï×AïA2ʵmÁœW½‡]âªÜ½&;Ô±ìØ"¦'h\Ž•*˜å•ÛkK¬±%Ö¶W±í÷ùqÎi1â½ÍÐ^¢EËÑ“"ß-³ÅǾMÆÖ“B¤šB,“%å°ÈVÓõŒÜ…pL–4ˬ€¤‚•î(#Ä K±åÅ$›UV;ëwh¤— ¡µÂb4Xç@(y«ZK‡ˆÁ•4Π$cÈ:Tà:YXIÎBD 1æÇâ­ °I[´·Aˆ‡²žì²ž<•õ¤e;iÙNZ¶“–í¤e;‘cÊGnU Ô*Fº £Wtn º+ HÁà­d ¥{ p[ÎDÈ7¥:Ì%­ÎQY;„AÊõpªöáôÑxÆ웃eÕUœmÁìósëPÇBÐÙŒ9ñ ˆŒìÁ¡ü"Çä—-õ€Ûà¡C©'óõd÷(êÉá6õ¤–zRK=©¥žÔROj©'µ3Ø”Úkn¯å¨”„™Z‚Ù¼ 0ͨłŒWæb‡þV¼%Læ[žÀ”¤ã ¸eRñ•CE0ˆ+ª £N¹ïòaT—câ€=ÔåW8!Môã`*6®axý©K’_ƒñÃtôé˜8«z™1Þºyà,ørÿ¸ŸôóãFRŒ´âT n8ˆ:UqOÔ*›~tÀºÑž#nG«èÈÓ<§´š¤´š¤´š¤´Õ…îÆqQH=Éë¶]tVÖr2b=èê²ø¯­ %KÇ\ŒñÖkÛ!ÂÍŒYÜ䟇tÉôÒÑq¸unp@ŸGòÝiBBƒËa~ ;Gbu{5 JîÁ   5„:Á±í¬­=¸ÕMYLhIQ90†|ÍS£·ØGwC÷MÛe—éO<¼ßÓÃe™ÀÔº#­’!îIkdăfgDºŽ6W¢]þ{bæÔR…¤Jsy=n%h|Ý’F ¢Æö¶@9%ĸÝ…‹ú­qèÄö·Çq‹÷½Š{]þÁ$îá"ë¾1(ønõ‹þ]¤¯.Oî[-j[õé‰Û¾KþßÝì~ØŸI¶Ê¶[‡ÎYçöÛ{`oð!ͳÉ(G@œvŒÈñ‰@màSn¯K¡£g:–Wj¯Ü^ý1 "Dôê‰o³ÐÑ}¨iãîÌ_ëá0…£ç|ÁLOÏa ¬#ˆÏ”bw0²³EÖ0²0ë)š¯áxÝÿ8=愃(Ú‚tsI·9‹Ô5M ºu¿¢ $sî~dš Â@šS© Û)$gd¨‰Úz î‚Ë6ËÝ0VZìöšµ°<퀪1.wª¢-éóFŠÍlöΉ¥&”ã-¿›‚ü£R´g¹ÚVN¦u“™OûóÚ|wúwv„tB Žè}~†Z'Ã{ß½ͯðæ­y=íþ}4˜™ŸóÏ‹ù³¦oÃÅ žê[(~Èígmc}öl»—:8n¤©WýÔ¿Öw êÏêf«7õÕŸõ|4è÷ªƃÉp4¾lØmti*ýUþíåäjøµzWo—5Ü”­~_Œg£õëÓÑÅE “ëa’\]Æ‹™)Rý±˜Ìë«úb®ÇªáFœÍFÕå´ÿg]õ‹y] FÓÁâúâªþRÍGWúºî¦“qõqZ£ >éÍj8B³Ñ¬²Äa}QMÑv5€'\]õoþ¾_ö§‹ë«þb^M.'ãúS5èk}³ÏýA½¥1ü#ôÈfYòc¯›¦f$¿}‚oÿc8~þøe­ãÇk0‡'…Ñîq%yÜq¥Umˆ¼´g¬ôøå–ȸ#Ô ˜Éq¦²jÛïö 5ÌT¬û†‘¶­ãGÚºÞ³êÇ“ýdˆ§ÍœÿŸDßS×y‘Ó£ž—NÈíº3¦"¿%Û”¹Cµ“Ô1+ 3.¡ ·­©¼Ðb[X›ºf¯\÷0S¹{íébt¹˜Ö¶ÜVûE¼]–£®‘=÷o9Äuáÿpæù endstream endobj 663 0 obj << /Length 1322 /Filter /FlateDecode >> stream xÚíXÍoÛ6¿ç¯Ð‚lÔbø!RT» X×fè€6ØaÝA–d‡…,:’ÜÔ—ýí{)[R˜EÅŠ`–L‰?¾Ç÷ñ{ä 0ü“€DqpLPÄXí.î. æ`0즈[æ^\½Û%Á}ñ ü÷S¡d¾¾¹¸º–,HP"¨n6!%± "‘˜7yðûâZmu± )Ç‹äå2d/^«}©[ûn£k;X§uY, _ísž¶éÊ›[½¤|q¯ª­}‘î÷µþ¤vië$§¥ÚV»¢rBïU{kGµn”;Ô'#¿hÐò›ŸÀ˜Q F1ŠpÄívonU㔶ۢiÝS{›¶ýÈélT[¸ÙLJ§cífëb_ ì¨W¾vvíµªz©i©O&ÙŸRUÅjjD¦C]çªJB³uŠ$“$)A ‘vëï1Çföêš0> C ¸nARÂŤÁpDDEIdqÊ'‰Bìã`€zµ ‰ˆPᄲP‰b¤âj¸½‚0äI;^z$…”rLÆnI«|¸&™ÒZ­êè Ž®Üºû%ä¬>ÇÇÍï‹z_T¹Êeêr¸Õž9/ì6˜!0q~ðâsðhìñs„8<{üœˆ$dÌô§Oš®:rʨ³ Q Õ½óÆ^ÂO]04ò€KK:pÄÙ @ÁÛ±\l<ÊA,æg Çúr‘öˆW"œôö˜2[„ðØ3’°õzŠ¡g°Åñô+ó ?IFáüàÑù€ú½^H㘢„R—Ïñ´Žfº†B¸×UÞ§Žó¿¯ã|F—ÏXÇÉ?¯ãô™ëø¹–".‘„ŒÆóy-¼ÑÌ›\Ó<íocå;3b”ŒE}ðÞƤJš¢‹QD];Ðßö‹»ƒ2%ßÞÅJ¸'—ÇÇ*w º nsþ…Ï%lüŸ&ìÿ¯a?‡aOœ±¯ü<„+ži‰‘êxxºBåú°.û.:ÏUëhiåÁµÊµ<µÞëfÚÑþ˜®¡™r“LïÒ¸Ùvÿ·¥Ä½°,­z¦§Žº¿+T¥ÍôrØô=X‡Á,x¢Îß-Cn¶ßuwG´†ž¹>Úwß„öw“f­®Í6wi›ÝšA÷éáøµ[µ²¸ÌÌÐ[ cŠ{Yén‘}øÖþà. º¡SWÝÍëÖ ¸„3Óeº.Êf$£Sv ÷äöÒiºìT_³ì$PŒŒµü¾‹ÝÇblì¶ÚÉÎŽ¼c^ØŸŸek¦a]ùœhðˆYÇƱ;UGfÝy^ÔëC]9™æ£Îa]?hb°GwxcW‰i©Ó'x¶ò롪ºÂHnÚ´nÃÃÞ>Yu×I4=¹t—ªêá2ÿŸOØ]&N¬ùAïöàb£”EÜ2’Eb‘ݪ°¹;¤µá™™KZÕ´*kÀ‹L$àV‹lŠ¢_k¾N¤ô¼6/òÚ¼ud3`ÿXõ à|î>®™ÙÛ¯u#=e‘n|>X—º#Šá°ª2•Uæ>€P`¬QFN4jñQw¦_¼½¹ø Vßø6 endstream endobj 645 0 obj << /Type /XObject /Subtype /Form /FormType 1 /PTEX.FileName (./gnmOverview-Biplot2.pdf) /PTEX.PageNumber 1 /PTEX.InfoDict 665 0 R /BBox [0 0 432 432] /Resources << /ProcSet [ /PDF /Text ] /Font << /F2 666 0 R/F3 667 0 R/F4 668 0 R>> /ExtGState << >>/ColorSpace << /sRGB 669 0 R >>>> /Length 866 /Filter /FlateDecode >> stream xœ½V[oS1 ~?¿Â«DMbçÊ×I Á*4í¡@Ö ÖŠË¿ÇIÎéÉÆV!ñÐÓº>ŸãØ_>GÃ1h8‡oÝø6¢2à6„ž€bDKp½€wpÙκ‡ë·ÏáãºÓ¨”U=×/»‡ÏHÎ>ƒ&ìåQlŽ5)‹,¯­ààñfçÝÓYŽý0;TVÀ5eðáþ` „VVÖ)dðQXªÀòÅ Ëà' `ïòžÉ ðÓ0¡K+KÝJÚÏÀN €ŒA­2øyØjtNbJÚ/ÀDSÝCéóË \“FïKm,*:X¤RïŸÂ9©„Òh ˜ZÀ1WBx' J`n[¡:å•=g°i;Ù¬Ò:¤ÒÛ²²Æh¤B¿Ò}ת0žQÞùpdda¼ l‡–•ZÙ³ù Ž ¼S„F$ÂùA_Þ`¸^¼ŽNn0°É"´ÏQäž™æ3M™ Ë]Ê'­“\O—_/®6ðy" W×ða~}±˜±àà|šoæ“3˜ï+ßrS "Äé˜oJW«¯W—‹Ë Ü9r難Î5ÐÁd C•(ÿ‡[gÔi¢°5C&ëåf±~§ÿ|+éC‹ÀY_Zñ]z UŸ_O¤òc¹Ø,ÓbzÀûmþiÉ]6Ú’†h½ È ç„* (×–ôÕr}öe ÑPre±Ê:(bË éâúG`Ž§NS*]OÂ+K¥=W+‘5ÈÏ}œ7;ÕXnÏë~ŸÎ.×7ø¸;‚ãtj4ÙDîáËoºßÊ«¥£ endstream endobj 671 0 obj << /Alternate /DeviceRGB /N 3 /Length 2596 /Filter /FlateDecode >> stream xœ–wTSهϽ7½P’Š”ÐkhRH ½H‘.*1 JÀ"6DTpDQ‘¦2(à€£C‘±"Š…Q±ëDÔqp–Id­ß¼yïÍ›ß÷~kŸ½ÏÝgï}ÖºüƒÂLX € ¡Xáçň‹g` ðlàp³³BøF™|ØŒl™ø½º ùû*Ó?ŒÁÿŸ”¹Y"1P˜ŒçòøÙ\É8=Wœ%·Oɘ¶4MÎ0JÎ"Y‚2V“sò,[|ö™e9ó2„<ËsÎâeðäÜ'ã9¾Œ‘`çø¹2¾&cƒtI†@Æoä±|N6(’Ü.æsSdl-c’(2‚-ãyàHÉ_ðÒ/XÌÏËÅÎÌZ.$§ˆ&\S†“‹áÏÏMç‹ÅÌ07#â1Ø™YárfÏüYym²";Ø8980m-m¾(Ô]ü›’÷v–^„îDøÃöW~™ °¦eµÙú‡mi]ëP»ý‡Í`/Š²¾u}qº|^RÄâ,g+«ÜÜ\KŸk)/èïúŸC_|ÏR¾Ýïåaxó“8’t1C^7nfz¦DÄÈÎâpù 柇øþuü$¾ˆ/”ED˦L L–µ[Ȉ™B†@øŸšøÃþ¤Ù¹–‰ÚøЖX¥!@~(* {d+Ðï} ÆGùÍ‹Ñ™˜ûÏ‚þ}W¸LþÈ$ŽcGD2¸QÎìšüZ4 E@ê@èÀ¶À¸àA(ˆq`1à‚D €µ ”‚­`'¨u 4ƒ6ptcà48.Ë`ÜR0ž€)ð Ì@„…ÈR‡t CȲ…XäCP”%CBH@ë R¨ª†ê¡fè[è(tº C· Qhúz#0 ¦ÁZ°l³`O8Ž„ÁÉð28.‚·À•p|î„O×àX ?§€:¢‹0ÂFB‘x$ !«¤i@Ú¤¹ŠH‘§È[EE1PL” Ê…⢖¡V¡6£ªQP¨>ÔUÔ(j õMFk¢ÍÑÎèt,:‹.FW ›Ðè³èô8úƒ¡cŒ1ŽL&³³³ÓŽ9…ÆŒa¦±X¬:ÖëŠ År°bl1¶ {{{;Ž}ƒ#âtp¶8_\¡8áú"ãEy‹.,ÖXœ¾øøÅ%œ%Gщ1‰-‰ï9¡œÎôÒ€¥µK§¸lî.îžoo’ïÊ/çO$¹&•'=JvMÞž<™âžR‘òTÀT ž§ú§Ö¥¾N MÛŸö)=&½=—‘˜qTH¦ û2µ3ó2‡³Ì³Š³¤Ëœ—í\6% 5eCÙ‹²»Å4ÙÏÔ€ÄD²^2šã–S“ó&7:÷Hžrž0o`¹ÙòMË'ò}ó¿^ZÁ]Ñ[ [°¶`t¥çÊúUЪ¥«zWë¯.Z=¾Æo͵„µik(´.,/|¹.f]O‘VÑš¢±õ~ë[‹ŠEÅ76¸l¨ÛˆÚ(Ø8¸iMKx%K­K+Jßoæn¾ø•ÍW•_}Ú’´e°Ì¡lÏVÌVáÖëÛÜ·(W.Ï/Û²½scGÉŽ—;—ì¼PaWQ·‹°K²KZ\Ù]ePµµê}uJõHWM{­fí¦Ú×»y»¯ìñØÓV§UWZ÷n¯`ïÍz¿úΣ†Š}˜}9û6F7öÍúº¹I£©´éÃ~á~éˆ}ÍŽÍÍ-š-e­p«¤uò`ÂÁËßxÓÝÆl«o§·—‡$‡›øíõÃA‡{°Ž´}gø]mµ£¤ê\Þ9Õ•Ò%íŽë>x´·Ç¥§ã{Ëï÷Ó=Vs\åx٠‰¢ŸN柜>•uêééäÓc½Kz=s­/¼oðlÐÙóç|Ïé÷ì?yÞõü± ÎŽ^d]ìºäp©sÀ~ ãû:;‡‡º/;]îž7|âŠû•ÓW½¯ž»píÒÈü‘áëQ×oÞH¸!½É»ùèVú­ç·snÏÜYs}·äžÒ½Šûš÷~4ý±]ê =>ê=:ð`Áƒ;cܱ'?eÿô~¼è!ùaÅ„ÎDó#ÛGÇ&}'/?^øxüIÖ“™§Å?+ÿ\ûÌäÙw¿xü20;5þ\ôüÓ¯›_¨¿ØÿÒîeïtØôýW¯f^—¼Qsà-ëmÿ»˜w3¹ï±ï+?˜~èùôñOŸ~÷„óû endstream endobj 678 0 obj << /Length 2861 /Filter /FlateDecode >> stream xÚ½koÛÈñ»……„D{Ü'É»¦ÀÝá\ï®Hm‘¤E®mÖ©T÷×wfÒ’ZÛRëÔ¬åîpfvvÞÜ./¾y“%³œäŠ©ÙåÕL*¢r>K%‰¢³Ëjö~þ§ÅR29_ÕÛu;üÖV?ÞÔï>Ù¹?.ío¿Û|HdÒé¾®vźLJ=¼^¿²PÃýVÛÑkûób«‹®o›OÿÉðÿâãåŸgÉlI)É¥1Pµ»ÕZ_ÊÀ¾êÏúyx(GˆF–1!½aRÓËÅ’Ê4ØŽaï÷ÕñÜÇ1êuãùüæ MTp€@Œ1šZbïéG <Ïs")OSn±RIógqätf7Ë<Ž À3‹#% q$ÉüÝ »Í溵bd€wƒœÙáÕ0·›ß­‡ºi7uáVí›ý¶mzm©f|Ì8͈r4/o=Oé¼’ž>;r8¼[Ðd^ôöaÛµÛ¶×>%óÕ½ý¾©4®}³£y&ŒìÌÌr‹ƒ¶@û6—Å Tίۮ. G­*†‚,–*eóŸ;U;Ú…cu«ËÚƒ—Eï8n¯bú4àYÆ'rÊÄ|Ý^×ýP—nÙÙË^c!înêòÆ.ïÑ”-rüO èjÜ€›v²ÎC%a\™pàÉðò!¡*r"aƒ”µn†Þâ,º=òͦu< ­[\¯Ýb DíÞ\-X2ß î Ïz2vC'«%Œ¤‚ §èû<Ô†&hÊf‹¿ÃÄ Ã ü:qÜÛ§•î´vï±sýΈÜƒ“HH„*EhF­y8MÝîáÉSž!œ 9BQ"ra¡à8!#<Ã;@½ŽœXs’îT8Í1lÙ¢2;ÀeÈcJ¤ÁÑŠ=ÒƒÒxwàÚD¶Á‰ÈÒ½D6 I¼X^Æ8YÉ€¡$ U#v5š‡¢àDÂzìnŒM©*漉Ã&åˆy›Y`_?JG´Ž<sfÝóÅO—Ÿ.(l?™ÑŠ`B% GrVn.ÞLf¬)e³;¹™åŒä0XÏÞ]üõâ Ó”Šà#œ9õy጖— 6ëvù‹5±„Ð’çváv¡ 0ºW¦×M°såhD4 fÐ\0çL|ý'êïmTs–†ø¿è۸§”‡ôþO ÌƾèQcÕIsÂꨨ#twÑ2â YJxN½°Îs#~R’|…>šS58À¿*VõºîýJ1Ø%CçO7Ì“ ö¸^7U‘á³Í¶"~>àoý6ê¨V'€ã-ˆ)÷¬B‘Lýé+xy–{€¢©¢!ˆä{¿Dpä ³lÊÙ^jÍn³Ò„`;eB!ä8<ó_ë[L¯´Õ©™´Ç) çbœÂàÄ!…1Ë.…áRÞåuÈ-aõ!IÆeœ(Ü|½Ù®8œåf?²ƒ=j nmU ÒlÕÌ4fzTÓô¬¸^@æP7ýàRŒ¦üÝB$óWvF“kbÞ¤dŽ¹JJRå|)ò`Mv䶩ûàø“øüáC"’È9chMgK°'ðZ‹XÓ$ì0pr‹h®L™Ž#i:83rõf¥lwXÌF‚k‚rÞŸæ?¢G)Á…FÄ?1–ˆ§*;tWâ؆FLè,~FÞƒš,œ¥Om5ÏdÈñw +•Î¿ÅeRÿó],õKH*Õ¹¬Â!G³©B™Ÿ9Oðñuå:NM:³Šµ š]»Ã‚ƒl¤0ì.5^Iúݦn®-\aÞ¶uošFQA&ÓÕ®s­«ƒ)ìF¨±ÚT„å.Áƒš&sÈÎU´2šˆJ>Ò4L´DÀ¿ںìðh+fÖö\Ž,Ší”=Á xÒIp£ršÇ#ÕßÞåÕWÜä>Žƒ¿D8P¦´Œ°0Ví£Ú&¦¾‚d‡´y{f56´ì+ {w'&_çšUÞ{ÓmcÒzÔ”f¢ÞC;ø5£évˆf`«†Ê851kµ,Kž.Ä0‘¨çtgÀ ˆ„Ú&3˨oIèr0± xs†kV]¿»‚º‰¶, [LEÍÅÔ§ñ–`Å|õ:"š2ç‰e[h&BZÛÓú_e4ä, qaõê…fŸ-0êCÙ×ÛS~f¾6…‰°$ÄŸ0àq¬´o§+‰±jA͹)†}øÜþµëèb=9 ÆÓÛ5÷ùâªõÍãýš¥kJ;ÅÀ½R÷I%›–vc%°¥Ýó(©Àƒ¹pô€¥Š°<ÿ ôHÙ@O-e;ÍÓ½±|":òø^B8ÏÒÌÕðj$÷ñ1O’6Ê£§œ'Ù“9¢ JžÔ{¤ÏVªÓšùÏÞ -c¥?äÈ€ ìWeβþG&¹1@æ?€t±2 ò)¨|ð ½Õ%#ÜßîbÅ}˜@'–b]¼DdL†-Tؘ÷€0øÀovu_4¥†±°³ãŒù\òô¿UɧÃX£›~w`Ò´WCvÇ"LTŸšbh“§>vRlÅn»¶Ú•Æ9SæBnQëû…’¦©F3S³áj§?íêNWÄeÛ  AÇÞ£·ÝŠƒ*é?à‡C'ëfóf IUÞ¶(o‹kd“a%Q®w•ù Ov§0€tÙ5šááj×”¾˜\Ÿ ®Ç¬¿l! þÉIŽ8rÈæä>Ï6Õ8˜”½cfŽ>OÛ7öÅ9<®TfyBöCÁ b¬²m¶ƒ_Îæ?ÜÛb¼ÒWæ;²)ÏóIÂÖÛwïêáƾ°ÿ ³öËëz§TQv­ÙÀÙOĆì䛵6J¦nóí„¡½Ö£lKšÏø¯­«X;Ti×T`VHÐ(œïÝ·Þ ŒùT ¤WSl<#MaáøàUç¾W»¨ÔÛ¢nb&&pèm· õz°ƒ«®ÝØÑáò>// ' ¦%”sFjµÖq†víÔFØÖ·™-ÿ _Hj0B^=DÝì ,*ÏÇßßmº Ò÷\SÏ5æËMÕeoŸ,M°þÆ=ûÛø`µª¬ˆéüÓ®h†úêÞæ_Âçf07â>WÀ¬+ pÝnÉ w§y¿Ó†¦u·)Ç^×VküÒÑ”¿-2üêao¬L­0L¤½‚2N%ûo#*‚„à‰/I¯ueŽ÷‰ËGýnàæn•S0w©é uWšÜ&÷ø·Ê]å“Õ‡nQìONraOØ“@di–Z]rªä^ ÷CÝ(êî&¿æíÛƒ9°‡ßCꛢ»…˜QoàØ?ë ¨Fìþ £ ª2Zû I§×µŽ^Þ‘wÏ`Äcac9™þ 8Äù8Žu;T cÐié¯-švx]ï(2F”úêЋ'LâFUø.’:(ðä" äˆ ¯<0’p—áå?—ýÎìÇ endstream endobj 683 0 obj << /Length 896 /Filter /FlateDecode >> stream xÚµVËn1Ý÷+¢®Ñ1~ŒS (•@ U( ÊbHœÔê<Â̤ ßÎuìiêÔIZh³qì¹>÷øø>ü~|ðú”`1ÈP&¨Œg!áé@ ‚° ƒñtðcxCF §|xCýÈF‰ÊàÏ"7•[2S7NêeÕ~Ž?‚Q  !(ãÜ!ÁÇF•b<¼®›V¯×¨H¼‚Ã!Ýd™µy¹‰CápØvÂ%àf~Ñ!S.šúJ—÷vévT+RYOu“wú©°|Ù2o.õt*¡Ší¨ŒÀ8©ËE¡m£ £g©…|}ªØ½å|)<„Û ,¯ÜŸªÁîk79ǘvîowá-ÛN7ºîn~nE+¼M}gÛú¯Ó¼Ë=R!0£þ¬ïÜÔ•>1 Okê*/ÜÚ›Äóª<‡º@^­üqÃ:Ø_¹áë²è¬©ýpä3ƒ}ÂôSæä(v `A(†;KSAp¸-oûôòfyiŠÛàÛá¢6m[W‡ÞÄJüÊ'—gÀñK]Í-'S˜öWö©2É Óšj£ûmYUöÓ ¿íò¦K– Oµ³± ¶íÝZ® Ƚm÷±¤=û‹ßõê r×]ï?û‡¼(ŽcÌ}dÌê¦\¡Ä/-«Éö°°D)†'¡HÏ) ŽG§ DòÕz63%¦õ2Ï<¥ÊæmÛ9I¦„vèhXÕgkÁe=B@’Êù– ¥ðááþ‡uÙÚæ^‘P©ÇS Î âEK²ÝÛ@¨õ"v3Œã {<ˆE’º’ŸŠ¤ˆ(¶Y†­q`ë ¡,g¾SŽ8ÍB´>ÑÂY ®;¯î•aðÂbÄ¥/á÷.X D"š©6±¹›Jï8cà˜…’ü»ãx”…NÖ4dÆxŸ2+*}¦óo Ψ b;sLÿ&é¸$gž†”H1úL×K…ÐÚT®ÔqÈ2ĤŒ+q¯(£G¸q•‘à ¼Ž~œàþ 1ØßPÔ'Ýî t*£>Åc}n9Ÿl—OŘêÃ=C”ð Eœè+“W}> stream xÚåXYoã6~ß_aû`·1ËC¢Ä´[ çC±-E€Øíƒ"Ñ®$gãíoï Yr”d·èñÐ'QçàÇáÌ|úüòÙG_§b¡ˆ’\..7‹X©Ä"‘ŒPÉ—Åâõ²ÛéÜl¦Þ®ÖB¦ËþZ¯~ºü$éH%X´ NÆ~ÉX9‹ £Ã’ÍŠÑe–÷MÔ¶Í~{=¯:’<ÈéÒT¦Îz=c!¦DEqXÙ5AsÖ]·ÆçìP"iú[ð.(9Z‘ˆ‹°à ebÎM¢’ÁŽÎûÎ9•µÞ;Ýõ¦‚îµjÜ5kјDldQÎXL *¬É®ûòàÖÅÑ:.Ö"…Õi²X3FT‡­²xy——ûÂzÅ“å¦m*¥Ë]kê>LWM¡K7ìöU•µFwîõêà–Úý¾ìÉjGlyym:÷m$m¼XSk7(L¥ëÎ4uVºÅ©sÿÍôþY¯"º¼]qºlÊ[ôYjJoâFét‡›}÷ ¥ ½ñP³ñÙ*B“áôïfƒ°H†/æ"ˆ‘øxX1õKÒÑAR&O¬Lt@ŒÅCŒ1·€‰x²B¥CL|oë c?oöuïfþp]7ØŽ>.óUÓwÃwêÂ}ˆƒ S¥6N„ˆ(‰Sþˆ‘éÌŸ€l­Ÿâ̽A¨Šô$"¿û,„'®'€üŸÀîÅCó' æBŒ].UÇí%\È¿1bþ0Ä |à#"ÁØ{BÌïC,ÞbW–ÊwX<±x,ŠÓ˜‹°Ù(ýk‹' E1SQtôGê}!~O·ÈII¬|vvß¾Ô·&ƒÖübµ–œ‡¶.åÆ•‡í0£Æñù5¸£"ïN>_ Þ¶Åh þ½c’$* žífm"³zÒ&P!Æ6CÏtBd$¡i|ÂI§§£ì_ $¯AÕÿæh¦œΆŽè9—3fchf­NUÝ;³Y¨×ÿ@(Àq),¶1 R˜ÏÁjù¶5½fîÑ,n<ú-o–aÃKÅ>Hy.Žòög’ÒaaëØcáþÂTá&$L/e™2š‚©¢©Î±9f¨¢>sĬSÚ\î›MshŸVÚNÔ~0ù«dY´+6ÍÕʼn˜Ïó ¾;Eÿ_š›o}¦vsÐ1L­ZϘ:ó«Æ“ým®oã™ÐÕÏ:ï§ ·ºt=DTƒ|9íö–l:­Ï=£û`ØÃèÛÆôPI> N>e'GÑúUÖB³e²›PA¯²Íê7ὩK}«Ëndu ¦#ºê°õ!0ÑûØîñûïPxÜØ›>ÛíÊ:3¦®Ï- ?¿ÏiŸ#aÊ^Ÿ|ýé±h¦ ZôÎ ŒGÿ±>;ýÙpv›w'u¬ “ô¡w’H94š©Ï!QŠ+Ÿ}uùìO*cM endstream endobj 693 0 obj << /Length 1402 /Filter /FlateDecode >> stream xÚ½XKÛ6¾çWÉEF× "%-Ðm“Šl‘¶¾5=Ðm+‹®¤}8òÛ;G¶´+7»M6'ŽhÎ7O~$ýóüÙË·ŒªIF2ÅÕd¾š°XE³I¢¡ŠMæù䯨|¯+½5©êéLrnýÎ^šÂnœË§ÏÌÒ”dLMfŒ‘LÊ çJóÚnMY[Wêb:S‚ÒˆñéŒ+™1Ɉ””µt¨Úl*3T»4âô gDÒDyå—oSÑ‹ 8#"MT½qSи±å”¤Ù覓 êÓž¾J £`Æ‚¸o.Q„fY·|ërSt[‡QçæŸ+Ý2ÅxFx—)´æÓ”F– ¹E£mdS7v ªI$‘[…iïw+,á·)“ÑÚUûY½3KûR¾ ?n¯ŠÆî Û–ÍOt¨¨žD5”Ô¸f¿C¸Öù3/gÑ Î•Æ䨆.]Y7ÕÁÇ…k6Ó ¡Å$‹ãaE ÎB}U¤Ë<Áê¥.Plà ¿Õ8µ;t ™Î„L£y§W覃|ð¬Áoæ¸Ønw®6hz±£OÙmèŽC…ÛÝqzជcÝ»t¾(×¾íte±~¾#ÊcßµBmÀå<ÈÇrá"7Ò›mÛ'‡nc#ý(‰äq·à,`Ýlìrƒfô> ôA/7ÖxW¯½Ó½Yશ©VûÞæAµ{1"Z×ô*‘õI†‹”ð8é|ü@™#KkjÓœ¤c&²Œž@ú= _½ ,õâEfå*ä^‡ŒÔ ë«íVWûTÒáî‡Ìø5ï0Î/º(ÎÇð×åÖƒ‚;PlŒýØ9vå»Ö‹ŸÃ°kw˜—~Ã4ˆW÷?œ…©[6XqˇŸÂ;ŒK­·¶lëÔ3kó³ÖQNIÆÙÐÛ•ÞÚb?Xþ|çl]»ò9‚溆±ÐËË÷àß;W®»lAr 8$°H¯ÍµÕåÝùÃÔ6¿ÒE}~Ò Ÿ¡±ß§px€par«»Y³‚AË]èÛ¡Ù>ÊŒ*„Èd4ƒ,ë}1!c§DÈLà# c²«iÖÚ™ÕÊ.­)<=7´©-=»Ö FkÚ6jÏ™7HïAáÏ&'AzSU® âÇ0\ëâ Éuˆ„ˆï‰6£þ”UüNñ´TÀ1ÃP8‰e:Ö-Fa×›†OVîÚ@ãCO2¡ 9 .iÜCäI|D¤â¢?S*ú 1¹jïà‹TÌXøÛ&<ÉIL]]š|ÔKØ”,¦GÄ,‘ˆ(àB#O!.ÝvWÀùB*Ø'fåU8À®VJŠcÐ2a8Aã1¸n›’GìQâ}¸q•?î²â€æVÆ ¬ùí'ìlúÕW ÓÚ›1Ÿ·ä ¿6±îòýêÍ'™ß•OjþD+Ò¿/ñS;0Ò·!~Øpí~bó÷š¼åM° ,?®×zgy€uï"ã´Ý×J2ú¸@OxdÔÇÉ73v'4Œ‰£±Tˆô„±Qú_yƒ¯>}üÔ;*ÇöØÆ­)¸2e)}‹34áÒ"Çjû_l¤@úÉ#Y­RÎ’ôáì e£­ͤx –…Óü ø­­QÈÖw Î®Ó¾å™0Þ-Q|2‹è›³^û7ăIàÁô³/æèÿÙWÁž¦¬;°ý›¹‡|mááWÕí{=WÝõß8Û{04úÒà´ÿß ¼ÃÂÈl7¼JÞ-Þª=¡úñfcŽo¹ÃÕ;LltÝYêlÞ{þ]ÛÞÎïà^•6»²&ñ©{—àsß,çáS•¦ˆ‹ã´[º®Œ¹ó$XÁTî¶#]W¥b°‘4Mƒñ8ó+Ÿ½™?ûä[ endstream endobj 698 0 obj << /Length 1060 /Filter /FlateDecode >> stream xÚ½W[oÛ6~ϯ0Ú‹YÞ)[.IKÑKÞ¶=°ÑÅÐ%M‹b¿}‡"eKŽÔ9@â%Šç;χ¿_½zK°œÅ(–TήV3!‘ŒÙLI‚°$³«dö×üÍ»³Óã… bN A9þçêž-(CqXò¾É¾˜Ò/*V~´µ)um‹¼ Ò„;ÉWo#ÜSèp\y ×#7ç63yâ:õs¿.üxgc—E“×~æ_?¬µÍýÓ/~¸lÒÚ-uN:r«Ê´“÷þ“Áú{:|enÅI·]Ø{,„§ +ÅJÌMj3›ëÚx‘ßÂÖ“ n¥3›~|{±.lUù‹°$ѵ,ø¢—·€ñŸE~ÝRl6tSg5éé¼Ëmmuj+›_ÑýÔä¹ûÔâWµ.ëE³ÞõBÿ+šm,Üëÿ »sðæOaÞß^-j}kÂt]„[D(ÎdS“‡zw[eßo‡þÛ¹Ç_o ýî®Z›&ÝOÜèªÓÔéÞÜM,´ƒÛä6»²°Â_HYïFH¡e•’Ïœ ˆ3OT´‡ÂÑÅÕѹeÅ endstream endobj 716 0 obj << /Length 3363 /Filter /FlateDecode >> stream xÚÅZëÛÆÿ~…à/ᡧ5÷I2iÄçv‘­smP$BI+‰1E*$uòåCÿöÎì,_ºµvzÀË}ÌÎÎÎã7C=»»zú5Í"c™fq·]hÃL&‰á,6|q·Yü½²m±9ååõR mì}‘Wkû)½jͲ8¥v]ÑS¥ýÔ]cmë·ôÜBצ>\ÿt÷·E¼XrÎ2­i£/_ÞzªºSÞÏ’eÜ3ó÷Óae›9É¢³MÞuÕúå\áÒ§_§rr4$Ä™Èüfw{{½©ŒÊzíÓÒÄgGÃ*‚A{Íu´«›‡e{´ëâÇ8kšu8•]q, Û´Ô±Î=5pÓ5yQÙ u¬èÙÚ®+ªß­ò»\l+£cÞäÛ „»šž¿Ù¦¾f&"[Àì&$ÈnßԧݦéØ‘$iÄi¨”q•ÁB·`à6 73¡D?3o®yíN[uDXL%²_³«ºÜk¦!6lÛÔº¥4‰Ž¾ûÊ]ñr˜;½éoA‘æw}#èùÿ*g¯õv ê‡óßðž-†Lžë¦…ýÒ &Å <ü£%Ã[@Í>Ðæ-(ºß{)Yg‰ôœ€ÐÚ¤só²Øí;VŽM}oI«¹N+:ÎL–ô¬$O¤øˆ¬ê ú1;g—,3RdJö¼Ò23^̇å%o^ÛMH,ŠÎõ(–4ÓB}LõXׇc ž5-øIÉÁO,5“ OµêùȲÌèÄñ!2pnòÂR~=åm¶DÇpoÿΛ޴ >°i©8F%J@»µ_*7‰ÒÐ÷çXW ®.åRªÞ´y Z• þé­öÌÄB™tdÆð,Sù±°˜iÚ”ö¯2VZÉêzÞjc°;Ü’£ ÎRžÄÆüéV—”qØjà@ÌÁ+¾_íŠkXqOPÇÃ×h쯧¢q°Þú`×2"™ñNô4SP¯“ ·Ã!EßX»¼Í›ÎzØ€7[Rs‹ûöp¢spÅV›–Þ‹ŠžùÎÎÁœ£ÑtyYtáÃñ”™”øx‰T”ög‚FÛ6ÔÌ« 5àV±k,¡U×_Oe8•šlë†Q?[DxÊŒôûãÙcÝZ¿Éʯy\l?HÚxß<Ë„ƒ!®g[7‡v¤æÐJ’ÅdÖÖR†,nŠ%EKÅ!çÓé\]î]ŠB’¸+Á'XY€]3лO(l¤Lˆ!7Aá ±KáúòðœÉl :TÂL2LØçžÇƒÍ}¶Ã¥ž&F@O™Æ±0qøøÒäàÈ!ÙNC}ÀÝïÑIåˤC\hðF\HóN`”ÿNV²)+hç=c`{L·b9Ý>xK dNý-y±rz8½Á†·Ý妀³X]‹8:uÖKÁ)4:;Ù{.ì-ZÊRùæzÖXtÛç JôGZF»,…âóq‰8tS$€4߶íå§É{k[¿jK]5ná]òŽç8ÞgBþTmÍpÁ#iJÃímiþ<€IrZ7ádÞNW†xƒ§ïƒ'Éχ½º,k”Êù"Þ–õn¹*Ê¢BûñSª¿ðÔ,Kæé'ÌGÝ©7,ÖÔ4椃ÞN¦}°män0EØL¢ý†øh³1Ϧé8/Î÷yˆíŒñx⸠J ‹Ó9SÁ½ ðLyúKh¿péd;Ü+uí.Lšep¶™ ¸’ãfFí!LMg3jŸľ‰ð&çÀpâtf—-ðúúÜ6¶f^×á˜!¦¦Yô³7¥º=5öD¯|q߀g9q5--Ï;jŠöµ·q%²ènßoTôéug+H;Êâ7ëÑ@UWƒ=f#x…®üå:ýÚtéÙvÕÆ!CnæžâXXÊùõž#{W,Sj¾ˆê ÚD§Ö7†‚^87šLQÃí¨wWô™d€!®b äawÔìHBÐðº!aé$rHÈÁ;¯9 Ù óiôýuGž¹sQöëÊò„k:?â °;‡Ï9L´É»œˆoÈ5VeoÈ{šˆ\¸A0f@$^°t´2Q¡WѶ©Ôrª„'DÅ®ùíäñõ9l Ú‚šm„@ 3 wèk*è¶8}ŸÌB%}ñíóÚág¡VŽ‘‡0EQæ«Òw¯.¸ýW5шQ=¯°ÁóøùWEŽ[È8zf›×x¶´ýŽüÆk'1n /“ÙEZš¿A’JÀ8¯©ý²‚l¯;¹Ò&ïA4žÛC½kò#À_êxe[°ŽõÞm%ÑàBÅS.˜Å€ìºîøéÓ§çó™ )«›] Íôˆ0)-Â}늞‡¼ô<:øÝR»ð£·Òlr:¼” ° ²²ÝÙZT/ #ˆ3ŒòÙt!üGÝã$óƒA¯«0š7¾‹2+è¡‹vc/ÚQniøˆ‚z‚F,Ît"ÞkÅj„ç+»ÎÉ3Š€8øÈt|' 2@v.Ñm1JâÈÃ<šŒÊçTg¨bÇ΋OäåN,U´òM€Nª4¨Ô0ç¾Èiœ´zÎvE=mq8ÒêŒtúúzðÓ6Î1¸ÎÍΈ€Y1êêÝ öæm{:Ø~*:~ìÑ¥á é2t9å((·¸„r¨VxS‰t–¼ª³OÔ e8BZiZvõ²74º÷âñ~ˆ¬Hå`ì‘·ƒ>ò]8 tÁ‘/¶¬éëÎ5¢ñ‚Š ° ¨œ/ljž>éj‚Ë‹£Vп |yKF 9ZÒ…J@5é$|¬Ë˜ø òS^…>ßhÜéú‹¡­ëòt¨Ú‹F:ÐâþgL;gV`ÄL ™ŽsïÔ\ìZw J“¯“mA‹ÇÌ%´{j˜0CZåœGpwiø$Ò>¦3—zœÀe|z ~ BçG_îì[>æ©%tÏÉzZœ1©Éù€yyO†¥É0çð•7ø5¦3Êbâ@´+–UZgèoH?û(£ßY7=äCæ諤+¿j¨”ö‰gº«%†U Æ=û¨úÍ­ÛƒQHÅ*ò_—þ;kuÀCôâs}ÿ¥*Ù´lݹA:ø€a_=ÃA´Ðá¢@ï3q2÷FTö¦éŸÓ繃À'·ùÁ¡Éì'îƒÑ‘ŠO‚y:9ÅÉrncª9ÏF>QX¾°LÅQÉWEš»yÉׯ+òï¥Ýì\)…›â%ƒÅõÝ.fA¿Ÿ0@oìƒ4ä:!ð–ÔN@òyÔÓR¨Ç‘bK]yÙØŽ»‡²Dp¥B˜?šÀ^‚a9Ë—Ç_Hx”HÕSBe‰¡šËÒÑ°ŠÿÝ‚K1?ã o9?™–€¤+búS}:W® 8ñ®‚fÂtÌç‰Sx¬÷yÕoQT¼Íäœ/ŸUöå£ñ#‚&taÙΕ¨áu¬`Ëõç÷líਠð¼b±¼€hpCöÎâ§Éá›mÝ#‚ÂÓoýÍPÚÁa@øiêã) í<ÜGÈ endstream endobj 735 0 obj << /Length 3436 /Filter /FlateDecode >> stream xÚÍZmÛÆþ~¿BpŠ†‡úhî_ܤ€ã$@‹M›ŠÂ6JZI„)R!)_EûÛ;oKQ:ÞÅiR lîÎÎÎîÌîÎ<3º/n¯^|›E©N·›…Kã´0‹,Uq’ªÅízñ&2Úª"ÎÍõv&jüÚ(5EÁ”µ¿V.ÚvÞ÷²a– Ö힉oµÎ˜ÜùCç{ß U³å±UíËŽ›þZ»èCµöÍÊ3;K3Q\öC ³ÊZ–Áe?à¾[WýøUÔõïÐÀ:u9T$˜˜:´,vØùëw·Z$‹¥âÂ9VöÛ¶ê{’`ò¤]µ¼ÖIt@n|}cýQ†Ûní;n’T“EdŠuÛyØy©ËÕ{™¶aÊÛ$Ñ“JþÔÕ0Ô¿9v0·»¶Iô(𹑠}`þp£<Ì åÑ·`ÝQ'0}µ>–u?§}E¦K¢¯«í‘4€¶J^"+\”drQp"ô°ƒe÷6qÉ~Óµw×7lþ9VHÖÏ5|þ›[µÜzfþ솿e7ǽï*š ½Õ®ìÊÕài‘×eS®Ëß¼Úú§„ÞUÃîÄýüŸÄ£ز:gìðÚNW-„Ó¿y½o×¾Ž÷eíŸóøpðg:>;°Ÿ=¶<1[~>®!­ërùù3y&„{"Œg6Ý°i¬œ–%äöeÕœoŠLG;¢þßÆs§îò^L Û¾9“|±ùÀŽÙ=ý—ÛwOmù“Û¿Ç•?~ÿoÈ([ÿmO¢Ëṳ́8ôî¤êÏšô˜‚h¨s/v>ú”“±Ýê´×Wá}m»öxà¦v7Æ=f¹ô ËÙâQ#¤Zîc&ýßXný´å\r“N-g³8É̹à…á‹ð¼f6-ÊÆC@3…Ž&/»è¾M¡FgŒ4•ð·jÖÕªÐuawØ•3SH Ù—ÑQ"#Í–iM;pߢdSß³¤¢t¹òAÿí.td‰u9”p Vg‘„žsñ5u•G["§Ö—ëµj!Æ7Cjà2.§ðcĵ»k•€D¤è€ :t<&‡ƒÅýÐVXzæÑpFjœLmD°Kœ:à©1'»÷<ò^ΦMnv7íê(3È’NÂ;6Na5h2sß|¹ê\0@N¼÷L•F:)TVú%+€@èhcbän@P9+C }41•`ðºØÅ:!÷Ú|Ðn¬™ öÆþÓÔ!5ðýÍ\n­b3ò f)²pOé¡’tQEs~?&S±²¾ì{¨„M£e{fý W5€ 0ÌÁa?Üy?·oå’Ø8ö„¯ùá¶U›´ÜŒ ðoúSú›8ËϺ%4 Û£G†][‹B›Vta‡”c]s«+›­é’À—¯7µ‘€qhŸ\=¸kñäI›$=·'^œ˜Æ• kÏG_ÁõR6êÛ=9‚;ö&@ÙŽ¥18Ä~ƒ7Lßv÷LýÆû›×e7 fþ593h º5Ê°?R»ÀáÒc‡^µ*{h•²}/h…‰ß4ÐçÈ|‘zìý‚I¢¦Ü_s3§'/… vص½à[Dv=7ÁÐknYÇ_:(üN¼éGÀá)jüËß¡¸)ÄÜ6{‚Ë_úrØ ¤ù÷EÈùþ|¬)õãA $>^‰~x¹KGk\\*o6dÌI«Û--Û"ËDÞ¦ÜWõýyHüáXöÕÓ¢ù0Êg2™Ã˜û E~ûüÝ#ÁðfÜï$$ÞÒA3:Tî­%/,ÉQcáYÑUåaº ðUÚ.¶–{ä¤ÕšÌbN©*±lø+©*øY›eQ?Tµ,<É:éš3YùBVyŠƒ°.Ü«Ù,«Ÿ„ÛΓØDÃëïAÙ0eéÁQÀ³½±i÷âbV%C˜ "3ç‚Há¤Ë\°çqÒ†Gx„ÄiÔ‚1¼#Ú€aEÙ•là;1 Ð%ÚÐjR¿áˆIÔLˆ‡tÓˆ§SÅijuâÙ¡û>®N<‘“@?Õ +:ùd,ȉÈñ}à7/øK¾þ'k ¤²P.ÆV/1 ç® pÀC !¥òj¸—á6°‰ ­o ñª¹SO0ŒaÀ’†¡¬G9•,·ƒ€ã{Y1¬|±uKÙCF™>¦÷’L’zº_ÓÂä½°Ñ«f|‰çÊOœØ±ñxÿeâÜEÎâTy0œaGæ8ôeN.4Bn Íö8Ô÷ü¨ ¹„Eƒ«ÕñPPȸBy&&„áfß•…×™†ë\G)7T”)áiDj®°&§éTS…¨ô²¸M/&5c¾“Ú`e†ôhO†/¥Lœœ²ß_è Óä,|Í¿`X1×ùÏ~Á¬+`j/¡; ®KÊF¶êlš˦6XÇD‡«»lŸk¸Ž~±ŠT©~é*£çàw„ÙóÏÜÚºˆM^üzk犊ŽW_Ý^ýp…^/Y¨ð7_ʨ8qÅbµ¿zó.Y¬a ^BlŠ|qGœû…ÊŠ8Sh¯zñÝÕ_¯¾¿Cž1¾wê4Õ´±—ñ ¾Dq<°y†—,‡•Çç|„XCªXi¿o¥ÄJ­“<“@8Œ¨ðÄzœ}’NU¶ˆõh‹Ý0^¾xqww’â¶Û¾øö¸¬«Õ ÌÖ?Tþ.>ì3v«%Xp ¸}ùK9§ƒuÿå<@ endstream endobj 743 0 obj << /Length 210 /Filter /FlateDecode >> stream xÚUŽ±jÃ0†w=Å6DŠNg+NƤ ´t)h+D,·©jeeÈÛÇUh¹áŽï?>~9‚nQŠ6B6çÀ¾™Ð÷°¬;Âò^Àú1 „‡È^æùÍxqñ?²½aëSG°[­4˜I4%[Zéáµ:ï—äj®ZY¡ÜÕœPVûi{»Zh°ÞM%øò1O ŽÃ²“›Æþb}ÁCŠa¹òGÑ>;Ç6e—Š0öÎÿ—„˜²õc¾Öoæ ær¢ëàHª4mé'dGÃnãM endstream endobj 724 0 obj << /Type /XObject /Subtype /Form /FormType 1 /PTEX.FileName (./fig-LCall.pdf) /PTEX.PageNumber 1 /PTEX.InfoDict 745 0 R /BBox [0 0 720 720] /Resources << /ExtGState << >>/Font << /F1 746 0 R/F2 747 0 R/F3 748 0 R>> /ProcSet [ /PDF /Text ] >> /Length 43210 /Filter /FlateDecode >> stream xœìÝM$É•Þû}~ _’‹1ÚûËò  @š*@¸ÐÜÅ"":G˜úö׎3w Š©$q$ÿ»y<û¥EtuUöS^æÿúöŸ=b3ÖÑ[Óòák2Í>Ycãñ럎ÿrüË[ÿÿ=±ÿõþþÍŸŽÿ÷í¿þ?‡=þû›;þCÿß_Þœ¼~üÇ·ÔLóGŠÖÄt¼Ù›TÖøËñ­“e¬>‚äSÿ‹:Ù K5¥Èº1&[MÓuz!»dR½^/ÉÿàNÍJlך[ãÐÆë»Ö_Ÿãõ:ëYÏzÖ³žõ¬g=ëYÏzÖ³žõ¬g=ëYÏúϸ~¶Z¾¿Ír¶‰kœåá ßÆþv+)TcGÁ™L­k”/ý²àŒ&å#ö|ȲnŒÉ;Sç[²¦ºëõ\ŒkÚ6-ï×8´ñú¥×çx½ÎzÖ³žõ¬g=ëYÏzÖ³žõ¬g=ëYÏzÖ³þ“®—RëÃ÷·¬-äl×8Ëþ+üTpf“â‘bûAû—n²jŽò¥_œÅÔ|ÄLoiŒÉG‰- Ù6âõzv&×÷ԼݴjRXãÔúë—6^Ÿãõ:ëYÏzÖ³žõ¬g=ëYÏzÖ³žõ¬g=ëYÏúO¹^[H­ßßf 9ÛÄ5Îòp…oã {ûý÷·ßýÑîøþgù˯G2.ö°ú¤.õGvÙäv|?~óËoïé¯~ÿõ#뼕/ûW¯‹}›5Ü×=v±ø.oµ¦#…6ne=Š7É®Q>¨WulqÆ•#– _±¯cr= u«^È!èÔŸ¯÷¯Öâƒ;5k’{Ðj\ãÐÆë»Ö_Ÿãõ:ëYÏzÖ³žõ¬g=ëYÏzÖ³žõ¬g=ëYÏúϸ~¶Z¾¿Ír¶‰kœåá ßƾßo*›ÈzD¨ÜêZ’Üã:GùÒ/ Îhb=bŠz‹¬ŽÉ5“ç[r¨Rlž¯÷¯–~t¿i¿.ì©%ãâ‡6^¿´ñú¯×YÏzÖ³žõ¬g=ëYÏzÖ³žõ¬g=ëYÏzÖÒõÒBj}øþV´…œmâ‹–‡+|Wø©àÌ&–#…"uhÿÒÍ8¿FùÒ/ ÎbJ;äÔÓdÝ“³r<éØò¸ %­«×ëòF˃{j.lZ•jwŽSë¯_Úx}Ž×ë¬g=ëYÏzÖ³žõ¬g=ëYÏzÖ³žõ¬g=ë?åzm!µ>|›-äl×8Ëþ+üTp6Sê¸es<­©zýåK¿*8«3Î1»qSé¡c²Å}ä‘^÷ÝõzÿjõGÏ|¯ÖÄø •¼Æ¡×w­¿>ÇëuÖ³žõ¬g=ëYÏzÖ³žõ¬g=ëYÏzÖ³žõŸqýl!µ>|›-äl×8Ëþ+|/8k0VŸ”FÁ™L k”/ý²àŒò‡Øå=´ñ–Ƙl“ÔxKãBNIÿTþ|½¿QëÜs}¼´þuʇ6^¿´ñú¯×YÏzÖ³žõ¬g=ëYÏzÖ³žõ¬g=ëYÏzÖÒõÒBj}øþVµ…œmâgy¸Â·q…Ÿ Î,Åg AžÉÞ¿tÅ5Ê—~YpSô¹õe¼¥1&ëÖm× 9ùêçëµÊa£»{j6oZ5¡®qjýõK¯Ïñzõ¬g=ëYÏzÖ³žõ¬g=ëYÏzÖ³žõ¬gý§\¯-¤Ö‡ïo³…œmâ«–‡+|Wø©àl¦ÈcÓ£©ò¥›7!­Q¾ô«‚³9cÃxÙ˺1&äáõ²N/ääåQîçëýÞŠÓ©Yʃ–Û‡6^ßµþú¯×YÏzÖ³žõ¬g=ëYÏzÖ³žõ¬g=ëYÏzÖÆõ³…Ôúðým¶³M\ã,Wø6®ð½àlAnô”[6‹ÇëuÖ³žõ¬g=ëYÏzÖ³žõ¬g=ëYÏzÖ³žõŸt½´Z¾¿5m!g›¸ÆY®ðm\ᧂ³™Žä›¯üsWÜL¬ãYvvìÑ{ãše/»boMuGÌy­s²QP{×+ò˜¾Ø¶Dí_2<Ø—èÒ.:Ëš—Ø›8sÞ_˘]®6°ïo«Ëì9ÏþõÌßç•êŠûúÒÆ™UwLŒk–¼îŠ£q~4ßaôÜ:'×#n¾»q%;g|Ý%Ê—Øí%ö·ÅR׬âH<ˆ=1ç-ñÅ írµ•®X»ÜÙÈžóì_Ïü}^ùç®8gdzìÊ8ÃWéoç,;xÝ©£cIÆ“ŽuN¶ÿmžŸ¡Wr[¸5ËQ»}í¥ä]¬Æ¶5«8›8sÞ_Î.WX銵Ëì9ÏþõÌßç•‰îH1Ëá}Á›×,;xÙ‡þJ?'8¯4gBÞ©¾ºŸ9›ìÅ׬âH<ˆ=1ç-ñÅ írµ•ŽZ»ÜÙÈžóì_Ïü}^ùtÔ=ì¢éUG=ÆLü[ÊMÜþ#åv3¾©/²£ÜNÎ4·fùÈ^–ÛÉš$'™d9ÚZVŽ9ùxžz­W²MóŠ3‘š‰?<ºï¥…]t²Ý9ë?¾‘ØÄ‘˜ó–ÀÀÀÀÀÀÀÀÀÀÀÀÀÀÀÀÀÀÀÀør†t¹ÚÀ¾¿IBºÜÙÈžóì_Ïü}^ù”ÛÕ¤¾ä`éŸVÍÉË×R-§Ñߧd¼_³¼×Us0Íö#šÐÆÊ1'ŸÏ³>æ•6N¾¾ÙÉ&»}‰Ò®_b”Óµç¼Dwq$æ¼%00000000000000000000¾˜¡]®6°R5k—;Ùsžý뙿Ï+ÿt.t__ë‘bY·%Wi¹ç,;xÝãÝ«>ñ˜srUèÞÇ•ì›}%r›Çw{‰YZî±–5«8bOÌyK````````````````````|1C»\m`¥+Ö.w6²ç<û×3ŸWþ¹+nƵqs§~dgj\³ìàeWœ­Üz,ï&Ž½ë,­tœ'Šè•lƒ‰qKô7Yík/5í¢3®®Y?‘ØÄ‘˜ó–ÀÀÀÀÀÀÀÀÀÀÀÀÀÀÀÀÀÀÀÀør†t¹ÚÀ¾¿IBºÜÙÈžóì_Ïü}^ùŸa‘ª‰²…lŠÿùÙ›¤þ³£Ï™³¼×Usg Æe²rÌÉÖó‹z%÷­‡´%JÇ?<-¹¿îò.F9ÆcÎKì‰M‰9o Œ/fh—« ¬TÍÚåÎFöœgÿzæïóÊ?Ý–Ü×W}’^ÖÛ’«<Éoβƒ×]q1.ŒGÿe7VŽ9¹"Ç3ëÞÇ•ìÇÓ¯Dÿ’!<ØKÌÒÁ?ˆÕ®YÅ‘x{bÎ[ã‹Úåj+]±v¹³‘=çÙ¿žùû¼òÏ]q3N…çM{,Δ²fÙÁË®¸X¹w:ö ÔqG²ÎÉ6¹³zì]¯ä`åM^‰þþí»âfJÝEgœ[³~#±‰#1ç-ñå érµ}“„t¹³‘=çÙ¿žùû¼òO]qñ&Ê£ð¢iã’äð‰9Ë^wÅA^‰¹o Œ•c–¯TçÞõJQ ð+!{{´/Ѷ]Œ&º5/±'6q$æ¼%00000000000000000000¾˜¡]®6°Òk—;Ùs(ŽYó÷y埻âdJÝpÑûŠ« mͲƒ×]q16±f9±YVŽ9¹ ç$ëÞÇ•ì²IeKäúâÑx%›hÅâ׬âH<ˆ=1ç-ñÅ írµ•®X»ÜÙÈžóì_Ïü}^ù箸›ŽªIã¾âêLnk–¼ìŠ«5¡QN»=·ÎÒJ—ùX?½’C3¡n‰\îg!_{)nÜ=gý4FbGbÎ[ãË6­öýMÒåÎFöœgÿzæïóÊ?uÅÕ›z^»â4*ae¯»â`r=bìïÆ•cN6Jý=ö®W²rÐÆ•(ÕØò`_¢õ»Mˆk^bOlâHÌyK````````````````````|1C»\m`¥+Ö.w6²ç<û×3ŸWþ¹+NR*˳ìôÞçZå¡vs–¼îŠ³iíˆÅK½-+Çœœw0½+9–y×ôLÈÊ?¼¯¸¿Â.“Óš—Ø›8sÞ_ÌÐ.WX銵Ëì9ÏþõÌßç•iåHÑššeÍ™ì×,;xÙ7+·GǾW7îšÖy¿¯X¯d_Ml[Bn¦öÚK“›™Ė׬ŸÆH<ˆ=1ç-ñÅ írÏûŠg—{ÞW<çÙ¿žùûüò¾âÖçz¤f[Ý¢iaͲƒ×]q0Ù±:ãÆ=Ñ:'çÎ6ôJîWêž³”íƒ}í¥Å]ŒÆ—5«8›8sÞ_Î.×VºbíruþåœgÿzæïóÊ?wÅɤ6Î'öã¼âVåÿÌYvðº+Φù#kü8acÎÍʭѺ÷q%§db½©Ž¿íö%ö½mb1©®y‰=±‰#1ç-ñÅŒÑåÎVºâÑå®FvÍ«]ùû|査âjš=RòÆÉ9Þ:“Òše¯ºboÇÛ»¤C+ÇœluïóJýK¶-QìxÎÞñÜ7“ò£XÛšõÓ‰±'æ¼%00000000000000000000¾˜¡]®6°ïo«Ëì9ÏþõÌßç•¿wÅÞöÙ3'Šî1ÊAÅs–¼îŠƒI¡¿›&­´¬³ÜÁ\uïóJŽUÎø¹¯pöµ—Zv1o׬âHlâHÌyK````````````````````|9ûÕÀJW¬]îldÏyö¯gþ>¯ü·§®8™$ϲËrû±ì WÖ,;xÝgSå4fkj+Ç,_)ç¹÷qEN_g!¯„ôØg!_¢«»XL²k^bOlâHÌyK````````````````````|1C»\m`¥+Ö.w6²ç<û×3ŸWþÛSW\M•gÙÍóŠ}ƺfÙÁË®ØYãÒK6IWŽ9¹`òì¹õJAÞä•Èn5á÷®¸É#ýÄêÖ¬ŸÆH<ˆ=1ç-ñÅ írµ}[]îldÏyö¯gþ>¯ü·{Wì¼qú,;7î+vÑ”ºfÙÁë®8˜˜Ø¿’§WèœlÌûŠõŠ¼_¶D Roïöµ—jw1ç׬âHlâHÌyK````````````````````|9Cº\m`¥+Ö.w6²ç<û×3ŸWþÛSWœLLG ëœ Wmk–¼îŠ³ÜºSYMø˜“«1›p½’c^÷,k¢Ø»}‰Îíb11¬y‰=±‰#1ç-ñÅ írµ•®X»ÜÙÈžóì_Ïü}^ùoO]q•¾VÎ'v㜠ïL´k–¼ìŠ½5¶öwÓL¶c嘓kÆÎ3(ôJvÖIJ%²7ùÇ÷7ý£XâšõÓ‰±'æ¼%00000000000000000000¾˜¡]® ìûÛêruþåœgÿzæïóÊ»wÅÞË Ç)öËã Mqk–Ø—]q0¡õwMЖyÌIot{×+Ù·q~Æ™È}ÅÏ+î{)a£±iÍ*ŽÄ&ŽÄœ·Æ—3¤ËÕVºbírg#{γ=ó÷yå¿=uÅÉ„r¤hMvcÕX¿fÙÁë®8›bX« ºrÌIÀõîôJ뉺%ä¾jû`_¢ »XLÈk^bOlâHÌyK````````````````````|1C»\m`¥+Ö.w6²ç<û×3ŸWþÛÛï¿¿ýîîpÇ÷?Ë_~=’œ6a«ÿGÊçœÇ­ÍÎßßßüòÛãû_úËß½õÌø>¾|–Gá%ÓÆmÉÁIG=gy/«æ`•:’´å²rÌã!|ó0g½’£3Þm‰’äi}»½ªæfB|sY³~˜#ñ æ²æ-ñÅ írµ}[]îldÏyö¯gþ>¯ü·ûmÉ}}kGŠaw¢ÉqͲƒ×]qÆ<67oºÖ9yoܬÉõJ¶ÍÔ¶%äø`/Ñ›œÅV׬âH<ˆ=1ç-ñÅ írµ•®X»ÜÙÈžóì_Ïü}^ùoO]q2Áj7<\Å´´fÙÁë®8›ìX³±ºrÌÉGÓfÏ=¯4Y´%úÞ\~°¯½´¼‹EÎ\ž³Š#±‰#1ç-ñå érµ•®X»ÜÙÈžóì_Ïü}^ùŸa¢Iõˆ¹Éη#,>´0›æ~öE¨rž²<‚/Ž{®åVê¼fyç¯;êfZ8bË&æñ™9ùj꺟Y¯´ Õô•û¹Ýƒ}‰¾lb´&µ5/±'.QsÞ_ÌÐ.WØ÷·ÕåÎFöœgÿzæïóÊ £.r/µé}Ý_ÕQëÂbåÜç¿~aÒåoùŠYîÿi+Þ?©&O=”Ï:šTÖ,ŸõËV<ãcÿçæ'”è,½ÿ<ïy]iNÊð+‘êýÁ†Kô£ÏßÅf׬âH<ˆÍ®yK````````````````````|1Ã.WXiŵËì9ÏþõÌßç•ÿv?å#&ãÃv"t,¦–³I鈵É[•cN>ËóuïãJ¶Vžx%úÞÊO„î{…ü)9{Î*ŽÄ&ŽÄœ·Æ—3¤ËÕVºbírg#{γ=ó÷yå¿=uÅÕ¤8N×Ð=&'GGÏYvðº+n¦Ê&ý+éýácN¾»ºb½R³¼É+‘Šâ»}‰Þnb²&ù5/±'.QsÞ_ÌÐ.WØ÷·ÕåÎFöœgÿzæïóÊ»wÅ}}Õ®Ø3©S4±­Yvð²+NãƒrvHÑ•c–¯ç]Óz%Û["Ûuçñ­+NÞ$÷(Ö°fGâAì‰9o Œ/fh—« ¬tÅÚåÎFöœgÿzæïóÊÿüdf)Ÿû|2¥ýü¸ã4¥hå«È(¦Ú5Ëx]5g倎qß³¬s ΄y(õ¼Ò¼){¢ ÷ãª9™êw±ãšU‰M‰9o Œ/gH—« ¬TÍÚåÎFöœe¡VËš¿Ï+ÿíé¶ä*Þ“ãÓØAvróœe¯»â&e´ç^oº³–‘çñóJM&Ú-ÑÿV˃}‰Îob¶&¦5/ÑùMÔÄœ·Æ3´ËÕöýmu¹³‘=ç¤ý뙿Ï+ÿíÞ÷õE…×äfdÙA4ѯYvð²+ÎÁ8yÐ_˜õÓ9…±`ì}^©ãé|W¢ï­µ{‰ÞÄð(–¼fGâA,yÍ[ã‹Úåj+]±v¹³‘=çÙ¿žùû¼ò??¢/°}“%ïGX|ha þoY˜³ÉO_ñùÐŒœŒ­ãÆç6žå—‹Üƒ=gùÈ^—ÛYøX‹‰ã¸ 壉«ÜÖ+Í˱ W¢o0?ÚKLòÂM,Æ–5«8›8sÞ_Î.WX)·µËì9ÏþõÌßç•ÿ@¹ådçXŠÜ´ü󪹚 þKó^ïâŒk–7ðºjn¦¸ñu¼¾õ1'–ÔëJ+¦å-‘ýjþïUs36mb±&Ô5/±'.QsÞ_ÌÐ.WØ÷·ÕåÎFöœgÿzæïóÊ»Ÿ¹Ñ×;ºá0êðMHk–¼ìŠK¿àGqîÆ Í:'_åvè±÷y¥eéÓ¯D‡Rïöýh¹w1×5«8b®kÞ_ÌÐ.WX銵Ëì9ÏþõÌßç•ÿùmÉÅJ«;ð|æÆómÉ%ëFµìÜxÅä¼fy¯«æl‚HÒ¿\+Ç,I—y܈^É6ÏG!ÎDöÒ§ïö“œAý ¶¶fGâAì‰9o Œ/gX·X©šµËì9ÏþõÌßç•ÿöt[r5Aý׿â芫•;¢ç,;xÝ7i¯cÿ[Ò[ªÇ<:èæÞÇ•ñnÊ–ÈɸÇã1®½HË}Š}/Á®YÅ‘¸DMÌyK````````````````````|9Cº\m`ßß$ÑòÙÈžóì_Ïü}^ùo÷®¸:“åIzÅ´qctò¥ç,;xÙW/•wÌu>èPg¹ºçÆÞõJ¶Íd¿%r3þ‡÷÷×}ÛÅ ›žó}ÛÅ‘˜ó–ÀÀÀÀÀÀÀÀÀÀÀÀÀÀÀÀÀÀÀÀøb†v¹ÚÀJW¬]îldÏyö¯gþ>¯ü·§®8ÊfRê±q$s-&Õ5Ë^wÅYžžK6ÙŽ•cN~ºw½ÒÒ< cÎÙ÷ÃãŽkêAl~Í*ŽÄƒØüš·Æ3´ËÕVºbírg#{γ=ó÷yå~ܱ”ÏþÈÎË#õ~zÜq­r*²YaGÞ¬©mÍò^WÍͤrÄêMÑ[ªÇœ‚3aa¡W²:_‰L-ö«inû^|X³Š#q‰š˜ó–ÀÀÀÀÀÀÀÀÀÀÀÀÀÀÀÀÀÀÀÀør†t¹ÚÀ¾¿IBºÜÙÈžóì_Ïü}^ùTÍŤ¾É´ðóª¹9“Ò‘b”¿ÉˆÆÛ5ËxY57oj=¢,Ô9…`Ê<ßc^‘Ç ¶-‘½lp·/Ñ»] &Å5/±'6q$æ¼%00000000000000000000¾˜¡]®6°R5ÇÑåÎFöœgÿzæïóÊÿ¼j–î:Ùe9ÙøV5`¡éÖ~ £Ž¦êüÊ8Iº“Üšå¿î¨³q­'»¬³œñáæàóJ­¦ú-‘ÚÝ^b2É?Š5­YÅ‘x{bÎ[ã‹Úåj+µv¹³‘=çÙ¿žùû¼òßî§4·þ‚<Á/邵òæ,;xÝ7ÛûßšÞì=æ¢üM÷®Wjÿ›Ý}oáÇ]qÛáe/.¯YÅ‘8Å™˜ó–ÀÀÀÀÀÀÀÀÀÀÀÀÀÀÀÀÀÀÀÀør†t¹ÚÀ¾¿IBºÜÙÈžóì_Ïü}^ùÜÏ\LL‡Ütmã_w?ó\˜äØŸÝϬ3±ŽNZÞWçѸ°fyç¯:ê`½©ý‚ •:ûb‚Þ%¾®ÔþÚ–Hñ~tô%º¸‹ÁIJæ%Jâ5¡óžÀÀÀÀÀÀÀÀÀÀÀÀÀÀÀÀÀÀÀÀøZ†v¹ÚÀJG­]îldÏyö¯gþ>¯üO;j)½K>’œåQ>P5G9$E¯B ¶˜×,oàuÕœsýðÆꇣ³¯«¤^WZYŸÎ©˜\ì%&)ÉÄRÖ¬¢&vQ:ï Œ¯eh—« ¬TÍÚåÎFöœgÿzæïóÊ»ÕÅÁVãäÑAaØwàF;gÙÁ뮸™èû»ñëVnÃxÞŸî]¯”*ñ•HIn¸Û×^¤å>ž[׬¢&Nq&tÞ_Í.WØ÷7IH—;Ùsžý뙿Ï+ÿÛ’Ç ÌRWW÷×1ÊÑÕÿü~fçLt㨠¯ïgyç/;jçM ýStò|?Y©srbõøÌæ•þ)¶=‘²œ8½Û—hË.Úš—(‰KÔ„Î{ãkÚåj+µv¹³‘=çÙ¿žùû¼ò?ï¨5Yžðç·¨š£)òäÀ8Ÿ$èŠ+=gy¯«æ,çG'[d£²Rç¾Ñ¶ªf½"FØ)¬sœïUsíú.»f5±‹’ÐyO````````````````````|-C»\m`¥jÖ.w6²ç<û×3ŸWþéè WÕ®ØWÙ·ræÆœe¯»â& L6¯›½uÎÔ8÷®Wj49m‰Mx졯½ä¶‰}/Ö­YEMœâLè¼'00000000000000000000¾š!]®6°ïo’.w6²ç<û×3ŸWþ·%d U íŸÞ–ì áHÑÊíÔò‚iuÍò^VÍÞ›œäœqe¬Ô¹o4éóúÖ•ÒL‹["SO׸Dù—ä3˜ó%q‰šÐyO````````````````````|-C»\m`¥jÖ.w6²ç<û×3ŸWþÛÛ>þõÍö¯jý¯ÿð÷·âÖÿ¡ÿï/orn´=þã[vR3GoM·XûÑûÎYvz ÌÑEÙÞx=š”·×uÜ^7k_¯ëx½^Ƨv¾>ÇíõhbÝ^×q{½È£ ¯×u¼^¯NŽ³>_Ÿãöz”£D®×uÜ^/rÿ÷õºŽ×ëÍ®×總MˆÛë:n¯ùí‡ëu¯×í¯l5ï‰h|Þ:ï‰"?L¶„Î[¢ÿx¨ÛǼæ=倗-¡óžèªÝ:o‰þRÝ>ì5ú=¡óž(&†=¡ó–ý‡üö‘¯yOD9teKè¼'ŠÜj¿%tÞÑnÿ¦ôÄœ÷DÿwxÿÔç¼'Šü)-¡ó–HÖ¤ýSŸóžèêþ©ÏyO9"}Kè¼%²•ƒt®Äœ÷D0uÿÔç¼'Šq ·D±ò”Ò+1ç=ä'«-¡óž(ÆîŸúœ·Dµ&ìŸúœ÷D0yÿÔç¼'²U´%tÞ­«û§>ç=ÑÕýSŸóžÈògJ¶„ÎWÂ÷×ý–Xóž&…=¡óžÈò»‘[Bç-Ñÿ]wÛ§¾æ=LÌ{Bç=‘噵[Bç-!¿Ýº}êkÞA~#sKè¼'ÆïÙ:o‰0~S÷JÌyOù=×-¡óžÈ&ïŸúœ÷D“?_´%tÞòèÜýSŸóžÈ&íŸúœ÷D{ø¥sÍ[¢ÿ»¾ÿâ¹æ=‘~ù\óžh¿€®yKô×÷_B×¼'Æï²o ÷D“o\¶„Î[¢ÿ»n÷O}Î{"Ë÷<[Bç=ÑLÞ?õ9o‰ê~5]óžÈ¿š®yO´‡_M×¼%šøÕtÍ{"?üjºæ=ÑäaÍ[Bç+lW·O}Í{"ËcA¶„Î{b‚|zqÜbá­ÜcñÓu%™Öå­kI~ÒÚÖÿ¡uÎ%ùÎz_X?¶0$ù×q[ãǦ(ö¶0}ì£qý›ðþmÛ¾ðcŸëßý懯˜?öáøþã C²/ü؇ãûwvýWÿmaù؇ãû‡ãþ°~ìÃñýÃé¿®í ?öáøþ B{øŠícŽœ¯X~È5ýpþðý­ÿ’Ñ1÷[rËŽ¹ÿÇãü´÷×Ãø¯¯×åAðòŸ©óõ9^¯Op½¾ù㨹ýmË·½•0~LÄþËNÿÆVvÿwéñmÿ»ëäëì_±,÷Ÿú?¥ñcw}Vý¿ g$ÿào¿þéø/Ç¿ü;? …ùÉßÿ!yý·ti™û:Ýî?þæŸþñ·Ç?üéßþÇÿ_ÿôË¿ÿíÿôÏúØ¿Gý;Öþ]oП£û¿î+_¼{yø“<`ªÊ¯¾²ð?ýéŸ~ý·ÿù/ýíé>¶ão©æ§ùÓø[èïúú“_Öœûå÷Ø/ë‡Ïãx¦G©qD»_õ5‰àûõ_cæ?”Æ~ø~o³{z?ýÇIsG’Bgüïÿ§ç;gùþêž»èÇ?éØ|Ö4VêÜ¿zGݼ’û7m±]‰Ôˆ´Ýs'oM¾CÜÅê׬âH<ˆ=1ç-ñY í¥g›üþ6{éÕ.¯yuÉg}›ÏüýXÓèÇïñËý…9Œ$¹/mβƒ×½w”“Sc7Éʨ‡¸FùÍi}wãJN^n!º5¯ßUº÷ÞaÜ¿u‰iÜÓ¡³Š#±‰#1ç-ñ‰ 饵M–Þ[{éÙ.Ÿóì’Ïü}^ùçÞ;Ë}H¡?þÀƒ¯ò§%ç,;xÝ{—ñ§ÀR”?<)+Çœl2~½»q%§blÛÕûãÞ;?Ux‰U¨9«8›8sÞŸØ^ZÛdé½µ—žíò9Ï.ùÌßç•î½›œ —Â||Z ~œôÎǧ½ì½ÃxEZ|çÆÊ1'äм±w½’Ó8¸ôJTû✓þºœùy‰nœ¨óßÅ‘˜ó–ÀÀÀÀÀÀÀÀÀÀÀÀÀÀÀÀÀÀÀÀÀø¬Æ쥵M~[½ôl—ÏyvÉgþ>¯üSïÝ×çr¤0+&;Hò„9Ë^÷ÞAæÓ8Q]VŽ9Ù"oAßݸ’åÑQaKÈ#ÄÒƒ}‰ò$šKŒ&Õ5/±'6q$æ¼%000000000000000000000>­¡½´¶ÉÒ{k/=Ûåsž]ò™¿Ï+ÿÜ{'yDkŠ#/;¨&º5ËìËÞ»Ž“ ~¬srMÝû¸"§¶ä¼%J2­=ØKÌr̃(ÏŽ×YÅ‘xKYó–ÀÀÀÀÀÀÀÀÀÀÀÀÀÀÀÀÀÀÀÀÀø´†öÒÚ&Kï­½ôl—ÏyvÉgþ>¯üsïÝŒMãþî<î÷ŽÎd»fÙÁËÞ;Zy1zSÇÝì:'gŸ{×+9¥yÎËLÔøâ|ï¾—ÜŖ׬ŸÆH<ˆ=1ç-ñ‰ 饵M~[½ôl—ÏyvÉgþ>¯üSïÝ7ãã‘ú.ã¹–1šÒÖ,;xÝ{“Üsw¢sN6 0ö®Wrȳ÷Ÿ‰ÚW<ž~í¥–]¡ÏYÅ‘ØÄ‘˜ó–ÀÀÀÀÀÀÀÀÀÀÀÀÀÀÀÀÀÀÀÀÀøĆôÒÚ&Kï­½ôl—ÏyvÉgþ>¯üsïL GJNZmÙA1­®Yvðº÷î¤<¥³ÇÊ1'׿r™{W²¯¦Ô-Qêzw÷Þ;›w±˜׼ĞØÄ‘˜ó–ÀÀÀÀÀÀÀÀÀÀÀÀÀÀÀÀÀÀÀÀÀø´†öÒÚ&Kï­½ôl—ÏyvÉgþ>¯üsï]å—)ºyOzrÆ•5Ë^÷ÞÍÔvÈ£9½vöc–“Ä«{Wrv¦Ä-QÝ‹sNúë}o—˜¬IaÍKì‰KÔÄœ·Ƨ5´—Ö6ùýmõÒ³]>çÙ%Ÿùû¼òO½w__å<ï4ÎQé;ˆ&ä5Ë^öÞ©_§tºÙêë,w–·Ù{ëyJg«[¢ãø\ËäMŒbñkVq$Äž˜ó–ÀÀÀÀÀÀÀÀÀÀÀÀÀÀÀÀÀÀÀÀÀø´†öÒÚ&Kï­½ôl—ÏyvÉgþ>¯üsïå˜)ŽcÀeŤ´fÙÁëÞ;_ŽX¢Tõ²rÌÉe9Ä[÷>®äØù-QÜÝ^b29<ŠÍ­YÅ‘x{bÎ[ãÓÚKk›,½·öÒ³]>çÙ%Ÿùû¼òßÞ~ÿýíwt‡;¾ÿYþòë1ž¡i«ÿGø¾%éËÃñýýøÍ/¿=¾ÿ¥¿üý×[gþ‡ï¯ÆÕ#Éy-£¸ÏVîTŸ³¼×µy31÷e}8cN>¯Íç•lÓº]^}ÅÍ^b5Õobß‹³kVq$.QsÞŸØ^ZÛä÷7IH/=ÛåsÎÚ%Ÿùû¼ò¨Í‹‰rÇz31ý¼6ÏNŽOÑŽV¿¿`š_³¼—µyö&÷/W­É㙡:'Wåoã­ë•ì½ aKô/™X›÷×­ÛÅqËüœ—Ø›8sÞŸÖÐ^ZÛd©Íµ—žíò9Ï.ùÌßç•ÿymžûîܨͽÿ@mMÊGŠÑÄ2Þ@1έYÞÀëÚ<™XªIúáŒ9ù('±è[W²w&•-‘³qþÁ¾Dow1›X׼ĞØÄ‘˜ó–ÀÀÀÀÀÀÀÀÀÀÀÀÀÀÀÀÀÀÀÀÀø´†öÒÚ&Km®½ôl—Ï9k—|æïóÊ 6ršyŽÝm¨Í‹T×)z“ÆíòÅš`×,oàumÞŒ ‡|qœŠ®sêo=Ìgzê•ìÖÉí3ÑWØÇJ~‰Õ„ö(æ²fGâAì‰9o ŒOkh/­mòûÛê¥g»|ÎY»ä3ŸWþ陞}}‹ãöð6îw/ÁĶfÙÁËÞ»x¹=–lê¨üu–‚>Ì'–ê•ì³çÙ%Ÿùû¼ò?¿]¼ŽcÅSë; ÷ÛÅ?°Pîx÷ñç÷™7+gŤ˜¥Ÿï＿åÖ,ïüeßÞœ z[»>«Tgiö×±äz%»*7¾_‰r>«ôÖ··ñá<ˆ­­YÅ‘x{bÎ[ãÒKk›üþ¶zéÙ.Ÿóì’Ïü}^ù§ûÌ[0^¿éçóD[2Õ¯Yv`_öÞãVöñÆ¡ë:K³ïVï­Wú—HyKä>ÿð>ólv“quÍ*ŽÄ&ŽÄœ·Æ'6¤—Ö6Yzoí¥g»|γK>ó÷yåŸ{ïlbçˆëÁè­ëÖ,;°/{ïbŠÜEæÙ0söMn7×½ë•æW«¯s¶rˆÌn_¢m»XM(k^¢$.Q:ï ŒÏjh/­m²ôÞÚKÏvùœg—|æïóÊÿü|•–L–›Ò£å??&¥™œÆíáYjs9;ÆÛ5ËxU›'kMócA c¥ÎÁ®Â}]©í1‘qåÁ¾D_wÑÉïÌy‰’¸DMè¼'000000000000000000000>§±zim“ßßd±ôÒ³]>çÙ%Ÿùû¼ò÷ÛÅe}ÕÇoV?vLhk–¼î½£q®/Èò7Y©sð똔u¥¿›¶DJrJøn/1˜XÅþúœUÔÄ.JBç=ñY í¥µM–Þ[{éÙ.Ÿóì’Ïü}^ùŸÞ.. l×Lé¯yçZ¨G¼üõ ³‹&ÿm ƒÜ›þ³;Ûåc·~ãôÎö*'ÉÌY>ë× 1ÁŽ{ù³þSó8zfÝÙ®WZXwÍ뜣±?z¨ì¥¿×±Å5«8b‹kÞŸØ°~µÉÒðk/=Ûåsž]ò™¿Ï+ÿ|g{3^>ÏãÀ™#9gJY³ìàeïí¬‰íˆÕš0~ÿBgyÔ¨ÕûÖç•l½ü6Æ•ÈÉ´=TöRÓ.:ùm‰9ë§1›8sÞŸØ^ZÛä÷7IH/=Ûåsž]ò™¿Ï+ÿÔ{»þ¥ìè¹Ëèæ]4-­Yvðº÷&×#ö¯òX9æäƒlhì]¯ä¾…ì·Dβ¥Ý¾Dw1šà׼ĞØÄ‘˜ó–ÀÀÀÀÀÀÀÀÀÀÀÀÀÀÀÀÀÀÀÀÀø´†öÒÚ&Kï­½ôl—ÏyvÉgþ>¯üÏOtqÎä¾%_åñŸ“â’Irðy3¶Ž7ЗÅ5Ëx]›gSå䘺~S`ÌÉÛU¸Ï+ãÀ–¼%²7éñVôKôa‹InÍKì‰M‰9o ŒOkh/­m²ÔæÚKÏvùœg—|æïóÊ?ß.^M‘ãÀ´ï}ÞÉìs–¼ì½;ióK1vñ¢sòYÎB{×+Ù&9©åJ̇“Ͻw3Ñ?ŠÅ®Y?‘x{bÎ[ãÓÚKk›üþ¶zéÙ.Ÿóì’Ïü}^ùÜ.^åkgM*Õã8¯…îiáó}æ}ßM!÷v¼ó(EýœåÛ—}{,¶Œƒ]t–w¸]×+ò)z·%R}qŸ¹÷&»G±¶5«8bmkÞŸÖp£—Ö6Yúví¥g»|γK>ó÷yåÞ·÷ýkÇ¥=ÿkúv](çºäœëâû¢|¤×ï4SÜšå¿îÛ³‰±o°ÍƒÔu–/_æc@õJöÖÔ=‘½|ÉÝ^b2Õîb1¶¬YÅ‘ØÄ‘˜ó–ÀÀÀÀÀÀÀÀÀÀÀÀÀÀÀÀÀÀÀÀÀøĆôÒÚ&Kß®½ôl—ÏyvÉgþ>¯üúö(ç§çŒu¨Í« rë¸3uÔæÁšf×,oàumÞäwb*&é[³üŽÃ¼Á}^ÉÎÉ—¸¹ÉïQìö«imû^|^³Š#q‰š˜ó–ÀÀÀÀÀÀÀÀÀÀÀÀÀÀÀÀÀÀÀÀÀøÄFH«M~“D³g»|γK>ó÷y城CÎ$}ì§Gׄ(gÖÌYvð²÷^ªñXݬüu–ÓaÂ<ÒF¯dyéžÈIξÙíKtuƒ‰iÍKì‰M‰9o ŒOkh/­m²ôÞÚKÏvùœg—|æïóÊ?÷ÞQî]—Ç~Ö:vPŒ¯k–¼î½³ÜP‹‡½ë<žKZçÞÇ9 ½Ô-Q‚q?¼ß;$Ê£˜ãšU‰1Ç5o ŒOkÄÑKk›,½·öÒ³]>çÙ%Ÿùû¼òϽw1ÍãÀãØc´&–5Ë^÷ÞÍ8¹›=š:ŽT×9¹"Õ½î}\ÉýŠµ["Çû£>—XMÊb kVq$ÄÖ¼%000000000000000000000>­¡½´mòûÛê¥uþåœg—|æïóÊÿüxð¾ ÿ-Ç$ÕûONúõ”¼iã-Æ`r^³¼—µyô&Ô#f?Ñ9¹«ò×+Ù69BýJä²~ÛàV›Çqú&y´çœU‰M‰9o ŒOlH/­m²ÔæÚKÏvùœg—|æïóÊ?Ý.£ ö¿åqÃz̦¦5Ë^÷ÞÉ$¹¾šRÆÊ1'ÙûzwãJöUþv%r’-ìöµ—w1ïÖ¬âHlâHÌyK`````````````````````|bCzim“¥÷Ö^z¶Ëç<»ä3ŸWþ¹÷.rŠÔçi4ïÉ×,;xÝ{÷·‘˜ÏVÆœ\¸ö>®d[-[¢¯(?<¼¿îü.6íš—èü.ŽÄœ·Ƨ5´—Ö6ùýmõÒ³]>ç¬]ò™¿Ï+ÿócRb6¥o)óÏIé_>ËS4½ì¢¿¾Ì‡5ËxY›'gZ:bIóyœ:'—V)>¯d×ÿÖ¶D¶÷[Ñ/1¸]ô&µ5/±'6q$æ¼%000000000000000000000>­¡½´¶ÉR›k/=Ûåsž]ò™¿Ï+ÿÚ¼ÉÁàR›çv¯Í?²0X¹Ñüç}{0µ¬ß'wžMôk–wþºoï_.1÷Kã@så†õ©Ž+Ù[ü–Èý-ýðx–eËb©kVq$Äž˜ó–ÀÀÀÀÀÀÀÀÀÀÀÀÀÀÀÀÀÀÀÀÀø´†öÒÚ&Kß®½ôl—ÏyvÉgþ>¯üÓñ,©›9ÅÅ£ÓS3Ù­Yvðº÷®&„#oš6æcN®ššçÞÇ•ìƒ|É+‘­œI³Û×^r{[Y³Š#ñ ¶²æ-ñ‰ 饵M–Þ[{éÙ.Ÿóì’Ïü}^ùŸßgž²ñ}K1 ðÓÛÅû×ñqÔäyÜ(Ÿ½®6ùýMÒKÏvùœ³vÉgþ>¯üÓíâ9˜¨Çûq,y_ØÚše¯{ïqpz,Ñ8mÌÇ,_©Î£eôJöÑIJ%ú–’°/Ñ–]L&¤5/Ñ–]‰9o ŒOkh/­m²ôÞÚKÏvùœg—|æïóÊ?÷ÞY q¹¿»è›qeͲƒ×½w‘»Þc÷£ËÊ1'å®sÝû¸"-¾/[¢Œ²·/Ñç]¬&Å5/Ñç]‰9o ŒOkh/­m²ôÞÚKÏvùœg—|æïóÊ?÷ÞÍT9¼çÇåÅ›×,;xÙ{glëï¦ÿmÜ«®srœ½·^ɶÍC`f"Gy“»½DkbzKX³Š#ñ öÄœ·Æg5f/­mòûÛê¥g»|γK>ó÷yå~±s&úÇ—~x»xßKŽbókVq$Äž˜ó–ÀÀÀÀÀÀÀÀÀÀÀÀÀÀÀÀÀÀÀÀÀøĆÔÍÚ&Km®½ôl—ÏyvÉgþ>¯üÓíâ}3®InP×Þ»š×,;xÝ{9ËE~`¾»1'çLXïn\É!ÎÛåg"'S~øX̾—v±çÖ¬âHlâHÌyK`````````````````````|bCzim“¥÷Ö^z¶Ëç<»ä3ŸWþ¹÷nò‚<3ŽÞ»öÍ…5Ë^öÞÕš¬ùÌ〓K&Îw§W²¯¦æ-‘³œ‰¾Û—hý.:ãÛš—hý.ŽÄœ·Æg5f/­mòû›$¤—žíò9Ï.ùÌßç•꽫7©)”ÙÍ×dœ_³ìàuïä䕘­‰q¬sr^f}wãJŽÎļ%r5Á=Ø—èÝ.FÙôœ—Ø›8sÞŸÖÐ^ZÛdé½µ—žíò9Ï.ùÌßç•)òÌ §ËÆÂ9Ë^÷ÞÅX9Å¥˜¦+Çœü¨òuïãJ¶Ù„°%r2íǽw–›ÌÄ\Ö¬âH<ˆ=1ç-ñi í¥µM–Þ[{éÙ.Ÿóì’Ïü}^ùçÞ»ÊfR̦Žû½›3É®Yvð²÷nÖxÄTLw³ëœ\_ÙtïóJËó~÷9÷½Ùø`¯½4“Ú£XóšõÓ‰±'æ¼%000000000000000000000>­¡½´¶Éïo«—žíò9Ï.ùÌßç•ê½›7.Žû»õÉš-šÜÖ,;xÝ{Ý«—²rÌòÍùÔÊy%Ûh|ÝÙÊ¡æ»}í¥Ô]ŒÆ¦5«8›8sÞŸØ^ZÛdé½µ—žíò9Ï.ùÌßç•ÜZ.ϱÔ;Î[‘½ç,;xÝ{g“íó¨îe嘥aw«÷W²rxË•Èqórｓie‹ñqÍ*ŽÄ&ŽÄœ·Æ'6¤—Ö6Yzoí¥g»|γK>ó÷y忽ýþûÛïþèw|ÿ³üå×#—{Xý?R¤§|äPd“ßßßüòÛãû_úËß½uæøÞù*÷®§¤¯?²uò<Í9Ëx]›7Sä!ŸÑýpÆœl–'{ê[W² ‰œLúqmތ˗˜­51¬y‰=qŠ31ç-ñi í¥µM~[½ôl—ÏyvÉgþ>¯üývqY_Ü‘B4ÉDãóše¯zïl½iõˆ¥™PÆÊ1'[MÑÍ畼IuKä.ý豘òzˆ»ä´˜9/±'6q$æ¼%000000000000000000000>­¡½´¶ÉÒ{k/=Ûåsž]ò™¿Ï+ÿÜ{GÓì‘|ß“;(&¦5Ë^÷ÞÙ8¹>™¢+Çœœ]ǤÌ+Ùe“Ê–ÈÕxû`/1™ÅêÖ¬âH<ˆ=1ç-ñi í¥µM–Þ[{éÙ.Ÿóì’Ïü}^ùçÞ»ÛÆqà~œsÒƒ9®Yvðº÷n&ÈÝìUNv‘½9¹hJž{WÆ–H͸ö`_{éÌ%ö½X»fGâ51ç-ñ‰ 饵M~“„ôÒ³]>çÙ%Ÿùû¼òO½·sÆËyÞ^Î)—yžæœe/{oçM’‡|†ñÀÎcÎò•æùÞóŠÜÄ^Ý–È}~´¯½4·‹AÞÄœU‰M‰9o ŒOlH/­m²ôÞaôÒ³]>çÙ%Ÿùû¼òϽw41çXí拱~Ͳƒ×½÷ˆÄœç1uN}Cqvöz%»&Ä•è+âžk)¯;»‹Ù„ºæ%öÄ&ŽÄœ·Ƨ5´—Ö6Yzoí¥g»|ÎY»ä3ŸWþ¹÷.&§#…*ù¾owk–¼î½«Ü”.§•ÛñLN“»Nq™WZ÷ªŸ‰TMÍö%ú¶‹MŽ{™ó{bGbÎ[ãÓÚKk›üþ¶zéÙ.Ÿóì’Ïü}^ùoo¿ÿþö»?ºÃßÿ,ùõHÆ¥ÃVÿTïµÙSãñýýøÍ/¿=¾ÿ¥¿üý×[gþ‡ïãË÷X ÁÄqK»&´5ËxY›{oœ<äÓə߲r̲Ñ8ouŸWZ6Éo‰ìŒýá1)Þ™XÅþ)ÎYÅ‘x{bÎ[ãÓÚKk›,µ¹öÒ³]>çÙ%Ÿùû¼òO·‹ûhlÁÌuì éºfÙÁëÞ;™àŽñ›ãhsSÿ[›·ºÏ+­_·D*/nï{ÉåQìÔœU‰±'æ¼%000000000000000000000>±ÑóM–Þ[{i9çÙ%Ÿùû¼òϽw1^ŽïŒ“ÒL)k–¼î½«IrL[wŠ9I•¿Þ^i^V^‰¾7÷ãÞ{lz›qqÍ*ŽÄ&ŽÄœ·Æ'6¤—Ö6Yzoí¥g»|γK>ó÷yåŸzï`åFðšiãžôàåÿÌYvð²÷ÎävÄfוcN>˜<ïf×+ÙŽʯDj&¤ûmÚEo‚_ó{bGbÎ[ãÓÚKk›üþ& é¥g»|γK>ó÷yåŸ{ïþ‚=RÌ&Œæ=d9$|βƒ×½w4UNqIÊÊ1'o×5וšäð–+‘ýS¿Dw1™äּĞØÄ‘˜ó–ÀÀÀÀÀÀÀÀÀÀÀÀÀÀÀÀÀÀÀÀÀø´†öÒÚ&Kï­½ôl—ÏyvÉgþ>¯üsïMiã<ï0Î9 MþÏœe¯{ï*GˆÇMucå˜åD•2ú9¯T/·¤_‰”äÁ›»½Ä"7¥?ˆÅ®YÅ‘x{bÎ[ãÓÚKk›,½·öÒ³]>çÙ%Ÿùû¼òϽw3MžcéæY,Ñ›Ö,;xÙ{÷ˆÏG²ÁäÑêϹ¥:Ï&×+Ùö-Ä-‘ª©åÁ^¢5Ù?ŠýMÌYÅ‘xk[ó–ÀÀÀÀÀÀÀÀÀÀÀÀÀÀÀÀÀÀÀÀÀø¬Æ쥵M~[½ôl—ÏyvÉgþ>¯üSïƒqåHýoÖ$Süše¯{ï(·Ë©-MWŽ9ùjìlõõJvUnI¿9šôãÞ;ÈolbÿjuÍ*ŽÄ&ŽÄœ·Æ'6¤—Ö6Yzoí¥g»|γK>ó÷yåŸ{ïlB>’ðuì šæÖ,;xÝ{“C7^Í)+Çœ|–£ZtïãJ¶Íx·%²5ñ‡çœô×û—ØÄj|Yó{bGbÎ[ãÓÚKk›,½·öÒ³]>çÙ%Ÿùû¼òßÞ~ÿýíwt‡;¾ÿYþò둤c·‡Õÿ#EzjGvIN`ùþ~üæ—ßßÿÒ_þþëÇZ9›üo[˜ë}ácKÿ‡ïý 5“äÁ›zü‘“7ήY>²—E}²¦úqÌL¿É srÙ¸ùC¯È=ò©l‰\Lqö%º¶‹NžG:ç%öÄ&ŽÄœ·Æg5f/­mòûÛê¥g»|γK>ó÷yåŸnPï닯üsïL £çNã&úTM¬k–¼î½‹ñöˆ¥ïU[ý1'ïä$Ýû¸2NY/[¢ÚûC>—8¾èƒXãšU‰±'æ¼%000000000000000000000>­¡½´¶ÉÒ{k/=Ûåsž]ò™¿Ï+ÿÜ{7ãäÁ›Uò}ÙIA>gÙÁËÞ;[Z7msÌ9ùp¦®W䌛¶Dnë`š{ïÝdÞD'7ÏY?‘ØÄ‘˜ó–ÀÀÀÀÀÀÀÀÀÀÀÀÀÀÀÀÀÀÀÀÀøĆôÒÚ&¿¿IBzéÙ.Ÿóì’Ïü}^ù§Þ;{ܑ䀕q¾wŽ²pβƒ×½w0©±ö룳×Yù™ô1›óJ¶YNY¿ÅšúÃqö½´¸‹Ñx¿fGbGbÎ[ãÒKk›,½·öÒ³]>çÙ%Ÿùû¼òϽw’'v¦Ô¿ð8ß;W©Üç,;xÝ{gSʳ3Mó1'ïMŠsï㊭¡½´¶ÉÒ{k/=Ûåsž]ò™¿Ï+ÿÜ{W“ë‘R•ÛÂûŠ3>®Yvðº÷n¦å#¦2vvÌ9¹&GµèÞÇ•ìÆYãW¢ï-ÿð¹–ýõà7±ô ]ó{â51ç-ñi í¥µM~[½ôl—ÏyvÉgþ>¯üSï]ÆQâ)¶yOz‰&ú5Ë^öÞeKš§¸èœ|”ê~ì]¯dM [¢„»½D/ÔƒXÚšU‰±´5o ŒOkh/­m²ôÞÚKÏvùœ‹vÉgþ>¯üsïäÖr9Ï;ŒçZ–b²[³ìàuïM§tFƽê:ËåqžM®W²«cïg¢4Sâƒ}í¥ØG±Õ5«8bOÌyK`````````````````````|bCzim“¥÷Ö^z¶Ëç<»ä3ŸWþ¹÷®2ËýÝqÜq^­©vͲƒ×½w“r>¦þ´³óx‚æ|®¥^ÉÁš–¶DßRvöµ—Ú6±ïÅ•5«8—¨‰9o ŒOlH/­mòû›$¤—žíò9Ï.ùÌßç•꽫31)ºyÎI ¦µ5Ë^öÞÕ›"§¸49šEVŽ9¹²žZ9¯äMõ[¢y9’|·/ÑÖ] òs^bOlâHÌyK`````````````````````|ZC{im“¥÷Ö^z¶Ëç<»ä3ŸWþÛÛ>þõÍkí±ÿõþþVB»ã?ôÿýåM±iÿø½“æÛ:ï‰jZØ:o ïåIÛWbÎ{"™àö„Î{¢šd÷„Î["8“÷ïræ¼'¢©ûg:ç=Qåž-¡ó–ˆÎøý3óžˆ?Ó­yO”‡ŸéÖ¼%’}ø™nÍ{">üL·æ=QŒß?±9o‰Ü¿3Ú?±9ï‰þÑþ‰ÍyOôïŒöOlÎ{B¢~%tÞE7s&æ¼'ô`¾+¡óžÐ?Âx%tÞßõ^ß_‰ßÿÁw׿þóÛïþèw|ÿó¡ßP¿üÝø«^ˆò·r„þ/qÿýýýøkÞýöøþ—Ÿ/í?|ún–†.-ãg·‡¥ñcK“Mò÷ÃÒôÁ¥ýûí—æ.íÿàÓmiùØÒlû …Ç¥õƒKûBôoÿ–¶.íß)û‡¥ÞZ]ú‡ïo±ÿl*ÿ$Ôä{@sÿO§:ÿ«ê)Æ÷Õ:o‰$ÿR®Äœ·ÄTÏÄöUúÚól›ÿ»íM…þ lÖÿzMm¼‘¿KŸÀ¿¿0õÿ„*ºÐþ5ë²kòŸ<².­­ÿ7첦~ô·_ÿtü—ã_þÃügæïÿÌÆùn¾ôÿ,oã?`å+ÿãoþÛ?þöø‡?ýÛÿøïÿëŸ~ù·ã¿ýïãÿéŸ~ýØÙþ½_ÿN&ôÿ˜Iú#àÿ~Zúê#(V¾åK¶Ê÷S²ô?õ¥ÿö?ÿ¥¿CÝÊö‰üMÿU?tÍŸÏÞrÿ̯ïÓ~Y³œ–Úc¿œ?’nó™õÂ54®ùšDðýŸþMÒüó¡ÐßÎã]{÷·Ól6wøþ Lÿ%Y~}—ÿУü0yßøíû,ŒÇæØ¿WöóÏåë…þ^‹×ëýçÎÛY°S³òýÏ®õ_vç¨ßR×7M^×q{õ¬g=ëYÏzÖ³žõ¬g=ëYÏzÖ³žõ¬g=ëÿ_?[T­?ßßf‹:ÛÐ5Îòs…oã ?ý1ë`œô¼e´Iî3š£|é—í¸QÌË}'ãOë(7eÍéôBs&—ëõÜÖ0¿´A~ûùÒ’ü®ç‡¦¯Ÿš¾®ãö:ëYÏzÖ³žõ¬g=ëYÏzÖ³žõ¬g=ëYÏzÖÿŸX/-ªÖŸïoY[ÔÙ†®q–Ÿ+|Wø© ÍÒûŒ×£:åyÎQ¾ôë§?ÉÍæ^þØÎè”u Vþ¤Øز^hYþháùº>ÞjwÏ]´°iUþÐɇ¦¯Ÿš¾®ãö:ëYÏzÖ³žõ¬g=ëYÏzÖ³žõ¬g=ëYÏzÖÿŸX/-ªÖŸrhãhQgºÆY~®ðm\ᧂ¶ýíÝMÇ–Þñ}ŠZJ‹Š÷—¥H ð˜ f3ð\]XPÏØÌ÷wœ8™QY¬!eÌ𿸟Š¿Îª+‘­‡©H«6²r”Gñr’ÇŒò¥Ÿ´e,xçå/}ŸF9z9 e·—Xål–;1—•Uwb.+o/kŒ^z¶Éo³—^íòÊ«K^ó×|Ì?ôÞ}“#²‹±rö¸sÁĶ²\ÁÓÞÛy©Ò] ÆëNÍÍ«ù|Å'k¢='|°«÷¿ôÞΙTïÅšVVqL܉5­¼M`````````````````````¼¬1zéÙ&Kï=zéÕ.¯¼ºä5ÍÇücï ã~ð¦÷{g“ëÊrÏ{ïd¢Ów£;5·íÝW¤Å÷áœð®ÉÑ+»}^K)»˜+«8&6qL̼M`````````````````````¼°áÂj“¥÷½ôj—W^]òš¿æcþ±÷.r”¸kIŽ —+h¦–•å ž÷ÞÕd{s¹_r;5—²vÎW¼o㉜kÂÛdÒ—ï÷î[ó.6ùfVqLl☘y›ÀÀÀÀÀÀÀÀÀÀÀÀÀÀÀÀÀÀÀÀÀxaCzim“¥÷Ö^z¶Ë+¯.yÍ_ó1ÿÐ{{k’ÞµFóîƒ\ÜÌrO{oïLi7‹ë²SsŽÇéåúŠ÷Q:øs¢VÓ¾x¿w_wi½‰~å%º´‹cbæmãe í¥µMþx[½ôl—<»äcþš×ücïL‘óÂíxnf¿‚,ë3Ë<ï½£| —’ÉqìÔ\âjÌç+Þ6㶠o)îÎ>Åw±ÿÄ­¼Ä>±‰cbæmãe í¥µM–Þ[{éÙ.¯¼ºä5ÍÇücïMm}ÝÎ'kúfb\Y®àyï]“ÞÊë²sæbÚê½õ•Væs=g®ÕÄû³Ã—Xä"ïÄjWVQ'vQ&4ï¯jh/­m²ôÞÚKÏvùȳK>æ¯yÍzûùýí§_ÝÍÝÞ“>ß’ÜrnoV"œ¿y7®ìýãöÃ?ÞÞïËïŸ/ù/ïo.ô÷Ro.5ãFm¼ÜW>³¼§µyp&È—³r¡²SslròËxëúŠTøÁoÅgïì%öìïÅÖVVqL܉}bæmã… é¥µMþx[½ôl—<»äcþš×üÃíâ!/‡¬s†dª_Y®À>í½£IQ/YwjŽXïN_Éuþ±ÁÌ©Ê“æ¯yÍývñàMê—d³ñª†öÒa´ÉÒ·k/­ù#Ï.ù˜¿æ5ÿxŸy39Ý\*ÆkŒ}Ý®,Wð´÷ŽÖ4?ÞM'³ÌœÆíãÚç+µÈéçD.rRúnŸ¢¯»èLÊ+/Q&NQ'4ï/j¬^z´ÉoG/­íò‘g—|Ì_óšÿ†ûÌ«©ý’Z6ínï_¾Ê*Ñøñ<јLh+Ëx^›G9¹\®Ov™Yv™(0_)}¢m¹ÊUìöƒ‰å^,ieubeBó>ñª†öÒÚ&Km®½ôl—<»äcþš×üÃíâ1ëorßzÄÄjR]Y®àyï]L°7'ýGªÏì‹< T¯]_)Á¸°M¤bÊ}§~^KÎ÷b‹+«¨»(š÷ Œ×5¤—Ö6Yzoí¥g»|äÙ%ó×¼æ{ïf¼Þ_טœ)ee¹‚§½w²&¶›³^ž(úq[9ØñàO¹öùJõrÌ9‘ü:@üÚ{79;}| 3맡§¨š÷ Œ×5¤—Ö6ùãM&¤—žíò‘g—|Ì_óšè½û`”»´«o²¯û°‹EZñ™—(§¨š÷ ŒW5f/=Údé½g/íV¯­yvÉÇü5¯ùÇÞ»š"w;'yle¿‚ìL+Ë<í½³•j\žžêØyäu€È|¥z¹–s"y¾x€Hj&ú{±Ø•õÓЉ]” ÍûÆ«ÚKû6™½´æ?Ž<»äcþš×ü§·Ÿßß~úÕÝÜíý7ùáó-—nöfõ'²Áú[ òãûÇí‡?~¼½ÿÞWß?_*ó_ÞÇWorw™›ÌQΟY®ß>m̓`Íɺlѹf¬Vâë•þ 9¿MÄþ~½Àþ™¹;°¶Õëw^m+oß© …¯ö´RGká;{Û#Ï–ö˜¿æ5ÿpvNÆå› EN)‘+(¦¸•å žÊÙÄxs6˜8žA9³«Ò¨ëµë+±oªÛD &¹QN¦Ú],Æ–•UÔ‰SÔ ÍûÆëÒKk›,½·öÒ³]>òì’ùk^ó½w5!§ýhÞ‹•;¨g–+xÞ{7ù͙ǥèl•ž]/]_éo&åm"ô¿´;ú¼”ÖN°_‰Ï3ª7ÖOOfÞ&¾[A _íi?ÞdB ßÙÛy¶´Çü5¯ù‡B¹8“bŸwr³\A4¶­,Wð´P.ÞT“'D¶Q†ÏìÚ:xc½«Éi›ˆ~½ÿK£Ü×]ÝÅ Ÿœy‰2qŠ:¡yŸÀÀÀÀÀÀÀÀÀÀÀÀÀÀÀÀÀÀÀÀÀxUC{im“¥÷Ö^z¶ËGž]ò1Íkþ±÷Ž²à|0E›÷b|]Y®àyïåLy7^wjvÞ”y'õ|%v2n!=¹“º$ʽ˜ãÊ*êÄ.Ê„æ}ãU í¥µM–Þ»Œ^z¶ËGž]ò1Íkþ±÷.¦É“½ ãàìjM,+Ë<ï½›q­¿›²Z}ÍNŸ 9®]_‰Ö¤°M„`\¹³—Xå6÷;±†•UÔ‰]” ÍûÆ«eôÒÚ&¼­^z¶ËGž]ò1ÍkþÓÛÏïo?ýênîöþ›üðù–ŒK7{³úÙÐÿ’«îýþqûáoï¿÷Õ÷Ï—Êü—÷þÕqrw’¯Ö¯¿“óÊrýO[óÚ/¨ÞJ6eþ[=ÎHÑ\ŒòÎWÙ°ËËOá<½þõüŒêõÍ3oß­àܬi¥ŒuïlmW\íš¾æ5þp vr¯µÔãI¯/›šV–/ÿ¼LN&•[kÆk =¢³ÎÄyù|Åe©ÛÏ çMüâÉ#ýRZÜÀl¼›Q½±¾yc`æmá»Fá;zZ©“g᫽í‘gK{Ì_óš,”‹ :çæÝÕÍW–+x^(W9îºTÓôÒGìïí¸òñ‚œ.RÒ9 §jÇûÓBNÏùÍs—çüî™· „ïT}¯Ö´o³î­íÛÑþŽm¼íð}^ãerߞ먯ƒ—/ß‚ñaeùòOËäæLK·þ£ÞU­qÜÖ<χž¯„j|Û&|Ïl¼=´É}=¸ ô&µ—Üî™· „ïTÐÂW{Z©“µð½í‘gK{Ì_óš,”ƒ©r…4Åã ²‰~e¹‚ç…r2.ÞJ›çdh”Û ó,Âç+Á›\¶ ßßþ— åh’½KQ½±~畺ò6€€€€€€€€€€€€€€€€€€€ð ZøjO+…²¾³·=òliùk^ó…r1V¡°r³\A3Ù­,Wð¼P®&„[­ãŽÛŒ}{óóÊÇ .Ž 8œoÆ–;ù¼’Üî¼VfTo¬ßy}`æmá»lž5­ÔÉ£î­íŠ«£]Ó׼Ưe²·Öx9-£êÙÉÞzSíÊò埕ÉÞ:“ü­Z)³eãˆÎ“êظ^éïÄåmÂS¾Ô&Ë¥Ôº^öiTo¬oÞ˜y›@@@@@@@@@@@@@@@@@@@@øn)|µ§ýx“ )|go{äÙÒó×¼æ?=ÊÁÄ0Ž·hZ(÷Ÿ´•å žÊÑw«Í476Žèl56ÍK×WB–¢Ï ßLúr¡¥?ÁdBšqy}}óÆÀÌÛÂw*há«=­ÊZøÎÞöȳ¥=æ¯yÍz(”³Ércq6¥+hÆ••å žÊÅ4{«ÎTíÂ5Öõ ¿ù‚‹Õ{8WÆ“o}r1>o^•CA4.ÏçÝ3oß©0ú^­i¥NuïlmW\íš¾æ5þé¡Ln¦ºq7r}·ó&ä•åË?-“3¶ÝZ–eãˆÎvw1_ ãìsÂ[yáN/К˜îÀfTo¬ßy}`æmáûfá«=íÇÛ*|go{äÙÒó×¼æ?½ýüþöÓ¯ý•ÛûoòÃç[2.ÝìÍêOdƒM·¤€~ÿ¸ýðÇ·÷ßûêûçoÙçb‘Û¢/ïKì_ÞûezÚ¹lBo<™”V–7þ¼ÇŽòNúgXÇñånê0úùJpÆÖmÂg9d§LŽw`ó3ª7Öï¼>0ó6€€€€€€€€€€€€€€€€€€€ðÝ RøjO+=¶¾³·=òliùk^ó_ï±ç<§fjü†>:'·C7“õ TSâÊòž÷ÑÅÄr+ë|­?éÐ\¨ÒÃRÐÇt'/o|ª§Ws3ª7Ö7o ̼M |·‚k³¦•6zÔ½³µ]qu´kúš×ø§ë!®™PoÎ9ãG]î<;pfùòOËäþòx¢wEklÑäùA}ÁÅ(uû1 7ô÷gIŸžõ›çdNãòúúæ™· „ïSоWkÚ7haµ¶+®Žöh‡/yº–ÉÞËÒ=§q"µORsÏ,_þy™LM·ÖŒ/cãˆru.óé+ÁË ÓçDÿ %ÞÑ'èÝFëŒËón÷ÆÀÌÛÂw*há«=­ÔÉZøÎÞöȳ¥=æ¯yÍý¤ ïä¶èÒ·µ¯”!_&Ë­ÏMom®rδF¹úçet16JËí´Å֘׾ù‚|!œÎ5)ßwyyY˜¸{ýÝjTo¬ßy¹¬¼M |§‚¶½£¦•*ZÛ^mmW\íš¾æ5þéáÎæjš<óÏÎgg’]Y¾üÓ29X9y£¸ùXErõ|¬¢¾ Q´ñ{¦½»“ו4é¾w¯æõcëw^˜y›@@@@@@@@@@@@@@@@@@@@øN…Ñ÷jMûñ6ëÞÙÚ®¸:Ú5}ÍküÓµLÞ8¹_¸H%-_>šÜV–/ÿ¼L&Ž;¶“nQkó4éùJ”=ÛD¿ üÅ'øõK)u£±iFõÆúæ™· „ïV…ïèi¥Nž…¯ö¶Gž-í1ÍkþÓC¡œä.h缩zÅÔº²\ÁóB9›lo5š2žN¨ÑYk켫z¾"C§mÂUÓÊ}^J_8Á"gSkTo¬oÞ˜y›@@@@@@@@@@@@@@@@@@@@øn)|µ§•BY ßÙÛy¶´Çü5¯ù¯Ÿ»¢<ÂÏÅjÜ7¼ªI~ÜЬ§yDglYYÞÀó>º™Òn¥ÊÃýäèl5ažI=_IVŽë8'‚•;²wúû×<ÀhM 3.¯¯Ÿž̼M |§‚¾ÚÓ~¼­Âwö¶Gž-í1ÍkþÓõ羿È1I–+ˆÆç•å ì³B9zÓ¤&Ž³@f–³6fÛ>_餫ۄ\ÒýÑΧâ.“ýÊK”‰SÔ ÍûÆ«ÚKk›,½·öÒ³]>òì’ùk^óŸzïhšœ¾1þ"WPLL+Ë<ï½³qÒÀ;SËØ©ÙãVï­¯ToÊ>Ñß´ wö“Iá^¬neubeBó>ñª†öÒÚ&Kï­½ôl—<»äcþš×ü§‡Þ»Ê³]ð³›OÖ串\ÁóÞ»™  ¼Õk×ìíÙ{ë+yœ©}NÄpíÔÏk)~ûµX»²Š:qˆsBó>ñº†ôÒÚ&¼É„ôÒ³]>òì’ùk^ó_?w;Ž;ë3Ñ}ýØí䌗SVš‰ãœ–L +Ëõ?mÍ“7)?Èã—3gÂ|Èå|¥_‰ÍÛDÿË—ïï×ÒÜ.99|fuâuBó>ñºÆè¥G›,­ù쥵]>òì’ùk^óŸ®w‹§(·“¹Ç))©ëW–+xÞ{'Sâ­9ùD6Ž(Å ó€—ùJ«RÀŸéáI–'èìf‘4.¯¯oÞ˜y›@@@@@@@@@@@@@@@@@@@@øN-|µ§•BY ßÙÛy¶´Çü5¯ùO…r1YŽ/IòTÊ~ÙïV–+x^(WÓÂM6IãYš3‡,Y¯]_É£5?'’¾¼=6ÊÕø¶‹Í¤²òeâuBó>ñª†öÒÚ&¼­^z¶ËGž]ò1ÍkþëǤlj¿¤ÒŒ÷_?¤ùGMîÇÈÁ„¶²¼§µ¹±"g³äy&ù̾­Û¬×+ýÃðq›H‰wö‰õ^,yeubeBó>ñª†öÒÚ&Km®½ôl—<»äcþš×ü×kó¾ÁÖ[rg÷×[óh¬Ü®W×7g“êÊrýÏ[ó$¢8ïåÐqÙ©9f“Wk®¯ä*—uNÈ—ürkM.÷bK+«¨»(š÷ Œ×5¤—Ö6YZsí¥g»|äWK®ó׼濡5ã˜oMø–Ú¼È=érÈw«ã 4SÊÊòž×æÕ$Û?ŒþŽõÃÑ,§»¬G_)Eû>'ú;*÷ç¿,±˜šw±ÉCgVQ'NQ'4ï¯kŒ^z´ÉR›Ï^ZÛå#Ï.ù˜¿æ5ÿéz6y±r º+nœ}~óÅ›–W–+xÚ{÷‘,'ÐTSGå?s¬Çƒ:õo£‰ûDsÒäïö)Ú´‹Þ¿òûÄ&Ž‰™· Œ—5´—Ö6ùãM&¤—žíò‘g—|Ì_óšÿôÐ{“õl;ŽI)YnŸY®àyïM•×¯.;5§(·›ëµW¼ rçù9Q“É÷—Ÿ¢»˜Lr+/±Ol☘y›ÀÀÀÀÀÀÀÀÀÀÀÀÀÀÀÀÀÀÀÀÀxYC{im“¥÷Ö^z¶ËGž]ò1ÍkþÓCïM‘{°ûWwœ—&?™Y®àyï]-CŽúî4Ë[X×>^ñ¶“öœðÖJý½ÛK,&†{±Ø•UwbŸ˜y›ÀÀÀÀÀÀÀÀÀÀÀÀÀÀÀÀÀÀÀÀÀxYC{im“¥÷n£—žíòÊ«K^ó×|Ìzè½ÇAÞ®5ãÇT/ƒ3Ë<í½«3^Ïh±ã^õ™k3q^»¾âC3ÙŸÞµ'ŬVïÄÚVVqL܉}bæmãUÙKk›üñ¶zéÙ.¯¼ºä5ÍÇü§kï]ƒqrvšî¬IŸY®àyïMŒãD•ª;G–Æ=γÉç+¾™Z¶ ïŒkwöy-ÕíbÿjueÇÄ&Ž‰™· Œ6\Ym²ôÞ³—ÖvùȳK>æ¯yÍzûùýí§_ûUÝÞ“>ß’”boV"Ez´7Ÿ¬¸ï·þøñöþ{_~ÿ|éÌyï|6AnéΦiµ_Ms+Ëx^›“õVw?¨9³k&Í›áç+É™˜·‰èówö)Z»‹UÎv™y‰2qŠ:¡yŸÀÀÀÀÀÀÀÀÀÀÀÀÀÀÀÀÀÀÀÀÀxUC{im“¥6×^z¶ËGž]ò1ÍkþÓÃíâ}!Ý|ˆ¦Œâ¾õve¹‚§½w³¦ú›ëâ8Ú\³<ÒÓÎʾ"'”ûm"¶u€ø¥÷îë®íâøf^bŸØÄ11ó6ñªÆ쥵Mþx[½ôl—<»äcþš×ü§kïÝ÷9º$ÎÊ[’gvÎ,Wð¼÷¹<ªzÙ9²·y=Ps½’ú%Æm"úÕû_{ï`B¹sZYÅ1q'ö‰™· Œ—5´—Ö6Yzoí¥g»|äÙ%ó×¼æ?=ôÞÉ´póÁ«Í{5±®,Wð¼÷.òäL¹{½èNÍrgù|,æ|¥¬‡~Îœô‘·ÇÞ;›”ïÅWVQ'vQ&ÊùQ Œ7´—Ö6Yzoí¥g»|äÙ%ó×¼æ?=ôÞMDñ>/Í{°Îä²²\Á³Þ;Xk‚ÜÍnõnö™åÎòèÆÎõJlr›÷9ãѯ½w“¼‰ÎØ°²~cbÇÄÌÛÆ ί6ùãM&¤—žíò‘g—|Ì_óšÿt齃õ&Èé$Ö„4® ššW–+xÞ{“êÍ¥£1×\ûµëÎùŠIžÌyLxçMúRï-×Òâ.FãýÊ*Ž‰M3o/l·Údé½G/½Úå•W—¼æ¯ù˜콓‰­Ï[ãó¸‚*ƒ3Ë<ï½³<úÒÕ G³ÈΑ½urt‹^»¾Ça.ç„«&ÝŸ¡rŠ.ìb‘²~æ%º°‹cbæmãe í¥µM–Þ[{éÙ.yvÉÇü5¯ùOo?¿¿ýôk¿ŠÛûoòÃç[2.ÝìÍêO¤H/ý¢‹3±ÜÞ?n?üñãíý÷¾üþùÒ™ÿò>¾|–[¸‹¬ô7àœtú3Ëx^›7Ó¤ÀϦùñÖG–Í£Ÿo]_ Ö4·Mx/íùnŸbð›è¬Évå%¿‰:1ó6ñ²†öÒÚ&¼­^z¶ËGž]ò1Íkþávñ¾¿ÊMÞuå⢠Î,Wð´÷vAjvW’>PsåL™×®¯øXäžøcÂû Mþn/Ñ›äîÅÒVVqL܉¥­¼M`````````````````````¼¬1zéÙ&Kï=zéÕ.¯¼ºä5ÍÇücïÝ/FîÁvú`ÎàŠÉne¹‚ç½w6!v¹™PÇΑ}¿ ¹s½«œy~N„$;vû¼–bïÅVWVqL܉}bæmã… é¥µM–Þ[{éÙ.yvÉÇü5¯ùÇÞ»Jöý ÛqP‹·¦Ú•å ž÷ÞÍ$9O¥™<Ÿ¹“æ/úŠÙävNxŸåKìöy-µmb¿WVVqLœ¢N̼M`````````````````````¼°áÓj“?Þd¢Ú£]^yuÉkþšù‡ÞÛ;ãè¹ë¸'ÝÓÚÊrO{o9JÅ÷wM'ƒk–RÜ<çd¾£É~›8ßÝ¥÷îë¶îb0!¯¼D[wqL̼M`````````````````````¼¬¡½´¶ÉÒ{k/=Ûå#Ï.ù˜¿æ5ÿéíïnÿ÷Íkímÿñ¿ÿçK ínÿ¥ÿï÷7'ë·ÿú–qí–­<™R.tTØå:ï—5¥(7¯Õ(wŸ«#m«£“?VG:W˸®µªi[ r"̱:Ò¶Zä©£ÇêHçjµr¸úZÕ´­9‚æXi[Ír£þ±:Ò¶Úä3ŽÕ‘ÎÕäÿôµªi[ÍÆŸŸ†¦mµ™˜ÏÕ‘ÎUg½â¾Þä?48×GÜÖ7þüÈfÜד<Öõ\q_¯r~ý¹>â¶Þÿn-ç7ã¾ÞZ¶õ÷õþ·yÞÖGÜÖƒ3aû|4îëýÍíóѸ¯yfì¹>ⶭݬkÜ×£ücx®¸¯ùÏ2Îõ·õdåø¢c]ã¾ä4ÿs}Ä}=Ë©GçúˆÛz_±Ûç§q_ò§fçúˆûz–_mÎõ÷õfòöùiÜÖ‹—?n;Ö5îëÙØíóÓ¸¯·ý—¡·õê÷_ˆfÜ×ÓþKÑŒûz5eû|4nëÍ™¶}>÷õdÜöùhÜ׫ Ûç£ñ\—§§óó™q_rœÕ¹>â¾^L³ÛúˆÛºsÆžŸßŒûz4¾në#îëE¦|®¸­{+ä{¬kÜ׃©i[q_/ÆÆm}Äm=Øýק÷õ°ÿú4㾞÷_ŸfÜ×ÛþëÓŒÛzôòÓc]㾞åϵÏõ÷õ&¿ÿŸë#nëÉË·ǺÆ}=™²}>÷õjZØÖGÜÖ³7nûü4îëÉ„íóÓ¸¯W“¶ÏOã¶^œ|[r¬kÜ×£©Ûç§q_¯òŸ¡ë#nëÕ¿}~÷õ¸ÿú4ã¾^ö_ŸfÜÖ›Ý}šq_û¯O3îëÅøíóÑx®Û¿ Ù¾ËÒ¸¯÷oCê¶>⾞å±.çúˆûúx²ø¹>â¶îÆXŽuûú8©î\q_ÿEß¹>ⶾÛx|y®ÿüþ…oM?ÿíí§_ý|Žûøntüð—ñ£¾Ðìôûÿ›ò$w×¼“‡¹ucÿ¤_ɾ1|ÛFçÆ/MûÎø;c’oa÷éwöo`s¼Û™¿m§ïÿG§ûåwF'úÙwÖoÜÙ¿ïßPï;Û·í ýÛR¿ïôÖêÎ_Þßú/‰}D~a’ÿ„V“ü}\ôߨ.Ëa|;;Ò¹*÷ÅcUÓ¹ê‚ׇ-éòŒÛzÿEÍŸøŒçºïÿPæt¬Ï¸­÷÷eOÆm½DýσçºÆs}¾Ùµ¾½÷þOÏñÑnç_¶¹C+¿áöoËêødÿ’ïÿù7·õïfªþò—ø'¶¹þ»´wºÏÿ™}ýw—þ»‡ì³b›<ʪS;þ¾ù3Ûúïõ>ŒmæÍßÂô"ÿÌG)¿27½ÈºþÎîÿb~Ûæòþ—ϽýÛ?ý¿>†ùO”¿þåõïé(eãßËå«þýÿëï¼ý§¿ýõö·Ïÿü¯ÿ§+ð—¾íW¡þMsÿ¦°ÿ«]Ò¿þç_ÿáó7½õþÓþݲ‹U¾)•ÿ­ïü—þ§þîþåÿã¿þÃÛ'ñÿSQÌöõ7—·Ðßóñ½î+ʹ¯Vnœÿ0ÝÇ5¸;zV¾ÉÜÁþ=FõÆú×fÞ&¾Oa¾ÚÓ~¼­Âwö¶Gž-í1Ík~þi›Uþíýó-ÉÕ؛՟ȆþWËúCQ¿÷å÷Ï—rZþÔÍã측j<-ú$ª5³¼ç}ôøó«þeÚøC-rgÚü3ßõJ)òÇ}çDÌra;½À 75œ`’??Ó¨ÞXß¼10ó6€€€€€€€€€€€€€€€€€€€ðÝ RøjO+}´¾³·=òliùk^ó×óV£ÏÆ·›ëLÿ1„¯²qf¹‚ç…òøÄ[”]6ŽèÜq˜ìz%GùÏ½Ï ùmïav^J Xå?ÊÒ¨ÞXß¼10ó6€€€€€€€€€€€€€€€€€€€ðÝ RøjO+…²¾³·=òliùk^ó…r“CmäŽä6.1øqAç7ÐÉ)^—çüî™· „ïS˜…¯ö´o«ð½í‘gK{Ì_óš(”ûþ\nNž2Ž`ãTú™å žÊã1TÍÉÃ[dãˆò°¥yvózEn·öÛD,«r¾ÊANù8Á(“Æå·{c`æmá;´ðÕžV e-|go{äÙÒó×¼æ ådªéœæ3ýB•§OÏ,Wð¼P.ruOs–#:åzôÒõ•ê䤋s"ùëñÌ ÌrÍ;XÊŒêõ;¯̼M |§‚¾ÚÓJ¡¬…ïìm<[Úcþš×üc¡ÜŒMãŽä4ÊÉve¹‚§…r´&ø[ Æ&Z£³í8Cy¾’‹©i›HýÍ}ùÈ‹¾µÝ-ϨÅX¿óúÀÌÛÂw+Há«=íÇÛ*|go{äÙÒó×¼æ åس<„/È…ÈDSÚÊrÏ å`’TÛÙ„Q†ÏìÆ×>_ÉÉÔºMôK.÷öy-µìb4.­¬¢Nœ¢NhÞ'000000000000000000000^×^ZÛdé½µ—žíò‘g—|Ì_óš콓‰ò¬À(xÈÓêÊrÏ{ïlŠ½µ².}DysmÝH­¯ä¼hÍýŠÛOæèë6o`1!θ<›wo ̼M |§‚¾ÚÓJ¡¬…ïìm<[Úcþš×üc¡\M–gFSÇiÔÉWV–+x^(7SÛ­ØqdÈmÆqÇöÁ~i˜¬IaÆåõõÓÓ™· „ïTÐÂW{Ú·UøÎÞöÈQ[Úcþš×üC¡Ü÷W;ž5èµPŽ&ä•å žÊ©¿0„XtãˆÎgÓôØõJu&»m")ÑwzÞÄx?£zcýÎë3oß© …¯ö´R(ká;{Û#Ï–ö˜¿æ5ÿéíç÷·Ÿ~u7w{ÿM~ø|Krʆ½Yý‰l°ñæm5¥ÞÞ?n?üñãíý÷¾üþùRFÿò>¾|wåQƒV+óbRZYÞÀó>:_nÎ%SÆ1Ø3û:z8Þº¾R¢q~›ˆíÉ‘)™îÅæVVQ'vQ&4ï¯jh/­m²ÔæÚKÏvùȳK>æ¯yÍCmålW«©ùjójœž+£¼lM‰+Ëx^›7óÍÙjÒ8ÓdæàŽ›Ôç+ýJ¼Û&ú—L÷öû%ûMì×âìÊ*êÄ!Î ÍûÆëÒKk›üñ&ÒKÏvùÈ},éq&:Íkþáø’ìäi2_ô ‚i~e¹‚§½wö&Kïmc§æ˜MœG¯è+ÞZcÓ6Qƒ4ý»}ŠÖíb0¾­¼Ä>±‰cbæmãe í¥µM–Þ[{éÙ.yvÉÇü5¯ùÇÞ;š$O¢tr¯¸\A‘ó¾g–+xÞ{'Sc7Éd;vjý+ÍV__ñ.¬w§%H¹¿Û§èí.Ž‹žy‰}bÇÄÌÛÆËÚKk›,½·öÒ³]>òì’ùk^ó½w‘îÙ¥$/÷+(VÎS™Y®àyïÝŒ *£ÓWnRåëµW¼¦Åm¢¿É|ôø« í^ÌeeÇÄØ'fÞ&000000000000000000000^ÖÐ^ZÛä·ÕKÏvùȳK>æ¯yÍ?ôÞ}ÿR.{“Æ9'%˜ØV–+xÚ{/G§H‹ŸÆµÏÃñîæ+ÍÎþ˜(qór齋3©Þ‹5­¬¢Nì¢LhÞ'000000000000000000000^ÕÐ^ZÛdé½µ—žíò‘g—|Ì_óšì½£qr¶•ãRä ²Éue¹‚ç½w2QOqч~Κ)«÷¯x›L+Û„~ÉÝ>¯¥”]ÌÆÆ•U›8&fÞ&000000000000000000000^ØpaµÉÒ{k/=Ûå#ëØÇü5¯ùÇÞ»˜ Ï±´z¿w3µ¬,Wð¼÷®r\‹ Å4Ý©99y¦^ûxÅÛ`jØ&J0>ÞÙçµ´¼‹Íø°²ŠcbÇÄÌÛÆ £—m²ôÞ³—ÖvùȳK>æ¯yÍ?ôÞÕÊ&Nn ×Xƒ±ye¹‚§½wu¦´qgxwŠ¯ìÞ{¾Ò’IûDÍr“ûnŸ¢K»èMô+/Q&NQ'4ï¯jh/­mòÇÛê¥g»|äÙ%ó×¼æ?½ýüþöÓ¯ý•ÛûoòÃç[’K°7«?‘"½Ä›sÕ¸x{ÿ¸ýðÇ·÷ßûòûçKgþËûøòEnò.&kmžedfyöimåžv9ì¼jΛÉóføùJ-rGû9‘û—ørmMˆ»˜ä"f^¢Lœ¢NhÞ'000000000000000000000^ÕÐ^ZÛd©Íµ—žíò‘g—|Ì_óšÿzm^½iáæZ11}CmžM•{Ä‹Iã$˜ÚLŒ+Ëx^›Wãòmœ•®b 9õë÷Òë+ÞÊ6Ñ/0·;{‰ý‹†{±Ú•UwbŸ˜y›ÀÀÀÀÀÀÀÀÀÀÀÀÀÀÀÀÀÀÀÀÀxYC{im“¥6×^z¶ËGž]ò1Íkþᔕf•§hFiµû´¾1¬,Wð´÷n΄4þÀŽw7³œK>Ÿ:_é_BŸ*:s¿¶ðÅÞ»ÛðïÄÖVVQ'vQ&4ï¯kH/­mòÇÛê¥g»|äÙ%ó×¼æ¿ávñfB_vÍÄüõÛÅ[0^7or¹¼dª_YÞ€}Z›Gyrì¹þ¡ÀÌ9¯›Áç+ÞZSÓ6Ñß‘Kwöƒiv“œ,3³ŠcbÇÄÌÛÆ ¾¬6Yjsí¥g»|äÙ%ó×¼æoÏR»R乂f¬[Y®àyï]L‘?ÊW’šsY×>^ñ®ï·‰þ%ç å×Þ»Ûv±Ê]ó3/±Ol☘y›ÀÀÀÀÀÀÀÀÀÀÀÀÀÀÀÀÀÀÀÀÀxYC{im“¥÷Ö^z¶ËGž]ò1Íkþ±÷n&ËÑ%Á9ß<Ù¾nW–+xÖ{'kMó79€%§±Ss¦ËÎùŠ¼9ý˜*þþ^òSôuœQ>óûÄ&Ž‰™· Œ5V/­mòÇÛê¥g»|äÙ%ó×¼æ¿~LJ«¦öKòÙøøÕcRäËW9<›¤µy’ãYf–7ð¼6ÆõËHÉÔ2vj.u=rt¾â}3-Ÿr3¼»·—L,÷bI+«8&îÄ>1ó6ñ²†öÒÚ&Km®½ôl—W^]òš¿æcþz»x²ã¨sß™ÆT)¼g–+xÞ{ä˜8nd¿­\ãºÑ|¾â}Ð#XVvþYïåO îÄWVqL܉}bæmã… ëW›,½÷è¥W»¼òê’×ü5ó½w3^ŽY錕+pΔ²²\ÁÓÞÛYå˜*çËNÍ5Œçyʵë+>&ë9!×Ó}^KM»èŒ +ë§1&6qL̼M`````````````````````¼°!½´¶Éo2!½ôl—W^]òš¿æcþ¡÷–A½ÛêD¹1|f¹‚ç½w0¹ŽhÚÑÙÏÜüq7»¾âC“3Ð ï£\ònŸ¢»Mð+/±Ol☘y›ÀÀÀÀÀÀÀÀÀÀÀÀÀÀÀÀÀÀÀÀÀxYcôÒ³M–Þ{ôÒ«]^yuÉkþšùÇÞ;™$ç¤d9e\® W–+xÞ{g!]m&ëΑ儕fçµë+Ñɽçç„÷&ä;û}ØÅb’[y‰>ì☘y›ÀÀÀÀÀÀÀÀÀÀÀÀÀÀÀÀÀÀÀÀÀxYcöÒ£M–Þ{öÒÚ.yvÉÇü5¯ù¯>SŠôÚ/)eÓÊ׋ٿ|‘cVª ãxpïä'3ËxZ›{+í¿«Öx?vjne=Ps¾â‹3Õ>Óܽjóf¢¿‹]Y?Ì1q'ö‰™· Œ—5F/=Ûä·ÙK¯vyåÕ%¯ùk>ænïû›Üä]æñà>šäW–+°O{ï ’kÖTÝ9óƵë+¾d¹'þ˜ð!­gÚûÞÛ{“ݽXÛÊ*Ž‰;±O̼M`````````````````````¼¬1zéÙ&Kï=~²Úå•W—¼æ¯ù˜콓<ÓËù,ãs_Lq+Ë<ï½³‰q¼§ïndoÏ£Íç+u< óœMÊýÝ>¯¥Ú],Æ–•U›8&fÞ&000000000000000000000^Ø^ZÛdé½µ—žíò‘g—|Ì_óš콫 òÌ&t¿‚0ÎS™Y®àyïÝL–C`’ܲ.×>ò8eÞ«>_Iy<Òó˜ðÁ¤û3TÎkimûµø¼²Šcâubæmã… é¥µMþx“ é¥g»|äÙ%ó×¼æzïàL’ãÅ­ÌËDy"çÌrO{ïàMÕ;Ïõðó•ÛñîôŸÊlõg–7ïìStuƒ‰iå%ö‰M3o/kŒ^z¶ÉÒ{^zµËoGO­]òÛÑcßçcþ±÷Ž¦ÈùàN w¹‚b|]Y®àyïu7W²Éã„ÍÞºóÚõ•bkÛDHrI»½Ä$‡«Ü‰9®¬â˜¸ûÄÌÛÆËÚKk›,½·öÒ³]>rÐ.ù˜¿æ5ÿØ{Óü¸¿»ŒÆ£5±¬,Wð¼÷nr‘r'yÒkŸ¹žïN_É~½;ÍÑIý½ÛK¬&å{±†•UÔ‰]” ÍûÆ«ÚKk›üñ¶zéÙ.yvÉÇü5¯ùOo?¿¿ýô«»¹ÛûoòÃç[2.ÝìÍêOdCÿ‹oNÎ$yÿ¸ýðÇ·÷ßûòûçKgþË{ÿòÎ8¹É»§µy09¯,oàim½ õæòødçÈòáÌë•MiÛDÈò—Ý^¢“¼‰ýKú•U›8&fÞ&000000000000000000000^ؽôh“¥6Ÿ½´¶ËGž]ò1ÍkþávñMÇ`ÖYÜÇljZY®àyïL’›á‹iãfxÍÞfiþõÚõ•äŒÛ„o¦}¹÷Žr¶Ë&fãÝÊ*Ž‰M3o/lH/­m²ôÞÚKÏvùȳK>æ¯yÍývñ8¨é‹—+¹Ü.þûüú}æãtïÓ<–òì’ùk^ó÷™÷ý¹Þ¼óÒj÷+HÁø°²\ÁÓÞ;9ÓR7M‚‘#{Û¿À¼v}EÞy>'ú_â}§~ŠÁí¢—slf^bŸØÄ11ó6ñ²†öÒÚ&Kï­½ôl—œ´K>æ¯yÍ?öÞÁÔró.™ÆdýÊrÏ{ï$G¶¸æå€Ù9²4ìež¯2_ÉÉd»MkZ¸³—¥î¿K]YÅ1q'ö‰™· Œ—5´—Ö6Yzoí¥g»|äÙ%ó×¼æ{ïbl¾yåFs¹‚f²[Y®àyï]M7W³Tõ²sdo‹I«³×WZ4Õn±_R¼³ÏkÉí^leeÇÄØ'fÞ&000000000000000000000^Ø^ZÛdé½µ—žíò‘g—|Ì_óšè½³5^ÎóΦy¹‚ìåKÏ,Wð´÷ÎÎ$9ÅÅš<îU×,'¤8½S|½Rª©~›Ð[Ôwû¼–Zw±SyeÇÄ&Ž‰™· Œ6¤—Ö6ùãM&¤—žíò‘g—|Ì_óš콃‰áæCgÆY,9™ÖV–+xÞ{GSÜhñ‹^ûÈò•Úºv}%—ù´Ð™S¿¤zgŸ¢-»8[™y‰}bÇÄÌÛÆËÚKk›,½·öÒ³]>òì’ùk^ó½w6Ùß¼/&è56ãÊÊrÏ{ïbší年Fþ6³wãQzíúJÿK'ωh¥þÞíSôy«Ü¤>óûÄ&Ž‰™· Œ—5´—Ö6Yzoí¥g»|äÙ%ó×¼æ{ï&7v{ï¤ïWP¼ ye¹‚§½wqÆÊ).Á¸qŠ‹fo«ãÚç+ÉÉ!æçDˆ¦å;{‰ÖÄt/–°²ŠcâN,aåmãUÙKk›üñ¶zéÙ.yvÉÇü5¯ùOo?¿¿ýô«»¹ÛûoòÃç[’ÞÞ¬þD6Ø~ÑiœÀòþqûáoï¿÷å÷Ï—Îü—÷þåGî]–ãÄå $“ÒÊòž×æQîHwÕ›¢ÎÈÞö¯1oŸ¯ä,pNÈ#;Ë½Ä çÇ܉ͯ¬â˜¸ûÄÌÛÆ ÒKk›,µ¹öÒ³]>òì’ùk^ó·‹—,¾÷~ÞÒ^ª)qe¹‚ç½w1±t9Í#^fvÞ„õÇúJÊrBù9!•þ½}^K »Xs+«¨§¨š÷ Œ×5\[m²ôÞÚKÏvùȳK>æ¯yÍývñ’Æy⥘¾ávñfBí|3yœóRl›YÞÀÓÚ¼Z“å '¦ìÙÛ&¤·>_ MŽ’9'‚•G‡îö)Z¿‹ý'må%ö‰M3o¯jÌ^ZÛä7™^z¶ËGž]ò1Íkþjó:Ž#ÏÞ¤o8e¥ö1½G¼OËHrŠùÌòž×æÁT9Í%›àÆNÍ­G¯ë+>zÙyLxßL¾·OÑ»]Œ&Ö•—Ø'6qL̼M`````````````````````¼¬1zéÙ&Km>zéÕ.¯¼ºä5ÍÇü×kóêÆiæÝMñjódŠ<Ã3I£-o Ê{›YÞÀóÚ¼È=÷®Úqýmfïœ)ópòùJr〙c"“Û½Äl¢½sYYÅ1q'æ²ò6ñ²†öÒÚ&Km®½ôl—<»äcþš×üÃáä}“‡pZSÆíâmæ¯yÍ?ôÞÍ'á´¦‡¶(…ùÌrÏ{ï`¢»99•EßÝÈòøO»zo}%69qýœðýMñtñ~-¥îb46­¬â˜ØÄ11ó6ñ†ôÒÚ&Kï­½ôl—<»äcþš×ücïL§h&Sµy/¦Ö•å ž÷ÞÙd¹›½˜6îf×, ódôùJ®ó •™C’›Üwû¼–Vv±Hý?³ŠcbÇÄÌÛÆ ÒKk›,½·öÒ³]>òì’ùk^ó_?¼EiÌ}÷fõxðVMòã8ð(µdëŒ-+Ëx^›7SÚ͵ ×/o}äñøÏõáè+±o·‰àäöÝ>E—O1[kbXy‰.Ÿâœ˜y›ÀÀÀÀÀÀÀÀÀÀÀÀÀÀÀÀÀÀÀÀÀxYC{im“?ÞV/=Ûå#Ï.ù˜¿æ5½]\ö9º¤è-íÙFãóÊrÏzïl½iõæj59Œ#{[Óë•r¼;Í¡™ö¥ãÁe=Ä] &û•—Ø'6qL̼M`````````````````````¼¬¡½´¶ÉÒ{k/=Ûå#Ï.ù˜¿æ5ÿØ{G¹w]ŽE)e\A11­,Wð¼÷ÎÆ•›ËI.YvŽì;˜ü¼v}%;!Ή×Íð×Þ;™îÅêVVqL܉}bæmãe í¥µM–Þ[{éÙ.yvÉÇü5¯ùÇÞ»ÛÆ1*nôÞΚW–+xÞ{79yÅUg¬¶úš[“ÎõÚÇ+rØy©ç„œ¡’ï?¯¥øMì×bíÊ*Ž‰SÔ‰™· Œ6l[mòÇ›Läx´Ë+¯.yÍ_ó1ÿÐ{;g|¹yçMsã ‚©ae¹‚§½wIi±‰cbæmãe í¥µM–¢^{éÙ.yvÉÇü5¯ùÇÔ‹òr{×è­ñne¹‚ç½w5-Ü\ &jc®¹¹óÆ+>9;æ˜ð.˜”ïìSôm›Ieå%ö‰M3o/kÌ^z´ÉoG/­íòÊ«K^ó×|Ì?ôÞ}•»´Ë¼‰ÞÚÊrO{oï“cgühäoGnGï­¯ø茋ç„÷EîKßí%:ë½XòÊ*Ž‰;±ä•· Œ—5F/=Ûdé½G/½Úå•W—¼æ¯ù˜ì½£œŽîZ0M¯ ›TW–+xÞ{'ô—::û™k‘ Òk¯xçLmç„·Až6ºÛçµär/¶´²ŠcâNliåmã… é¥µM–Þ[{éÙ.¯¼ºä5ÍÇü×$÷aœ`¼qåëǃûb¼×cQÊxÍ”²²¼çµy•;Ê]òpNÙ©¹VéÆõ­WätöbÏ o›ñ_|Žg¿–šw±ÉŸ̬â˜ØÄ11ó6ñ†ôÒÚ&Km®½ôl—W^]òš¿æcþávñ`Mtãöð:niÞ´¼²\ÁÓÞ;8“ÛÍ¥j|;5—bܼ^_‘#W²;'¼ÙíS´i½ ~å%ö‰M3o/kh/­mòÇ›LH/=Ûå•W—¼æ¯ù˜콃ÉòèN'wËd¹i|f¹‚ç½w”jÜÅ0¨9sŠ&ÌÎ^_ñ¾õœ›áóý#7OÑÇ]L&¹•—èã.Ž‰™· Œ—5´—Ö6Yzoí¥g»¼òê’×ü5ó½w6EïÒöãxð0ΟY®À>í½«±¥ËÖÄòì’ùk^ó½w òøKW:3îIïWüÊrÏ{ïhb¼¹àç-3§þ–fg?_iÉø}¢ZÓ¾x¾·Üïv±µº²Š:qŠ:¡yŸÀÀÀÀÀÀÀÀÀÀÀÀÀÀÀÀÀÀÀÀÀx]ÕÕ&Kï­½ôl—<»äcþš×ücïM»´£©Ú¼WÓÜÊrÏ{ïbrèï&ÊA*²Ssr&–yíúJ‡·œ¥‰/žïÝ×û—ØÄj|Yy‰2qŠ:¡yŸÀÀÀÀÀÀÀÀÀÀÀÀÀÀÀÀÀÀÀÀÀxUC{im“¥÷Ö^z¶ËGž]ò1Íkþ±÷n&Éyá^Ù¯ ÿÄÙ•å žöÞÉÊ1åÎg¹dÙ©9Úã|ïùJ&úm"åÕû_zï¾îÚ.:9fæ%ÊÄ)ê„æ}ãEÙKk›üñ¶zéÙ.yvÉg}ÉkþÓÛÏïo?ýênîöþ›üðù–äÁ˜öfõ'R¤—:nÏñöþqûáoï¿÷å÷Ï—Îü—÷ñå‹nÒÆå)ßV–7ð¼6ÆöËðÕØ0vjŽÙØU›ë+5K³NȇóÅãÁS0¡Ü‹9­¬¢Nì¢LhÞ'000000000000000000000^ÕÐ^ZÛd©Íµ—žíò‘g—|Ì_óšÿzmžœirIÕ”ü µy2-t>È1çòª‰ueyÏkób¼½9׌ÓÂ]s(rܹ¾u}%G9èåœÐ?(Øí%f“ò½XãÊ*êÄ.Ê„æ}ãU í¥µM–Ú\{éÙ.yvÉÇü5¯ùo¨Íû_ú%•&¾^›÷÷"G«T“ƃC³3¹¬,oàimž­ r+¾5u|83‡$o}¾RÒxëÇDÿŒÒýì«6o’7ÑÉO̬¦Nœ¢NhÞ'000000000000000000000^×^ZÛä7™^z¶ËGž]ò1ÍkþápòìMÃÎãæ¯yÍ?öÞÉD9µ%ʽârÕØ´²\ÁóÞ;›RnÎz¹Ù©Ù;Öµë+¹ﶉÖáì×Þ;v±È“Ag^¢Lœ¢NhÞ'000000000000000000000^ÕÐ^ZÛdé½µ—žíò‘g—|Ì_óš콫ÉuœkâÇñàÅW–+xÞ{7Ór7ã/ríšÃ(ÚõÚõ•LrÛDjë¹›×Þ»™à7±X“íÊK”‰Cœš÷ ŒW5´—Ö6ùãmõÒ³]>òì’ùk^ó½wß_åìq¸\A4ѯ,Wð´÷îéR—û%ëNͱhc¾^iýZÜ6Q’)÷ú½¼í;±´•UÔ‰]” ÍûÆ«ÚKk›,½·öÒ³]>òì’ùk^ó½w_{²½ ãÁœ¥˜ìV–+xÞ{gä—ñ$OÙ©9:ÓÖµW¼õ&Úm¢6)÷wû¼–bïÅVWVqL܉}bæmã… é¥µM–Þ[{éÙ.yvÉÇü5¯ùÇÞ»J–ž;ó½«5Õ®,Wð¼÷nòðMç³<¬R®]sÊÇ-úŠwÕ¤¼MÔ´N/¿öÞÕÔ¶‰ýZ\YYÅ1qŠ:1ó6ñ†ôÒÚ&¼É„ôÒ³]>òì’ùk^óŸÞ~~ûéWws·÷ßä‡Ï·$‡Ø›ÕŸH‘û%Ålb¹½Ü~øãÇÛûï}ùýó·llýëÙë¾û®ý—÷~ÙÎÄ82×ÇiÖ`Z[YÞøÓº½z¹“ÝEkl;5§dBÐL_ñ>G5®BþR·÷u[w1˜W^bŸØÄ11ó6ñ²†öÒÚ&KÝ®½ôl—<»äcþš×ü×ëöjårŸœüåÏÔíºÑÙlÒCßþËûÛßÝþï[ã¶oØ~üïÿùR~»ÛéÿûýÍÉúí¿¾EïŒk·lçÑìutç庬Ϙ¢Ü®¯ëÑ„º­¸¯“ʶ>â¶Æõë÷õþ@ÜÖGÜ׋<1õ\q[ï7®kÜ×Çóçúˆûz–ÿRá\q_oò':çúˆÛz ò·à±®q_Ïò`Ös}Ä}½™˜·õ·õ<þƒ…c]㾞ä?º8×GÜ×›üÓr®¸­/~=Ö5îëIžP{®¸¯Wù2Îõ·õþPÙ>?ûzÿéöùhÜ׫qÛç£q[o΄íóѸ¯G“¶ÏGã¾^LÙ>çz²}åü|fÜ×£qn[q_/ò Þs}Äm½ÿƒÏÏgÆ}=È/ çúˆûz–óŸÎõ·uoå¿“9Ö5îëAþ,ï\q_Ïò à¹>â¾ÞLÛúˆÛzð¦nŸŸÆ}=ËQXçúˆûzÛ}šq[~ÿõiÆ}=í¿>͸¯WS¶ÏGã¶Þ!oÛç£q_8×GÜ׫üÎq®¸­ggÒöùhÜ×£)Ûç£q_/¦miÜÖ‹3vûü4îëýwºíóÓ¸¯ùsãs}Äm]~ÛÚ>?ûz0uûü4îëEàq®¸­7»ÿú4ã¾ö_ŸfÜ×óþëÓŒûzÛ}šñ\ÏÖËOß?5îëYþ„ÿ\q_oò-ɹ>â¶î¼Içç3㾞L‰Ûúˆûz5-lë#nëÞw~~3îëI˜r®¸¯Wù~ä\q[ýíû ûz4uûü4îëÕØíóÓ¸­GgüöùiÜ×ãþëÓŒûzÙ}šq[OvÿõiÆ}=î¿>͸¯¹?â\q[ï+aû|4îëAn9×GÜ×û7$Ûç£q_Ù9×GÜÖËxͱ®q_göë#îëã¿m<×GÜÖ÷o(ï/Ïõþíòãw­ŸÿÖ¿öó[áñêøá/ãG}!Ê_Ê­ÿrßÿï”ï†]óN¾!þúNùOBóÝÎð;ËøÅiß¿mg²I¾­Ýw¦oÜÙ¿¡ÍñngþÆýÿêt¿³|ÛÎlû¿Ô„»õwöoÏû·×ûÎö;û7¦~ßé­ÕýßmbÿU±Ï´q³Tÿ&L£ï¿Wøù]Úu Èw±·u×u ù˜yŸHãû¿sBó>Q³ü|NhÞ&än.ùo–×ÄÌûD¿Ëžš÷ }oçÄù^Ç¿`Îsûÿ²}°¡Û—íß+Ï[ºþ’ïÿOø·÷¹þMbÿËØÿÔÆþÝQÕÿûþâÿÔÆ:þµN6Ú?³Ï÷+íÿº7þ~ùSûú…ößÆdߟzƒ¾û2ÿí;¯¿9û¿xßBÞFóþ—ϽýÛ?ý¿Ì…ù…¿þc1^é‡0þµ[¾îßÿðÿãí?ýí¯·¿}þçý?ý—–¿äôm¿ ôï}­tòw´Pÿó¯ÿðùÛÞ}±ò]¯üß[ôSûo}ë¿üó?õ·÷/ÿûÿõþØ>Œÿ¯Bÿ®Ÿ¿K¼åþ®oZÿX18'C¬ék^ãŸÞþîíÿ~/|£ endstream endobj 752 0 obj << /Length 221 /Filter /FlateDecode >> stream xÚUÁNÃ0 †ïy [‰dqÒ´ËŽ &¸ å†8D4e•Y†ÄÛÓ-©²d[ßoÿ–ø½bRnAÈñNÂ[ _„õW±–+Â:^Áæ! Â]$ÏK¬­^ôÙ­!›ÃV‚fº=˜ ôÀ„^rÅ°`Fxióû9¹– ÅÄ]K%òfo?ìho Ö»Sic‹ªù¾$— éT]ùô1¯SS©Éæñl}ÅSŠ¡tùX>9G÷6åÕ-ÄÑùÿ&!¦lýœÚWó8hÆ•Š²c\ÔTwɽ!¿M^R¨ endstream endobj 727 0 obj << /Type /XObject /Subtype /Form /FormType 1 /PTEX.FileName (./fig-LCover45.pdf) /PTEX.PageNumber 1 /PTEX.InfoDict 754 0 R /BBox [0 0 720 720] /Resources << /ExtGState << >>/Font << /F1 755 0 R/F2 756 0 R/F3 757 0 R>> /ProcSet [ /PDF /Text ] >> /Length 40898 /Filter /FlateDecode >> stream xœìÝA%9v¦é½ÿŠ»¬ZEÒH¹¥h ¥ 1hÍB‚JBÒÕP Aÿûáá94£™Ç•{$Z€Ëç]TE~vŸkv##Âós Ú¿¼üõã_©9ï)z×Ê#ÖìšÄìO_ÿôøo~éÿÜ'Öÿÿ›ÿô]Ìÿ÷å¿ÿ?ÿø‡—ðøÏý~ òú㿼„]ÉšËåñú)»ÍÏüËã[G÷±þ±ÉŠÜÿϤ´¹¶?RnnÓ•šÛîò6VÚ‘’ú×s"—àj¾Ø§¸…UL®l3O±O,☰¼L````````````````````|1ct¹ÖÀ¾¾X—;Ù™gÿ:çïù˜ÿvïŽûúV9fÚ8ƒêŠŸYÎàyW¼»X©—òX9r–:ع#Ås9'úe»ØS,®Ä«ØÒÌ*Ž‰‹Ø',/_ÌÐ.WX銵˵FöÈIû×cþžçü·—¿üþòáñýåÿ~}dòÃ?¼þƒ,ˆý$û™…úøþúøÝ/¿|ÿsùû¯YXBråÍÂkAýÇïý¼š‹û#o^N¿_y®…™åÊŸvÔÙ»Üi¯®mcåÈ9D·i»=ìIÞõœH=_íÙQ7¡1ÈBËú³0&qLX^&00000000000000000000¾œ!]®6°¯/2!]®5²G¶þõ˜¿ç9ÿŽººÜO¡U¹÷ø§:ê±P:êúŽ:G—ûyÅ}^yv1Î,Wþ¼£Þd×T6•#K Ÿ’~fv¤×Ú2‘v9ÁÕ>E©åO1ÉÆ–§Ø'qLX^&00000000000000000000¾˜¡]®6°ÒQk—kì‘­=æïyοÙs£¯¯õ‘ctUïg®²9‡e9ƒç]ñîbx¤º¹mì¢9‡M.AÏ]ìÕ…¸L¤(·v¯ö‹œôE¬ûÌ*Ž‰‹Ø',/_ÌÐ.WX銵˵FöÈÖ¿ó÷<çß¿Ÿ¹/ýZu©üÔýÌsatÛö–Ý–>p#t“£s ²FÿÈd'ê4³|dOËíâ]ŽTú‡;¶ËÖœý~óz¤„N¶e"Ùeµg¹Ýdo”E ²Q‰eýé‹8&,/_Î.WØ×™.×Ù#[ÿzÌßóœÿ@¹]]’Sh®„÷«æ]öc>-0ËrÏ«æÍU}cÒK9ûèvkþõHéGò¶L¤4wŽ¾UÍýõþ¹.b’°<Å>±ˆcÂò2ñÅŒ8º\?X©šµËÕüË‘­=æïyοÙs£¯¯á‘cgÆžeÜ?mYÎàyW¼» »^gm™Gξ¹ìÜÇ‘â½ÜÊ}N¤"[N¯ö‹KûU¬~fÇÄEì–— Œ/fh—« ¬tÅÚåZ#{dë_ù{žóo»â&Û{ä¸I¥ÜÏ`nßg–3xÚïcËçÔOy=·æì«ìS1ÎÝŽÔ]n‘>'Ò6»ò{WܤÐ^Ä 'mY?1±ˆcÂò2ñå érµ}}‘ ér­‘=²õ¯Çü=Ïù÷·°òYêê(›H¿»…Å]ÚFµ¼ë%fçëÌrÏ«æñDCù0š^úÈÙ—fÕ¬Gj“>ýœHIÞbµOÑ·UL.…™§Ø'qLX^&00000000000000000000¾˜¡]®6°R5k—kì‘­=æïyο_5KwÅ'¹ùgžè7V·åtÔÙíò¿â|W^eÏ ËråÏ;êÝùòH5Ù¦šåÆëYÏÛ‘êeëés"yyîâjO±H»~÷8³Šcâ"ö ËËÆ3´ËÕV:jír­‘=²õ¯Çü=ÏùtÔÙù~ ­¸í';j]X³Üxý~GÝœ—Gÿ%yà`¿ò\i3Ë•?í¨û‡³É†$›ky¬9?WÎ#»w­,isåÇ÷Q7i×18¿Í¬? cbÇ„åeãËÒåjûú"ÒåZ#{dë_ù{žó訫l(R|vµ½_5×趢Õò¸ºæÑ(k– x^53L©ÊÃeåÈÙ·ÙíÐv¤mLΉÔ\ûaÕÜ_÷q“ÛÒÌSýû!Ž ËËÆ3´ËÕVªfír­‘=²õ¯Çü=Ïù7OìëKÑ­2ƶ"µÊmÐ–å žwÅŵöHû.´¬YJíh·rÛ‘½ØÖÑ–·êöëS OqÛVqw%Ï<Åm[Å1ay™ÀÀÀÀÀÀÀÀÀÀÀÀÀÀÀÀÀÀÀÀøb†v¹ÚÀJW¬]®5²G¶þõ˜¿ç9ÿþmÉ5ÉMÓ%ŒÝ;~fë ](Û~äòû™«kû#ÇÝÅÑ£· {{X–+ÚQ7ï6ÿH9Z»­9û(wtÏÌŽô¼_&ÒîÚõ^éÙQ7WÒUlefýY±OX^&00000000000000000000¾˜¡]®6°¯/³ËµFöÈÖ¿ó÷<çßÜÏÜ¢”×yó®Œ;ª[rm›YÎàyW¼ÉHÚ›mЬ9‡sï ;²ï.‡e"SXíó\úI/brqŸYÅ1±ˆcÂò2ñå érÃÜû¢i—æÞ–­=æï9<Ûû¢e—õ|qlݪ‹if9ƒç]qq->Òæ]Ô&\sMn·½/ôH9ïY9ç}Þy|¼é"ö¡Î<Å>±ˆc¢œwAc```````````````````|Mct¹ÖÀJW<ºÜÙÈÎ<û×9ÏÇüÛ®¸º&ÏÐk²WÆë#úàržYÎàYWý¸ŒTìžhËÙ'y ¬œGÚx*à9‘’ ?Ü'¹ærk›Y?1qû„åeã‹Úåjûú2»\kdlýë1ÏsþÞGßs÷ouœArµÌ,gð¼+Þ\ÞúÕìº{…åì«ÛµçžGê6zîc"móÎãkW,çR÷UL.ú™U‹8&,/_Î.WX銵˵FöÈÖ¿ó÷<çß}¦ž”Ï©=JÈnKï>/úì²< /˳åª ûÌrÏ«æâªlæÜd_ Y9röMî¬ÖK×#m<¤ïœH»‹íbŸb¨«¸»ìgžbŸXÄ1ay™ÀÀÀÀÀÀÀÀÀÀÀÀÀÀÀÀÀÀÀÀøb†v¹ÚÀJÕ¬]®5²G¶þõ˜¿ç9ÿíÍmÉÕÕí‘c±-,Bp©Î,gð´+^ªé$Ï»ÛÇÊ‘sÈ.Y‘nGJr{Z&¶*§´Úó\š´Ü±†™õÓ±OX^&00000000000000000000¾˜¡]®6°¯/³ËµFöÈÖ¿ó÷<ç?p[ru¡Ÿä¾ËÖÄ·Û’?°0·âJù- ÷$ üm ó›w|{uˆ.¤þIT—Æv!!ÉÛ–å³~ÞŠo.•þ‘Dyn¡¬9{ï²ý,é‘"ÕX&Ò.?¬ö£«~“ qfÇÄ"Ž ËËÆ—3¤ËÕVZqír­‘=²õ¯Çü=Ïùo÷;¨¥.—‡þE©çå ªómf9ƒç]qqûþHy“›ÀeåÈÙgWí!€v¤yÙ‘äœèç®;yœb¿ìEì±Í<ÅVqLX^&00000000000000000000¾˜¡]®6°Òk—kì‘­=æïyοµ”ÏÛØØ9‡Ü–\¥îÍqwelÖƒK~f¹€§UsôÎׇܩƇ£9K·m5¹©ýÃX'R|V57—âUÜÓÌúaŽ‰‹Ø',/_ÌÐ.WØ×—ÙåZ#{dë_ù{žó¨š«óòl¿ìüö~Õ£óeì½½«cr{˜Y.À?­š7·µñahË®Yvè­e·#u—ÒûœHÞÅî€ÑÏeßV19ŸgVqL,☰¼L````````````````````|9Cº\m`¥jÖ.×Ù#[ÿzÌßóœ¿jŽÁmylÂÑòªæì¶}TËql¶ûûÄ™åžWÍã)…©…qaË9l®mvéz¤&Ù%äœH›ì½Ú§è·UÜÝVfž¢ßVqLX^&00000000000000000000¾˜¡]®6°R5k—kì‘­=æïyΠjNòÁ”ƒ+íUsu¥>r›D÷ Ø‚Û¶™åžVÍ›w>»â÷¶°òYOa¯÷-,>°0•àR|ï ¹ñYÁW]wT§ÝÕ}f¹òçuq9=6tiÍÙ79 ýÌÆ‘¼m bysõº…ó³«mwÃÌ*Ž‰E–— Œ/gH—« ¬tÔûèr­‘=²õ¯Çü=Ïùooîg®.§Ñ çqŽ9ÈÌ–å žwÅMÞ"íYöy–sYvÙóÜõÈޜߖ‰þ–ò¸ÀÇÛ®¸–û³— 8,OQ ôCÔ ËËÆ3´ËÕöõev¹ÖÈÙú×cþžçü·{WÜ××þú]»oääR›YÎàiWœ;);{쮌;’5gŸå–ìqîz¤øèrX&r”[¬W{Šcð"ÖmfÇÄEì–— Œ/fh—« ¬tÅÚåZ#{ä¬ýë1Ïsþý½/¤|<«ïÝ-,òX$Û*oZ‡ï®ú™åžWÍÅ¥úHµ›cÛÍ9ç­H·#-É»žiŸ$¼WÍÙÕ¸Š» ifÇÄ"Ž ËËÆ—3¤ËÕVªfír­‘=²õ¯Çü=ÏùoonK®.•GÞ6Ù‰¸ŸA .„™å žwÅMÞ"UïŠ^ÝÈRjÇY“ë‘–].ËDNn¿nqŠ!.bñ.噧â"ê„åeã‹Úåjûú2»\kdlýë1ÏsþÛ½+îëwy^“ÊYÎ Éõ,Ë<íŠËØ#õK°«yl¬lçnGöâJY&dænw\¢KÛUÜËÌ*Ž‰‹¸—™— Œ/fh—« ¬tÅÚåZ#{dë_ù{žóïoaÑx9IïRz ‹’¯¼Er»tÔ–åžWÍÅ%­š«ÞÐXWn9»V/ö£ÛÊU,ufÇÄEì–— Œ/fh—« ¬tÅÚåZ#{dë_ù{žóßÞtÅÙy}^ÜÇì²{†e9ƒç]qqÛöH%º =÷ÈÙ—ñP¾qîãH yÚ:!Oókû<—²_ÅÖfVqL\Ä>ay™ÀÀÀÀÀÀÀÀÀÀÀÀÀÀÀÀÀÀÀÀør†t¹ÚÀJW¬]®5²G¶þõ˜¿ç9ÿíMW\Ýy+®Žs¬Þµ2³œÁ󮸹";lD¹ZÎ}äì“KÁÎ]´"§pN?﫾wÅU îSìç²ù™U§¨–— Œ/gH—« ìë‹LH—kì‘­=æïyλwÅ5¸²=r .è$©©-Ë<íŠktMvØØ•#çàe[qîv¤_MIËD®òˆ»Õ>ÅØV±ÿC˜yŠ}bÇ„åeã‹Úåj+]±v¹ÖÈÙú×cþžçü·7]q’‡Øå­ºPÆì.×™å žwÅÅÅ2šï2öÏÐ,ïgW<Ž”~d/ËD_‘ÜgWüUlqfÇÄEì–— Œ/fh—F+]±v¹š9rÖþõ˜¿ç9ÿíMW\åä¼õ±q_qó²y…e9ÿ´+n.ïýjªËzGòÈ9$—ì®i;Òš Û2‘“œòjŸçÒßôû¹ÄmfÇÄ)ê„åeãËÒåjûú"ÒåZ#{dë_ù{žóß^þòûË_üUx„Ç÷”ÿûõ‘]Èÿðú£|î'¹5çëãûëãw¿üþñýÏýåï¿Þzæ?~ïo\μEçÛ¸€ä¢ŸY.àiÕܢ순ö '*+GÎ!»d[5Û‘ÚÆ¥ÙÏ2úV5÷×cXÅÍå4óûÄ"Ž ËËÆ3Âèrµ•ªY»\kdlýë1ÏsþýªYºëô(a“ýŽoUóJ‘¾ÅtÔÉÕòȱ¹6 Øvéß-Ë•?廓 ­Šóa~š¥L϶í‡Ù«Ûã2‘žÝݲËñ*Ö<³Šcâ"ö ËËÆ3´ËÕV:jír­‘=rš·?Ûü=ÏùtÔI~H¹Hû~Õ\]Øy cgÇ潫qf¹€çUss©=R-²†\úÈ9z·Ù½âv¤>.i»×ØS¬®n§(çÊÌ*Ž‰C´ ËËÆ—3¤ËÕöõE&¤ËµFöÈÖ¿ó÷<ç?P5ï.å~’Áµös·Cë½ÿðævè-ŒÎo¿uáþn+¾ù wˆçX层òY'ÙÞIJ|ÖÏZñÍGW}ÿH¢ÛÃX9röÕmº?Ê<"dôËD r·ûjŸbH«¸ÉÝê–§Ò*Ž ËËÆ3´ËÕVZqír­‘=²õ¯Çü=Ïùo·ºXÖïrßó¸ó[Î`w)Í,gð¼+.²HÚ«ì "+GÎ!Î&x©eöКStíjO1k˽ˆû>³Šcâ"îûÌËÆ3´ËÕVºbír­‘=²õ¯Çü=Ïù·]quÁûˆ½l(²…ÑßZ–3xÞ7¹U:ÕÍmzî#çPæF#óHIºYÉÌ©“ÛÅ>Ïe/‹ØÏ¥Ÿ„eÇÄ)ê„åeãËÒåjûú"ÒåZ#{dë_ù{žóïî¶!åóÖO¡Vjï [˜ö,;…¼»÷E.éÃÃ6®<9Ÿg–+ÚQ‡(—6¶ÑÏld¹Â¢w+Ï#¥È¥Ÿ)þøáƒòºßW±ÿµ™§Ø'qLX^&00000000000000000000¾˜¡]®6°ÒQk—kì‘­=æïyοßQïJ?É=Hãü~Õœ¤û–j9ë%îÒ¿[– x^5çÓ#µ¤Ï]´œCue·K×#%Ë®ÙçÄÖœ/{Šy”䫸û™U±OX^&00000000000000000000¾˜1†­•ªY»\kdlýë1ÏsþUs’N8ÕìöTÍã1‰R-oc—æèå},Ë<¯šÇHª»«cj͹"aVÍzdß\¨ËDò²YójO±Ž’üû¹ø0³ŠcâuÂò2ñå érµ}}‘ ér­‘=²õ¯Çü=ÏùTÍ»ÛävÓûUs ò¤À£Kã¶ì¸ÉÛ–åžVÍ}¤”Gj›ËûX9²l'­j¶#{39'ÒÈ«}ŠR’Ÿâæ¶8óå$NqLX^&00000000000000000000¾˜¡]®6°R5k—kì‘­=æïyÎ{ùëÇ¿¼øþ®þ±þÿßü§[qÿ¹ÿïÏ/²í´ü——¤fNÑ»66¬ŽÚk–3½ X inR’Ëey]ãòúîêúºÆóõþ™„ý|Ýâòzr©.¯k\^ßåɈçëÏ×kMKŽ×-.¯'Ù‰ä|]ãòú.·Ÿ¯k<_oÁùí|Ýâòzr[Z^׸¼¾Ë(çëÏ׃÷ò Èc`æu"És&— ÍëÄ.ÿš,š—‰þïC]>æ™×‰äB['4¯]õë„æe¢¿T—{æu¢«qмNì.më„æebëÿÊ/ùÌëD»¼œš×‰}ÜœNh^&’_~¥ô ËëDÿ5¼~ê–׉ÝÅõS·¼Ldïòú©[^'ºº~ê–׉]žºLh^&JW×OÝò:±É_ÈX&4¯»ìã¾Lh^&v/9=',¯›üfµLh^'vùkË„æe¢z·­ŸºåubseýÔ-¯ŵõS·¼L´®®Ÿºåu¢«ë§nyèêú©[>'bÿµ—‰™×‰þ‡Î¶Nh^'Š<§v™Ð¼Lô_ëaùÔg^'6—Ê:¡y(㡸ç„æeB¾Ýº|ê3¯Û؈êœÐ¼Nèß:'4/›~Û÷˜°¼Nlò=×eBó:Q\Y?uËëDsmýÔ-/ý×z\?uËëDqyýÔ-¯íòGçÌËDÿµ¾þá9ó:Q.|μN´Ë 3/ý×úúGèÌëD‘oA/š×‰®®Ÿºåe¢ÿZ÷ë§ny(ò5Ï2¡yh®¬Ÿºåe¢ÆËŸ¦3¯åò§éÌëD»üi:ó2ÑâåOә׉rùÓtæu¢¹t™Ð|Nl¾«Ë§>ó:Qä΄eBó:¡>'4/A¾qLX^'ÊåOә׉vùÓtæeâòuçùuè9ñ—ßðåí¯ÿôòíþŠñíø¿?Œÿ×9Ë]›|zz‹EôrÅ»ëöìrº¬+Zײü¦µ¬Ûâ‡Ö…å+ëuaýØÂ-Ë/ÇeaJ[˜“|ØËÂü±&ô/Âû—mëÂ}6¡õ[.ïX>öáÄþïAÿ‚d]ø±'ö¯ìúŸþËÂýcNìNÜÖ…õcNìNÿsm]ø±'ö/ÚåÛÇ>Ùžq¿ü+×ôÃùã÷—þG†ü§jOèy}h,%Ͻ ï¯oú3#.¯÷?¶åuËëúµÍñºÆóõÜ¿FoþxÝâòz‰ò»ßñºÆåõ6þÔ:^×x¾^ä+üýxÝâòº^ðñúyýã-ûx—ôËG½%ýh\³|º¸ý|þÛëúÉÅÖ¥ŸY×ÿC³]b]î?wAßÏÿ̲þSšôí~êÝúÏtÿlü¦óËŠü Ðô×ñÏ,ëÿ^Dý+ÙÇ¿àýø)?øá×?=þÛãŸÿ?;6û•￲¢þÖ:¶1-ý?ꃞîßþîïþö÷¿ùÓ¿þø_÷Ë¿>þþ?þîŸþô±ßüúfôë6ýƒE°ÿë¾òÉÕË¿úg-OÓŸ¢ÿú§¿ûõ_ÿç?÷ËÓóX>ŒßÒÍßôÏÞ—­_õù_”¿Ì\ö c¿Ì_S×xL&ê‘üz4Ö,BìÿñÚ¿0°Ÿ”Žýðz®÷F¾¹žþkµÉî²ów­þÅD*3Ë/ûg7J¦8~¦Sÿ "ŽßO4gæóÛæ‘þ[P¬Ë„ÜZ/öýø²~kœYÅ1qû„åeã³ÖKk›üú2{ik—l]ò1ÏsþÝýdAÿA¾Ýb߃ú·öHqÜבcqI/1˽ˆ–åž×æiìó[æ·é4g_äfýpôˆ|óÃ/)ÏïÔÝkómܳwŠyÜÇSÎïüéÄ"fÝÎaäeãÒKk›,µ¹öÒÖ.ÙºäcþžçüûµyŒò—ôËå¯Á¼_›Û#÷é0þŽI¬òl-Ë<¯Íõ¯á—&+EVŽ,7úgûë/v¤eÙUáœÈ^þRÛjO±Œ¿ˆzŠÕE?³ŠcbÇ„åeãÒKk›,µ¹öÒÖ.Ùºäcþžçü·[õb“?÷ŠîÝåä,Ë<í½·ñJ’½~uW¦‘sèsv/½Ù£ì•sN¤,;Ü­ö)Ê6±§ÆÖ‘š§â*Ž ËËÆg5¬—Ö6ùõeöÒÖ.Ùºäcþžçün³?61¿Ý.þ…)gê½{Ÿy?ﲇ'½ò,ƒ²,Wþ¼oß\KýSl²¼¬9‡â²=¦ÁŽ4/O :'ä™DåbŸ¢<4é“Ëuæ)ö‰E–— ŒOkh/­m²ôíÚK[»|dë’ù{žóï÷í[Ï#õM¶^ÿ™¾]¦\Ú~˼ͧÝþì;fÙ þ·,ôòÝŸ_( ¼Yøƒï)ô+Ü’‹>Ÿ¹Êc€-ËÏîóï)ì.l”êüNÍÈYŸU«ÿ^Œ#¥Ñ}l"·ù½œû÷Š ^Ä}ŸYÅ1qû„åeãÓÚK‡Ñ&Ë÷´—ÖüË‘­K>æïyοٻfkÎçÑs·ñ¨Ó¤ð¶,gð´÷NÞmñ‘ú;¥±ëæì›|›dœ»iÞ¥¼LÈvôñbŸçRÚUlefý4ÆÄEì–— ŒOlH/­mòëË쥭]>²uÉÇü=ÏùÜK_]¬âãOßÙ>¦]z³éÍÛûÌûXL¬ÞÈ•'··™åÊŸ÷í›üµ€´'WÆd5gŸœ·ý~ôÈøîA\&r‘«=Åè꾊i\fÇÄ"Ž ËËÆ'6¤—Ö6Yúví¥­]>²uÉÇü=Ïù÷ûö\꧔›kØò=eÙ©fÜ–>n¥O»kuf¹€çµy‘B?•è¼®YæºÙMøz¤„ùY›oW\·“?ÅN-âî¶4óû?.☰¼L`````````````````````|ZC{im“¥6×^ÚÚå#[—|Ìßóœ³=Kª®ÄGÞÆ#YûäàÂ>³œÁóÞ»¹Ú²oº+5‡¹Îo è‘Ú¤¹?'²—cVûû¹bÌÛÌS짨–— ŒOkh/­mòëË쥭]>rÖ.ù˜¿ç9ÿÛÅwWó£Ä涟{b©.L±þàQ§oï3ïç]ûʘe3¹ò$że¹ò§}{—’ö*••#ç0±:>3;R79¡s"y¹Ý|µ§å›q3«8&.bŸ°¼L`````````````````````|ZC{im“¥o×^ÚÚå#[—|Ìßóœ¿oï |z”­ÉIþLß® s«®ý\Q¯ e—øýû¨÷ nml·ñ‘í.ç™å#{^Ôe·ïêØ-Gsöåøö‡)ÞË÷(ΉþCýásKsve»Š-̬☸ˆ}Âò2ñi í¥µM–¢^{ik—œµK>æïyοÙ&Wtãó2¾™P¼ÛÓÌrÏ{ï&÷° PõÜG–o Ì]âíHÛ\,ËD?%Ÿ/öy.:Å~.ÁϬâøÇSÔ ËËÆ'6Bmòë‹LH/míò‘­K>æïyοé½KpÛþÈ1º²3Ø\‹3Ë<í½K‘§¡n¶…ºæ¼ì$3ÎÝŽ´ì¶°L¤âêõ^òSôaûÇÐfžbŸXÄ1ay™ÀÀÀÀÀÀÀÀÀÀÀÀÀÀÀÀÀÀÀÀÀø´†öÒÚ&Kï­½´µËG¶.ù˜¿ç9ÿ¶÷Nò¬Ñ¼ÅÑ«÷3Ø]3Ë<ï½³«i½º‘s(n³VߎÔ(Ï(='Rž½÷ÞÙE¿ŠÅ¥zý¼ÆÄ"Ž ËËƧ5´—Ö6Yzoí¥­]>²uÉÇü=Ïù·½÷.ÝsŽÅ¶Dß½ÛüÌrÏ{ïæü¦w¯•šåÎò¹‹‹iqœû1!O9õ{ŠÕmí*–}fÇÄEì–— ŒOkh/­mòëË쥭]>²uÉÇü=Ïù7½w_ßÒ#Çæ6í½7—ÚÌrO{ï½Kqåyì¢9‡ñÃ8w;R³lA~N¤6ïf¿õÞ{p¹^ÅšgVqL\Ä>ay™ÀÀÀÀÀÀÀÀÀÀÀÀÀÀÀÀÀÀÀÀÀø´†öÒÚ&Kï­½´µËG¶.ù˜¿ç9ÿ¶÷N.l¼yaä ŠlpbYÎàyï]Ò]\ÂØ£Esɵyîz¤ÿ`ׯ9í²¹ËjŸç²ï«XœO3«8&qLX^&000000000000000000000>±!½´¶ÉÒ{k/míò‘­K>æïyοí½w¹µ±ˆcÂò2ñi í¥µM~}™½´µËG¶.ù˜¿ç9ÿíå/¿¿üÅ_…Gx|ÿGù¿_YNÁ?¼þƒé{z¤]ˆï¯ßýòûÇ÷?÷—¿ÿzëÌÿø}¼ýÞWn›k£Ú¯EF,Ë<¯Í“ŒŒó7škÎÁ»m^ºiý3Y'ò&íùjŸâ–V1»fžbŸXÄ1ay™ÀÀÀÀÀÀÀÀÀÀÀÀÀÀÀÀÀÀÀÀÀø´†öÒÚ&Km®½´µËG¶.ù˜¿ç9ÿövñâª<³È¼œAs)Í,gð¼÷®.ÈÍð›m°¢YÀi‹Ï#{v­-ÉËè«=ÅÝåí*V?³Šcâ"ö ËËƧ5´—Ö6Yzoí¥­]>²uÉÇü=Ïù÷oï B|¤²¹¶ÝoÿÀÂ,—ËÂdóðwoPoÞù:ö/›|d-º²Í,ÙÓ¢¾¿ÁÖß.÷CãÖvÍÙWçu¥)²—º_&úO‡}£âVÔ÷sÙãUì?o–U±OX^&000000000000000000000>±!½´¶É¯/³—¶vùÈÖ%ó÷<çßÜ Þ6õ¹iì=Ó²«qf9ÿ´÷N²UŒlE³ic>²<1´Ø^ìv¤&¹GþœÈ›KábŸçÒü*fêÌ*Ž‰E–— ŒOlH/­m²ôÞÚK[»|dë’ù{žóo{ï"{Àä­ã šóaf9ƒç½÷îv¹ý¾Øí÷šsØ]³ ÉíHI.çe"¥»}Š¾­buÛ>óûÄ"Ž ËËƧ5´—Ö6Yzoí¥­]>²uÉÇü=Ïù·½ws%?rÜ]–'‰fß_÷3Ë<ë½³÷®Å‡<Šttö–s(R´ËÊyd¯rËù9‘Ú¼ßýÚ{Ëë±®b›Ö-O1ÖU–— ŒOjÌ^ZÛä×—ÙK[»|dë’ù{žó÷Þ[ÖWyðf’Â]Î »­Í,gð¼÷N.ÈcEw9eY9rÇ..óHõÓÖœ’ËþbOqsi¿Š{žYÅ1qû„åeãÓÚKk›,½·öÒÖ.Ùºäcþžçü»â”}²x/?üÄó4manUªü÷ž§™}q^ö÷²¡¹\y•Ì-Ë•?ïÛw·ùGÚÇ?eåÈr…¶rÙ“ki™È^žzºÚS,®”«ØZVqL\Äþ–— ŒOlH/­m²ôíÚK[»|dë’ù{žóoï3o²‹­¡½´¶ÉÒ·k/míò‘­K>æïyο߷÷1<Šn«¨Í³|g@¶]/c[˜¸»=Ì,ð¼6/.Éf;A®XVŽ,ß(vk¿ÙõÑ©ÇD .\·xŸb–½âqw~ŸYÅ1±ˆcÂò2ñ‰Pf›,µ¹öÒÖ.Ùºäcþžçüjó$û¶Ÿ\ ?u›º.̵ºæ?зW·ÉcN£óã[ ›—e–åÊŸ÷íÍÙl§ ŸÙÈr‡¿Ýà>ìEv¶9'RvùjO±ÊÇ}Šý\b™YÅ1qŠ:ay™ÀÀÀÀÀÀÀÀÀÀÀÀÀÀÀÀÀÀÀÀÀøÄÆ–g›üú"ÒK[»|dë’ù{žóèÛwÙ·]ö®ñ¨Í·à²ì _]ûÚlI6…±,ð´6ߢ«ò Ø:ö»yXÓëMèóÈÞf%¯9õÜ.ö)†ºŠ›Kyæ)ö‰E–— ŒOkh/­m²ÔæÚK[»|dë’ù{žóßnÕ·¬ßõ©¡›öÞ»‹uf9ƒç½wq^ö®IòøTY9²Ü¿';w=Ò‚ a™Èã¢W{ŠY¾[pKšYÅ1qû„åeãÓÚKk›,½·öÒÖ.ÙºäcþžçüÛÞ{wMžÞ]Í{ò.í3Ë<ï½› í‘ú¹ê3R5示×s×#{qaèoÚÅžâ¸Éü"ÖmfÇÄE¬ÛÌËƧ5´—Ö6ùõeöÒÖ.ÙºäcþžçüûÛ¤ôAo9/í§¶%×…%xiÜ~aÚÇ“Mßݘ%ï Èòº×{Ú\)3ËGö´¨O㥩ìöèTÍò Óh{ÒعÝ~H»ü°ÚS ’±¿eœYÅ1±ˆcÂò2ñ‰ 饵M–¢^{ik—l]ò1ÏsþÍ ê)¹Í]ÌuÏõT\Í3Ë<ï½³Ëò4ÔÝå8VŽœC{ö©)yÜ\LlÙ±Ïsii‹‹afÇÄ"Ž ËËÆ'6¤—Ö6Yzoí¥­]>²uÉÇü=Ïù÷oPO›ležk•3ù™ûÌÏ…ûþûÌw—äŸÁ•ñÓìO3Ë•?ïÛûÈÓP7Y ŸÙÈ9æ÷ ôÈî]+Ë„|¸×MÖO1ÄUl.ù™§â*Ž ËËƧ5´—Ö6ùõeöÒÖ.Ùºäcþžçüúöâöø(1ÊníïÖæýíK5ù>v´É›‹ÛÌrOkó\“›ðƒ˜²räìG7>.ÝŽìÍmq™HA¶ªYíSÜÂ*F—ÛÌSì‹8&,/ŸÖÐ^ZÛd©Íµ—¶vùÈÖ%ó÷<ç?P›7×£ÄìÒÏm‡® sÛ¥°¿oß\•ç”6WÇ7 rq)Î,Wþ¼oÏRè§ZÜ>ž¹ªYvŠ·Üç‘Ú?š´L¤8ŸÊzïÛ“Ëþ*îufÇÄEì–— ŒOkh/­m²ôíÚK[»|dë’ù{žóï÷í}Aè']štÒï×æ»óå‘ãØ|F. ÉÃG-Ë<¯Í«Û¶ñ´R¯+GÎÁ{ÅÛ‘}w9,ÛèõW{Š»|á"¶}fÇÄEì–— ŒOlH/­m²ÔæÚK[»|dë’ù{žó¨Í‹ì»ž›—­T~æ6u]XBvÛ¶…)^¾3c•óìW^¢«~f¹ò§}{ ²}*› ÈÊ‘³ï„ÝüoGÚ&àœHIþÀjOÑ»ZW1ºPfVqL,☰¼L`````````````````````|bCzim“__dBzik—l]ò1Ïsþ}{“ý`r+®|`w—~-IvO÷.‡–ìZ›Y.àymžäfýTv·éÊ‘eÿš`;âØ‘–]ŒËDjr·újŸ¢ßW1»-Ï<Å>±ˆcÂò2ñi í¥µM–Ú\{ik—l]ò1ÏsþÛ­úÎ¥¸yÓ‡œö3h.ì3Ë<ï½w×ü#ÉãNÇÎ4š³Ï.ÙSDõHñEvš9'乨õbŸb,«X]N3O±O,☰¼L`````````````````````|ZC{im“¥÷Ö^ÚÚå#[—|ÌßóœÛ{7WƒöÜc ™=º­Ì,gð´÷ÞƒóòLÔ*w¸ËÊ‘åÆôh­¾Ùç®î–û¹mébOÑ»”¯â¾Í¬â˜¸ˆû6ó2ñY 륵M~}™½´µËG¶.ù˜¿ç9ÿþ6)}ÏcwñÞß&e×6²uÉÇü=Ïù7·‹ïÅÙÊN.çD?·˜/öy.u[Å*wÍ[VqL,☰¼L`````````````````````|bCzim“¥÷Ö^ÚÚå#[—|ÌßóœÛ{7·ÕGŽÅµÑÍ× ƒ–å žöÞÕËs7G‹?Î]ó؈ÜZ};Ò¼=UÔrŠ®´‹}Š>®bp±Í«a½´¶É¯/2Ѷ£]>²uÉÇü=Ïù÷·IÙ«ì'^üæjz›”]Þ9z×ô²<~Ó²\À¸’Öæý ò#íIÞGVjní¬ÍǑܪ|oà˜JÿÇOÕì¯Ç°ŠI-O±O,☰¼L`````````````````````|ZcôÒÖ&Km>zéÙ.Ï<»ä9ÏÇüûµy ²y I6p¹ÕæZX¥Ç¿oÏn/³Ý(_«ÛÂÌråÏûöÝù4>Åay™ÀÀÀÀÀÀÀÀÀÀÀÀÀÀÀÀÀÀÀÀÀø´†öÒÚ&Kß®½´µËG¶.ù˜¿ç9ÿf{–¾¾Éã7ç­ô-ÈBËrO{ïæ]Œýj¢ÛÆ÷ 4çd¿˜qîv¤ìòM…s"%Y±Úó\šËí*Ö2³~câ"ö ËËƧ5´—Ö6ùõeöÒÖ.Ùºäcþžçü›Þ»Eäñ›U*i9ƒ$ûºX–3xÞ{o.É]ôÕî¢×œƒ—ê~œ»)ý-Ë2±—ÒÅ>Ïe¯«˜œÏ3«8&qLX^&000000000000000000000>±!½´¶ÉÒ{k/míò‘­K>æïyο¿¿J n“[Î7¹ýgöW™ ³lþ[öSn¿aaÊE>´ß°°Œ›âßÝ ¦eùBŽI6x—Ÿ¤ÝÕ:³ü$=ÿÖ@qÅ÷÷‰ãfý‡åìw¹G_zõÈÞz¶ebÛ]ùáÖëý\Ú¾Š»‹ifÇÄ"Ž ËËÆ'6¤—Ö6Y¾5 ½´µËG¶.ù˜¿ç9ÿo $—ËCÚóŸíÛÇÂⳫ?ÙðÛ;nÎï(ê«l¤“ct»|dÅYfY>²çE}s»<0¶ºM¿£¹BaõãHÞwù‹ Ç„<þõfŸb(§X¼wi›yŠ}âmÂò2ñiÑK[›üúb½ôl—gž]òœ¿çcþ¾w¬ßåÙ¤^¿™P|’¿>`YÎàYï]|t­ö«Ùä!­²Rs‹ò— d¥)ÁË_#8&rö¬ö)ni7WâÌSì‹8&,/ŸÖ½´µÉÒ{^z¶Ë3Ï.yÎßó1ÿ¶÷N®ÉðEîs—3Ø]Ê3Ë<ｋ û#%ïê>Vj®mÞ\oG¤zþœÈéØ»çÞ{g—·«XÃÌ*Ž‰‹Ø',/ŸÖ½´µÉÒ{^z¶Ë3Ï.yÎßó1ÿ¶÷®Î7Ý€|4ïÁ»’f–3xÞ{7¹m=•êrç>²lã‹»)MÞòœØ²äÕ>Ïe‹ØÏÅû™U§¨–— ŒOlH/­mòë‹LH/míò‘­K>æïyο»!¹é[?¥½ºí§ö·…){¹Óü·,lrçú{’—\” È« cCò°¹ºÍ,ÙÓ¢>D—óC¶™im¬ÔܲKöÓ¡GrÝäÚ‰œ²äÕžbp-¬âæB›YÅ1±ˆcÂò2ñ‰¸Ï6YŠúÑKÏvyæÙ%Ïù{>æßÜ ’Keܾë ê»óqf9ƒç½w–‘”ûäØLÝrk.Ù¦2zD^ªÛÎXNÉù=ˆS^~‹ÛêÌSì‹8&,/ŸÖ½´µÉÒ{^z¶Ë3Ï.yÎßó1ÿ¶÷Þ]É£m̽‹af9ƒç½wuM+Úì1£–[q±Ú¹#r¯{ŒçDÞú¹´‹}Š±­bsyŸyŠ}bÇ„åeãÓ£—¶6ùõÅzéÙ.Ï<»ä9ÏÇüû³„âj[£·üþ6)ýíkÒçvŽKŒãNuËrOkóåq¤©Ù‚]VjnéØÏ\äÝ»ìÏ ©ôÛoÁ¥z÷2³Šcâ"ö ËËƧ5F/mm²Ô棗žíò̳Kžó÷|Ì¿¹]<ŽûÊs,ò@Q9ƒþuf9ƒç½wv[èWÓÜ>žRª9Ë–ìö”R;²'9rN$ï¶ýbŸçRö«ØòÌ*Ž‰‹Ø',/ŸØ^ZÛdé½µ—¶vùÈÖ%ó÷<çßö޻ܙžã6¶aég0NβœÁóÞ»J9Ÿú¹FmõGξ¿¡=ƒÔŽ”âê¾L¤ { ¯öy.µ¬b“Ýb,«8&qLX^&000000000000000000000>±!½´¶ÉÒ{k/míò‘­K>æïyοé½7/ÏË̱ÿQ¶èZ™YÎàiï½WÚ#•,ûËÊ‘³/òœÎqîv¤V·ùeBÎíº‡Ê)ú¼ŠÑmqæ)ö‰E–— ŒOkh/­mòë‹LH/míò‘­K>æïyοí½7Wü#÷ÂhÞ·"›„[–3xÞ{'Wë#¥Ý5]9²¼S›½·)»ÔÝçDösóó{ï\L«˜]3O1¦U–— ŒOkh/­m²ôÞÚK[»|dë’ù{žóo{ïâö¦ûšŒý½·æ¶4³œÁóÞ»ÊÖ)i¶7¹æì«\‚ž»)E6-?'žîs²íò¦q÷3«8&.âîg^&000000000000000000000>­QF/­m²ôÞÚK[»|äc_›¿ç9ÿ¶÷n®ÕÑsDZ¿wŠò@KËrO{ïþ±ay™ÀÀÀÀÀÀÀÀÀÀÀÀÀÀÀÀÀÀÀÀÀø´†öÒÚ&Kï­½´µËG¶.ù˜¿ç9ÿ¶÷n²ñ·ôÜúdÍdËrO{ïâÝÖôjÆÝìšÇåöLN;²{ÛÅÅrÚ]Ëû<—=¯bp~›Y?1±ˆcÂò2ñ‰g›üú"ÒK[»|dë’ù{žóozïÝyó.nã ’«ef9ƒç½÷ær}¤½ºÖÆÊ‘ÇN⶿·iE¶79'Ru)\ìó\ZZÅ$›Œ[VqL,☰¼L`````````````````````|bc ³M–Þ[{ik—l]ò1Ïsþmï]j£çnc§•RÏ3Ë<ｋÛ÷GÊÍ5mõGξ{´è‘âãØ£å˜Ð­WVûöŠ»\„å)ö‰E–— ŒOkh/­m²ôÞÚK[»|dë’ù{žóo{ïêŠìç½Ës0ûìÁÅ4³œÁóÞ»Éâ)o.Ôqî#Ëåû¼ß[´ìj[&rz²¿w}‹‹¸{y¸¦å)nquÂò2ñi í¥µM~}™½´µËG¶.ù˜¿ç9ÿ¦÷îë«ìç]ÍûÞÆ™å žöÞ{¿ŒüH{sAÏ}d¹³|³ÎÞŽô ½›ÝrŠOö÷ÞGÝ÷6³Šcâ"îmæeãÓÚKk›,½·öÒÖ.Ùºäcþžçü·—¿üþòáñýåÿ~}d)£ýÃë?È‚ 7¥'l~}üî—ß?¾ÿ¹¿üý×[gþÇïÏΗGŽmÞ.¾»f– x^›·¥‡Üü^Ç+šåœ›m“bGZ±mb,ç.Õ‹=Åìv[YÅ1q[y™ÀÀÀÀÀÀÀÀÀÀÀÀÀÀÀÀÀÀÀÀÀøĆôÒÚ&Km®½´µËG¶.ù˜¿ç9ÿövñ±ŠôÜaœcõ®ú™å ž÷ÞÍåí‘Jv±ŒsYÞ)&;w=Ò’<™óœHõnŸçRÛ"ös ûÌ*Ž‰SÔ ËËÆ'6¤—Ö6ùõE&¤—¶vùÈÖ%ó÷<ç?p»ø.û‰—X\l÷ÛÅ?²°¿Sð¿aaê§ìßÜ þ‘…ÙË–0ïÞÙ^ƒŒå°ËN4òYoòÍËòY?møkt{ì'¸Ë ÊÊ‘³ßܦ+íHéVØ–‰tüÅ€[Ãß_÷u;UfžbŸXÄ1ay™ÀÀÀÀÀÀÀÀÀÀÀÀÀÀÀÀÀÀÀÀÀø´†öÒÚ&Kï½´µËG¶.ù˜¿ç9ÿ~Ã_½ìœ^B“;æoµù¿¿üõã_^¼ó¾/XþÿoþÓ­ÃÿÜÿ÷ç— ¯?þËKŠAN(E/Ï•+ÕÖ\³\é}bæäfHn«ë„æubwy_'4/Û8ÏsÂò:Ñ?Ñ´Nh^'ÆOÏ2¡y™H~lLX^'¶±YÐ9¡y(ãïEœš×‰6¾ƒtNh^&ô/8œ–׉ââú‰Y^'šKeмL”èÊú™Z^'²l/´Lh^'šóëgjy™Ø£‹ëgjyÈã»ç„æu¢ŽGœš—‰þëe_?SËëDÿÇõ³¼NTÖOÌò2Ñ‚üE“sÂò:‘\^?1ËëÄîöõ³|NdïåIÃÇÄÌëD»?š×‰]~ãX&4/ý—xZ>±™×‰M~KY&4¯E¶­Z&4/ý%¿|¦3¯›|So™Ð¼Nù½p™Ð¼Nôßõ¶uBó2±EW×ÏÔò:Qd7¯eBó:Ñ.¿ÓͼL¤xùnæu"_~§›y¨n_?1ËËD²kØ9ayè&¬Ÿ˜åu¢ºmýÄ,/%¸¼~b–׉äöõ³¼Nì®­ÿZ^&öàüú™Z^'’‹ëgjyØ]Z?SËË„üá¸~¦–׉MþÚÚ2¡yü/š—‰æ/¿ÓͼNl—ßéf^'Êåwº™×‰vùnæsBžCÞ–Olæu¢Œ{Î ÍëD_󜚗‰]^>±™×‰,O3_&4¯ÕµmмLÄ(7Mœ–׉,{æ-š×‰ê²_'4/[peý*Çò:‘\]?SËëDu~ýL-/)È-"ç„åu"]~§›yØ/¿ÓͼLdùnæu"]~§›yØ]\?1ËËDé_­Ÿ˜åu¢e´~b–׉þ•Ñú‰Y^'ôaõç„æeb×Çú–× ÝÁðœÐ¼Nèßõ<'4/—¯zϯ‚ωþÕùÛ¯®ý§þ%{´/ÙÇÔãÿþ0þ_$ùalý±ÝíZ ò•ûûKû¿>ýl.K·.ÝÇïn—¥écK³Ïò…÷eiþàÒþõvI×¥åƒKûO|¾-Ý?¶´øþ_RÛuiýàÒþŸýËÿËÒöÁ¥r;Øeiô^—öÿKýwSùiè¿Uy¹C̲,òö_Uo&6ý:{äu"grš×‰þ‡L Ë„æe"ÇñÇò1ayèÿi06Gš×‰þUªË„æe¢È²íç„åuB¯ÿœ8?ñ_¿ö‘/ó–ë_j?ÔÇ ùÊõçê…ý?²Š-L?µ°%Ðô¿¹ÿfaî?›AßÑÿÔ:ùO9}ß{¿þ3¿ë]|?uEþ}ÐÛê-ýß’þ¥š¬«ó_ý¿~üËc+:õ£~ýÓã¿=þùßøýt³_wñþë.ê¿ìAžWÚ(!äÿöwÿ·¿üÍŸþõüÃÿú»_þõñ÷ÿûñ¿ÿôw¿~ì·þõ{ÿjtëÿAšõWñÿýfé³`÷òe{öU¾&–¥ÿµ/ý×ÿùÏý õT–Oä753ówý3é¥ôÏüüZû—™e£`/ Òü5vËÇü·ñS“th‹5‹ûý ]û‰ùÐÐ/çzïæýrú¿Òý×Pè¿Ï×qg”²Â¢üðô6ÎñBÿM~¬ÓØÏp·™ÇØÿÓ5¶óõþÙìW×4/_îZÿÒÉ¢~Y,¯¯ZÝâù:ëYÏzÖ³žõ¬g=ëYÏzÖ³žõ¬g=ëYÏzÖÿŸ_o-ªÖŸ¯/Ö¢Z:£•Ÿsøçð»w汉n”oë~àïéo²@(YÚù~âYî4³('þ´Þ· †íÁuå;GVÈå|]oËZ]Ó6¹áÔ²|ßÛâÐÆë§6^·x¾ÎzÖ³žõ¬g=ëYÏzÖ³žõ¬g=ëYÏzÖ³žõÿ.ë¥EݼջڢjüeF+?çð-Îá7J;ƼË_šîo]å/úZ”·~þ 4ùû‹!di¢û:©I“ÑW»$}½Ž¿-´ºÇY´mѪüµ#‹C¯ŸÚxÝâù:ëYÏzÖ³žõ¬g=ëYÏzÖ³žõ¬g=ëYÏzÖÿ»¬mÖŸ²ëçhQ­ ÑÊÏ9|‹søMAÛd?•P6}ìØe‡‹òÖÏ Ú}¼ÐvïËF ²¯….Ó1xÙFêx]ž`ö£hûË!žX3×dVˆ‹5^¶x¾ÎjV³šÕ¬f5«YÍjV³šÕ¬f5«YÍjV³ú?Új«'µX|}±zÒzÆ­VœÃ·8‡ïÍg_\ö>å-û[µ-Ê[?m>Ç€ÚØÖ /)Ä:7¿ÖQ¶ú¯çëµÌ.÷Ö|nòTK²—½&³¶°Xãe‹çë¬f5«YÍjV³šÕ¬f5«YÍjV³šÕ¬f5«ÿíÖzR‹Å׫'­gœÑjÅ9|‹søMó™]-°Ey6oë*}³(oý´ù6—¿ù/»Èj a“§Œ3bŒò Ìãõ:6r]YÊ<-uÁöÝÒ°Æ««Õ_¶x¾ÎjV³šÕ¬f5«YÍjV³šÕ¬f5«YÍjV³ú?Üj­'µX|}±zÒzÆ­VœÃ·8‡ß4ŸÍyù{èÞ%i>kpÅÏ(oý¬ù”¿LµŒ RšFaª›]Ä\içëµÉ¦¤+{œD[°V,«¯®VÙâù:«YÍjV³šÕ¬f5«YÍjV³šÕ¬f5«YÍêÿ€«¥žÔbñõÅêIëg´Zqßâ¾7Ÿ5ÊþŸ!Uy˜Sëäö6£¼õÓæss9¹ÜÞ¯…« r»î.w¿>‚÷nO3Ë?í…›l§üæê(Ê5nMêññI1·Åóõ}ÿñŽ´ý4j<59àgÜ88°| üÿV²WKÚײ×JÛ#[I;ÇoqNß÷>¸mOB+m¼ýæZœYÞþY•|‘'´ÙšWVjÎÉe½‰ØŽÄèeßÜc"öÚêdy݇UÜ\l3O±O,☰¼L`````````````````````|ZC{im“¥õÖ^ÚÚå™g—<çïù˜Û{÷ÊxÀ\ÒÚ}w!Ì,gð¼÷ήJÿîgc®9çç¹#1U}RÞœhIvþXíSŒ~ûIÔ™§Ø'qLX^&000000000000000000000>­¡½´¶ÉÒ{k/míò‘­K>æïyοí½wéžC»W÷3Þm~f9ƒç½w“í´E.Ú˜[.®;w=ÒvWË2±7y‹Õžbu[»ŠeŸYEXE™Ð¼N`````````````````````|VC{im“__f/míò‘­K>æïyο»O·Tï­=âæ?²á¶¼}“MV²ÛGµ6—ÚÌrOkóå)“!f—Gåo9µùpJ;û[ĶLȃ%¯öƒËõ*Ö<³Šcâ"ö ËËƧ5´—Ö6Yjsí¥­]>²uÉÇü=Ïù7·‹‡äÂ6¶ßÆV.¡¸Rg–3xÞ{g—d? ýÇÌ)çjŽsG¤—Çrµ<뽓ÞMοȈe9ÿç­y’">Äæ¶}¬ÔœFÖ+×#ÍÛ,–KqÉ_ìSÜÒ*f¹­ÞòeâuBó:ñY í¥µM–Ö\{ik—l]ò1ÏsþýÖóê’ÏùÇló{ïîŠì\R\›û§ç(›«¬,+xÚ{gïzœK.s••K>[}=£—w„^­Ê«9wûcÛÅàŽbÙÄ1±‰sbåmãUÕKk›üþf½ôj—ϼºäsþ1Ûü绤ŒÕµ±$Ù¿%|¾MÊøò-ÏšÜÏGÚóáR·,7àŸÖæÙ…±ŒTÝ1)°r9äùtýpæ‘»;Žm¢GWûmbr¹Þ‹õ°¬âœ¸ÇÄÊÛÆËÚKk›,µ¹öÒ«]>óê’ÏùÇlóÏÅyÙÞ|üA‹û&/ï\YVð¼÷®.é&0QïNsõ®ÛÚç‘ÚÜâÅ&¢OÏŸOÉ߉=[VqN܉cbåmã… é¥µM–Þ[{éÕ.[¶.Ùæó9ÿ±÷î.Ê6+ãp”ÁÕjYVð´÷>¼¼ùSö?Ò¼RsÉç&0z$¦. _}ÜÝýÖã×ZÚ±‹Á…dY?9±‰sbåmã… é¥µM~“ é¥W»|æÕ%ŸóÙæ?ôÞGtÙÏíÄÛ|âüÈR¯,+xÞ{'ÙíE¾@×»Ó\ª;Ö/z$Æ$eþ9}–*~·/Ñç]Ì.EË&Ž‰Mœ+o/kh/­m²ôÞÚK¯vÙ²uÉ6ÿ˜Ïù½÷!›®Èvâq¾˜óh.d˲‚ç½÷|sf(EäJÍuÜ]XkŸG¢<ß¾MD_]ûyï]ä]ž›8,›Ó.Ή•· Œ—5´—Ö6Yzoí¥W»lÙºd›Ìçüç¯Å<²kcIòù^‹9¾|ÕG¸k•(Á¥dYnàim^¼óå&@Ÿ¬¬<ZVm®Gbë­¢+ÇC~W°ÛV›w—ã½X½eý0çÄ8&VÞ&000000000000000000000^Ö˜½ôj“ßßV/mí²eë’mþ1Ÿó×÷:·EY½wvG´,+ðO{ï$’l<žôîVžµû\»‰ãHñׄ¬mm~îï{ïe‘wbë–Uœwâ˜Xy›ÀÀÀÀÀÀÀÀÀÀÀÀÀÀÀÀÀÀÀÀÀxYC{é:Ûdé½µ—Öü‹eë’mþ1Ÿó{ïCö½Ì×nÞd?•,Ë ž÷ÞÅå|@†_¹çÃðz$Öêr¹&b,.ý¼÷>\ó»X¯–Uœ›8'VÞ&000000000000000000000^Ø^ZÛdé½µ—^í²eë’mþ1Ÿó{ïæÒq }>þ>Kùî-Ë ž÷ÞÝ•t“}X¢vöškvÙž÷žGbk{ô•}´'Â{ïæzßı–X,«8'.Q'VÞ&000000000000000000000^ØHº{HžÏ{Wí¥W»lÙºd›Ìçü‡Þ»y_fèÕõ¹ÓÊôݲ¬àiï]£ì@.që+=W®I6ñžk×#1.Äk"úæüý³ä—Ú.&—Ë&Ž‰Mœ+o/kh/]g›,½·öÒš±l]²Í?æsþcï]•g°»<6.+¨.6˲‚ç½wq>ÜBN®êÚ5Ë-«³×#ñèRæŸ1TySçn›x¸TïÅ’-«8'îÄ1±ò6ñ²Æì¥W›,½÷쥭]¶l]²Í?æsþcï]å%›²¯Iﲂæ]®–eÏ{ïî‚îV^tíš[’WuêÚ瑘«õþš£—Í\vÛÄæŽr/¶dYÅ9q'¶dy›ÀÀÀÀÀÀÀÀÀÀÀÀÀÀÀÀÀÀÀÀÀxY£Î^ZÛä÷7ë¥W»lÙºd›Ìçü··?|ûýŸÂ-ܾÿEþçÇípá¸ù›×?‘ Æb ò>Ìïï·ßüòÛÛ÷¿Óß¾|p!L>Ì[2îme¹§µy‹.µ[H^ö>‘+5—._wÞº‰épµ\1ÒäﶉAò&Ž/-«8'6qN¬¼M`````````````````````¼°!½´¶ÉR›k/½ÚeËÖ%Ûüc>ç?<.Þ²¼?3t¿Xoŵò¬àyï}¸C†÷.ÏÍÏW®M¾€®}‰±ËÞ/çDô‡Ýÿcïå·›X\ –Uœ›8'VÞ&000000000000000000000^Ø^ZÛdé½µ—^í²eë’mþ1ŸóŸ?.Þ’;Ò-Eø~x\ü‹¶ô…ç̫˲­¹—6ý}í>[–;Þ·7ù¥@ÈÑÝØE³ì“Ög6ÈoŽm"Æ$în_bˆ»Øå/ÐÊ&†¸‹sbåmãeÙK¯6ùýmõÒÖ.[¶.Ùæó9ÿ…¾½È¦é±yâýÓÚ||ùÒnó>æcê}|ÝdYnàimÞƒëãËåºp_¹Ž/¿¶gÑ#ÉWó5S±}Ïjóq>…]Œò)®lâ˜ØÄ9±ò6ñ²†öÒÚ&Km®½ôj—-[—lóùœÿ°=˸¾ÕÙ×·4W0/\YVð¼÷>\È·PÆ@˜Wj®Ùµ¶Ö>ÄÃÏÊß&¢oökƒÇÞ;»Ãß‹µYVqN܉cbåmãe í¥µM–Þ[{éÕ.[¶.Ùæó9ÿ±÷®Î—1ŸœŸÄôîJ°,+ðO{ïæRº…£Iu/WjnÙ…õ¨»‰W§®9VkÆýCï]eÑwb¯–Uœw☈WKñú†/Ö&Kï={ik—-[—lóùœì½eð(Ï`gWd{ðè£kÞ²¬àYï}p‡¼\ôÎåJÍí°-ÕבXæZΉ˜ÆMÿlŸYËÜÄ(›­¬¬âœØÄ9±ò6ñÂFÌÖ&¿¿ÉDóg»lÙºd›Ìçü·½wr9ÝbȲÁŠ¬àp½[–<ï½³<¶Ž!‡y¥fÝ\×>Ä£ÊCìçDLãKþìyo9ïë..–M›8'VÞ&000000000000000000000^Ö˜½ôj“¥÷ž½´µË–­K¶ùÇ|ÎûÐ{)È£¯ÂÈ º Õ²¬àyï=Fü¸›îš^©Y¶j±ç½ç‘x—ò5ã!Ûìö%Ʋ‹ÍÙ²‰cbçÄÊÛÆ˳—^m²ôÞ³—¶vÙ²uÉ6ÿ˜Ïùozﱘ07X)³yÑ¥bYVð´÷ÁyÙÅ%èÞä–[rUßÚ¹ŽÄ”e çD ‡lj¾Û&zy(ýN¬É²ŠsâN¬Éò6ñªÆꥵM~³^zµË–­K¶ùÇ|Î{ûÃ÷·ßÿi¹}ÿ‹üÏÛáÂqó7¯"ø±èd‘ßßo¿ù巷刺ÿxèÌÿø}|ùÙ†ËÛ5«Þâøº‡e¹çµyv±#­GÝW®Õ•õ¸¸‰Å˺ΠùCþéãâc-%ß‹=ZVqN܉=ZÞ&000000000000000000000^Ø^ZÛd©Íµ—^í²e¹0ÏÇÃuþ1ŸóßEÞ—)‡7í½›«Ù²¬àyï]]®·Ð‚ëzåÊÕU¿Ö>Äìå͜焬íÉãâEÞº‰Í…`YÅ9±‰sbåmã… ÛW?{oí¥W»lÙºd›Ìçü狇cî/^ƒüáóÇÅ»Kí&/ß Un Ž{K–åžÖæÑ»"{ÈD×æ“ò+·æüÚ!FÄØÖ‡³²ünàÞ¾Dw1¸Ø-›èã.Ή•· ŒW5´—^mòû›Lôt¶Ë–­K>{ì‡|Ρ6o²Ã‹¼Øs\ðim£;d³ó$»ÃÈ $Z–x^›'׎[x-Ü5×ê¢}8óHL~m½¾r(®üt—•q>†]ÌòfÑ•M›8'VÞ&000000000000000000000^ÖÐ^ZÛd©Íµ—^í²eë’mþ1ŸóŸ×æ1ÌÝÍ[••}^›²é¹l²~Ì÷ŽÆæR°,7ð¼6Ÿ½|(Å¥¹ÁÌÊ=»n¿u˜Gb¶‡õW ÷À˜X\ö÷b©–Uœwb©–· Œ—5f/½Úd©Íg/mí²eë’mþ1Ÿó_¨Í¹›ùå)öÏkó溼Ã3¬—’¦ ›Ã¬,7ð´6O^ŠüPª¼“ôýf¹^Úë!C»&d»õöóMZºtõwb+–õÃœwâ˜Xy›ÀÀÀÀÀÀÀÀÀÀÀÀÀÀÀÀÀÀÀÀÀxYC{im“ßßä”ôÒ«]¶l]²Í?æsþÛã;=St!ߢ¯RAË ²+ݲ¬àyï\·pççþ,+·x¾ÓSÄ<þ¯‰ùjÏvg_k©m³ó‡eçÄ&Ή•· Œ6B¶6YzïÙK[»lÙºd›Ìçü·½÷áRšÏw×¹ÏKª®5˲‚ç½wqEž+o.Í VV®ÍÕµ‡Œ‰eî¸~NÌ™Þ?K~­¥×]¬.fË*ΉMœ+o/lH/­m²ôÞÚK¯vÙ²uÉ6ÿ˜ÏùÏ·IIÙåkùÒ6)©¹C·.éó‘ödçô•åž×æÝÕ>–¤ —[×Ü“ÜÞú<"¿(ýšØëOwçCÙÄì]N–M—¨+o/kÌ^zµÉïo«—¶vÙ²uÉ6ÿ˜Ïùo‹ë«l]2Ææ6)9»X,Ë žöÞ9ºÞn¡4tíšÇWêëqq=›m=¾rÊ¢_bÊ»˜ä•M›8'VÞ&000000000000000000000^Ö˜½ôj“¥÷ž½´µË–­K¶ùÇ|ÎûÐ{g×ý-ú¾Þü™«Ë‡eYÁóÞ»¸Po!σʕš{qÇz#¨‰eõš'܃¿³M<Ü‘îÅ,«8'îÄ1±ò6ñ²Æì¥W›,½w•^ÚÚeËÖ%Ûüc>ç¿}è½›ó²;I”]QÆ ïJ¶,+xÞ{wy»g(ã‚0×®¹eW×æçzDn*o²…Küék1ÇZjÜı–ñ%WVqN\¢N¬¼M`````````````````````¼°!½´¶Éïo2!½ôj—-[—lóùœÿöØ{ÁÅz‹þpÇ|-æ‘\K–eO{ï#ÊV(¡Ï¹rå|®]Äè×Þá+ûfO³?ôÞc-=ìb’ãWVqNlâœXy›ÀÀÀÀÀÀÀÀÀÀÀÀÀÀÀÀÀÀÀÀÀxa#Êî!³M–Þ{öÒÖ.[¶.Ùæó9ÿùk1/O†‡q÷¯ÅüÂ…1%y„ý×_(_±–Ï_ÄyŒ;)ã£;ææ1Gu>Z–l~º?-êÙû%  è_ÍÍË×Ð{™{ÇÄkB6}‰÷ö%¿‹Å¥fÙÄàwqN¬¼M`````````````````````¼¬1{éÕ&KQ?{ik—ßΞZ»ä·³Ç¾Ïçü·¨WYLhu= ^Æ`°,+xÞ{7Ù»%”î¼>\¯¹ÎbtíóˆÜMˆ×„ü¡ÿtc–q~,z»;ªeÇÄ&Ή•· Œ—5´—Ö6ùýÍzéÕ.[¶.Ùæó9ÿí±÷×·<`÷óù’\ê–eO{ïe‘¡ŒÉÙê¯\£ËkCr=Ó¸(_1tÙKf·M .·{±Ë*Ή;±ËÛÆËÚKk›,½·öÒ«]¶l]²Í?æsþÛ‡Þ{,&ÝB Ò“Ë Š;šeYÁóÞûpIöS)®ëÝi®É•¾Ö>ÄØdo˜k¢W~ú¼÷XK©÷âø“•Uœwb?,o/lH/­m²ôÞÚK¯vùÌ«K>ç³ÍûÐ{WÙR%ô šË º«Õ²¬àyïÝÜáo!—ùŠÐ›åZ\²ç½ç‘ƒÝf¤Šßík-­ìbw![VqNlâœXy›ÀÀÀÀÀÀÀÀÀÀÀÀÀÀÀÀÀÀÀÀÀxaCzim“¥÷Ö^zµË–­K¶ùÇ|Î{콫—7tJÏ]æç5Ê;@W–<í½k*=äàêÜÅeeYÐÚÅED¬N}Mô³/Ñ»]Š–MôÇ.Ή•· Œ—5´—Ö6ùýM&¤—^íò™W—|Î?f›ÿö¡÷N®øm/–Z\8ö½Xž÷ÞÙµv Ñžf_9{WÚZû<c’›:'böÖÎÇÞ;»˜wq| Á²‰1ïâœXy›ÀÀÀÀÀÀÀÀÀÀÀÀÀÀÀÀÀÀÀÀÀxYC{im“¥÷Ö^zµË–­K¶ùÇ|ÎûÐ{Wû|ï¥ïsÝ¥lYVð¼÷nÎ×!'—µ³×\üÕêÏ#1R½ŸÑû'ïµ”—o¦{±zË*Ή;±zËÛÆËÚKk›,½·öÒ«]¶l]²Í?æsþÛ‡Þ»Ë ,å=–m® Ew$˲‚§½÷‰²‹K[oÄ\¹®é.&ëˆìVê5!kK÷Ï’›8ïÅÖ-«8'îÄ1±ò6ñªÆꥵM~³^zµË–­K¶ùÇ|Î{ì½[’/5{n]ãáj´,+xÞ{g—ó-¤(ź\©ùrKzwóH ^b¿&úaû›?öÞIÊöM_­YVqNlâœXy›ÀÀÀÀÀÀÀÀÀÀÀÀÀÀÀÀÀÀÀÀÀxaCzim“¥÷Ö^zµËg^]ò9ÿ˜mþÛ‡Þ»¸TfÏæ{-[s=X–<ｫ+ò4{–kÊ•šK Çuí󈼥3í½ZïÿØ{Wù›Øäõž+›èý.Ή•· Œ—5´—Ö6Yzoí¥W»|æÕ%ŸóÙæ¿}è½»;ä=–uíÅÒ£ Þ²¬àiïݽkñÒár™Wj>dD×¾Ž´<×~NÔäÚ½}‰¡ïb?YÙD™¸DмO`````````````````````¼¨±zim“ß߬—^íò™W—|Î?f›ÿöö‡ïo¿ÿS¸…Û÷¿Èÿü¸òbLóú'R¤WÙ(åW¾¿ß~óËooßÿ6NÿñЙÿñûüòU¶Ï.ÎGÚû!²²ÜÀóÚ<;æ‡Qæƒæ+/û¶è‡3ÄÐÜѶ‰–\-w¶‰I~[p'–òŠsâN,‡åmãeÕKÏ6YjóÕKk»|æÕ%ŸóÙæ¿=>.>®ïº¹IÖÞ»¹Ü,Ë ž÷ÞÕE?îfÈq^©Y z»»y$ú(u÷5Q³ËñÎ6±¸£Ü‹-[VqN܉-[Þ&000000000000000000000^ÖÐ^ZÛdé½µ—^íò™W—|Î?f›ÿö¡÷î.Äù|w‘Ö“®T˲‚g½w’7oöy7³1·|ŒµëóÞv¤Ï]Á¯‰·`¹ÖR] B­¬Ÿ†N\¢NhÞ'000000000000000000000^×^ZÛä÷7™^zµËg^]ò9ÿ˜mþÛCï|tIv')²YAv­X–<ｓ<”bÕVßò‘]ÕíÁíH;\HÛDóöµ–žw1»-«¨—¨š÷ Œ×5¤—Ö6Yzoí¥W»|æÕ%ŸóÙæ?öÞ‡”ðá(.‡¹‚æüaYVð¼÷.òvÏ0€ w§9×ûZ»iŵ°MToO³?öÞEn{«,beeâuBó>ñª†öÒÚ&Kï­½ôj—ϼºäsþ1ÛüÇÞ»¹²öóö²‚\Ì–eÏ{ï.ïî !¹¦½æœÐµë‘žä ›×DM²¹Ën_bŠ›¼+Þ²‰2qŠkBó>ñª†öÒÚ&¿¿Y/½Úå3¯.ùœÌ6ÿ¡÷×7yûWNÊ æÆà+Ë žöÞã6Â1ä(U½\©¹\w§Gbè²Åø5ÑçMﶉÑá^¬Ý²ŠsâN+o/kh/­m²ôÞÚK¯vùÌ«K>ç³Íì½çåÍš^ÆdÕ•`YVð¼÷..åy7EsÍåܽ{‰)¹£_q|ÉøÓç½Ç‰êïÅÞ,«8'îÄ1±ò6ñ†/Ö&Kï={ik—-[—lóùœÿØ{7É¡7—æ£wÍ[–<ï½»;Ò¼›>wqY¹Ž[°V‰²·x¿&b÷ÿóç½›<æ~‰c-¡ZVqN\¢N¬¼M`````````````````````¼°!½´¶Éïo2!½ôj—-[—lóùœÿöö‡ïo¿ÿS¸…Û÷¿Èÿü¸²ùˆ¿yý)ÒóXR ²²ïï·ßüòÛÛ÷¿Óß|åÂà»´úÞ—íü>Ö\ηЫ«yÞy’ ÈW–;Ú·Çèjœoõú©j Nk_q=KYÏ°¯ºüŠ`·/Ñ·]L.Ë&Ž‰Mœ+o/kh/­m²ôíÚK¯vÙ²uÉ6ÿ˜ÏùÏûöèe#òØ“<ŸþP›ÿñûÛ?Üþß›wÞ ¶ÿýŸÿõ¡Ã·ÿ6þïooAÎßþû[ .ô[ñòBM¹ÏÙ€k”Û¼?­éÈòÌý<›]j×Ù™¶³Õõ:;Óu¶ÎuÙYMÛÙñéåëìLÛÙ*o;=ÏÎtm^¶°±³š¶³IvÎ9Ïδ-òßœggÚÎvù]Çyv¦ë¬ü•¹> MÛÙââõihÚÎv—Ëuv¦ëlðÑ•ëÃZq?¸–·ó3îç»ói;?ãv>D¯lÅýü!;]çgÜÏ7ù]Îu~Æíüø›½^ÜŠûùñ§u;?ã~¾¹P¶ó3nçÓøÿ¢íóѸŸÏîØ>ûùêêöùhÜÎg/¿‰;ÏkÜÏg¶ÏGã~¾ÊrŸq;xù•Øy^ã~>Éÿï_çgÜÏÏïK×ù·óãŒß>?ûù$¿w»ÎϸŸ/òÍê:?ã~~|WÛ>?Ûù]Û>?ûù"o¾ÎϸŸïû·¡·ó-î߈VÜÏû·¢÷óÍÕíóѸïÁõíóѸŸߢ·ÏGã~¾¹´}>¯óÑw\ŸÏŠûùìjØÎϸŸ¯®ûíüŒÛùœ¿>¿÷óÙŶŸq?_]®Ûù·óòϦëó[q?Ÿ\;¶ó3îçÇ?yóv~Æí|òû÷§÷óiÿþ´â~¾ìߟVÜÏ÷ýûÓŠÛyùMüöùhÜÏùuüu~Æý|—®ó3nçèŽíóѸŸ?ä·Ý×ù÷óMö}»Îϸ/QÞ~ž×¸Ÿ?\Ú>?ûùæŽíóÓ¸¯Á•íóÓ¸ŸÏ®mŸŸÆý|s~ûü4nçÇOIqûü4îçóþýiÅý|Ý¿?­¸ï~ÿþ´â~>ïߟVÜÏWyÛËu~Æë|òã‡í§,ûùñcHÛÎϸŸ?ˆÔíüŒûù.g®ó3nçÃ|qÌy^ã~~n°wŸq??ÿCÄëüŒÛùýÇÆó§ÈëüøÑöã¦?þ:~ÞëçÝùÓèüŸßÍÿÕ㟃ãïƒñÿ㯦üÄz òC硫¿AÆJö Ó×.œï:î®Ì_¼2ò#ì~åñÅ+Ç°%ß]Y¾ve¡û+ë¯Ìã_?ÒÝ•í‹WŽÆÇÔû•ýkW¦ñciܯŒÞë•ã__Æ·Ä12þ"È¿‘i’ šþ{ÈÃé4œé:Æ?ç~™zzÅíüø±¬úë¼Æíüø¶Ñûu^ãu>ÆCþ±hçWÜΗùcÅy^ãu~ÝŽßînþ«ßúð¶ìwÛ9~d•¤6Ù´S>»ß•ûÏü?¼lü¼1~^˜—å_qYHsÝóºøk®?EýzþW\Ç·C¿Ü¯ùjòO³ªO ýš›“o²ã_Ðä²bŽ‡¾mseÿÃ?ßþ×í_þƒoeiýÍÿæŸæ¿s ,Í…–¯ú¿ùÿþñ··ÿò×?ßþúã_ÿýÿŽov¿+Ç×¾aŒÏwüü6þ-ìÐÿGúßþ§_ºõñ§ãÛÏ¿þǸòßþõ_ÆÝýÛÿùçÿ§_¶Oâ?Ó&¬ÿ7Õ¼¥qÏ祿X”eÇÐ/öÿ÷ц¿Í¿Ùoãø±}\åŸ Ùþz|m꧷rÿ¬á‡[‘½tÃøY½è‹Ýòø»süKùÊòÿµÏ<Ìqþ5ã§À2¿]¬<~êIë»™‰ãŸ³Íoãßûxg›èåÂ;qü€²²ŠsâNlÑò6ñªÆꥵM~³^zµËg^]òÕc?d›ÿôÁC¹`ü!„*×ýŠÿÐ](¿jôéÓÿÐ?Çä‚üJó¼óCr[Yîüyß>< Ç0Ó¼RsKçoõHÌMž#<'ä9“ã~“<`±‰‡üÂeçÄ&Ή•· Œ6‚·6YúöÙK[»lÙºd›Ìçüã år,.ö1ßlMþãí•eÏ{ïùŸÔGßä¿f“+5‡bÿõ¾©ö¾¬,ÿ}ò}§~­¥§]lò°ûÊ*êÄ%êD½þ[ Œ—7b·6Yzoí¥W»|æÕ%ŸóÙæ?Î<²«XÈ]úôÏï.·¹Õ¢÷)ÎM›4Ë <­ÍÓ<#»nùyë+÷(¿ ˜·®Gb:d ™sBÞ•WZ›ó!îbýÝV6qLlâœXy›ÀÀÀÀÀÀÀÀÀÀÀÀÀÀÀÀÀÀÀÀÀxUÃzéÙ&¿¿½´¶Ë–­K¶ùÇ|Îx\|\_ä¡ï䊾 a¾TaeYÁóÞ{¾EM^.Uô-š[—ùèÝÍ#1ë×Wþùãâã| »˜å½%+›8&6qN¬¼M`````````````````````¼¬1{éÕ&Kï={ik—-Ëúxøê±ò9ÿùãâ)Ì÷|ÖC®ûôqñ4_¶cqAo ÉËW–x^›W ½ÚËgŽ¡Ÿ»¬¬#e~F×D²ÎÝ6±È«`ïÄZ-«8'îÄ1±ò6ñ²†öÒÚ&Km®½ôj—ϼºäsþ1ÛüÇÇÅ»óÇ|<ü˜¬ ‹·,+xÚ{gïR¼…±Ö1 Wj®ã–ÖÚõHKìÛ„têýþQôk-¥ß‹cÑ+ë§1'îÄ1±ò6ñ†?¬M~[½´µË–­K¶ùÇ|Îè½óȺ[yÍ{ήv˲‚ç½w2äæRWjnÝ…Õêë‘Ø’¬åœN>ÿt›”±–Vw1Ky¿²ŠsbçÄÊÛÆ 1[›,½÷쥭]¶l]²Í?æsþcï}¸,{´¹ýøXAu½Y–<ｋì@ã¹vÍ9›Ÿ¯#Õ˯‰R¬콋óe«KÙ²‰2q‰:¡yŸÀÀÀÀÀÀÀÀÀÀÀÀÀÀÀÀÀÀÀÀÀxUcõÒ³M–Þ{õÒÚ.ŸyuÉçüc¶ùÏ·ÏY6/½Ë#çŸnž›¼>3Fïò¼Å#¸P-Ë ø§µyw­ßB?äyt¹õ•«+mÝú<S´‡á5Çæò}%‰cy—xxw$Ë&Ž‰KÔ‰•· Œ—5f/½Úä÷·ÕK[»üvöÔÚ%¿=ö}>ç¿P›W×ä™ö*…þ§µùøòMvŸïî”È.ËrOkócWŒfÙ¾\®ÔÜ¢kë…¢z$ï|º&ä·駵ùåžïÄ-«8'îÄ1±ò6ñ²Æì¥W›,µù쥭]¶l]²Í?æsþÃ.+ãú.»´ißeÕ‡eYÁóÞ»¸Xçû7×ÝiŽÑõ¼Ö®GÆ׎}›¨Cúé.+ÇáJº{°¬¢Nì¢LhÞ'000000000000000000000^ÕÐ^ZÛdé½µ—^íò™W—|Î?f›ÿØ{7äìC*鱂âe[ñ•eÏ{ïîr¹…Þ¤X—µÏ,¬„µMÊ:’Ú|#èí>ïöµ–7q¬%xË*ΉKÔ‰MÄÀÀÀÀÀÀÀÀÀÀÀÀÀÀÀÀÀÀÀÀx}cöÒ^þðþ&³—žù—3¯+ÎùÇlóŸo“rTÙ|npž·IùÊ…e¤/ì¯R‚¼Z4†èŽÙø—¹åúÊrçOûö]9n¡®èg6sôÁÅõÛ„u$ÙxýšŸAºßòü}ØÅ$¿vXÙÄ1±‰sbåmãe í¥µM–¾]{éÕ.ŸyuÉçüc¶ùÏûöâeuá˜ûš^›gwÈ6äÕ5½êB°,7ð¼6?\Ë·ÐÒÚØeåÞ]Y¿ŒÐ#òˆ|j×D ɵûGà/1ú],.7Ë&F¿‹sbåmãeÙK¯6YjóÙK[»lÙºd›Ìçüjó$ÑÇ£ÈÆ1Ÿ×æUªëçÆ Tï’·,7ð¼6ï²õz Íù8o}Ïzëz$.‡m"—^›7—ú½X꽨»(š÷ ŒW5´—^Æ›õÒË8óê’ÏùÇló_¨Í‹lƒº—%üšÝ]ôByi­ÿ‰ Cî®ûÏ‹úqÃ]ŸAó#«ÉånY>²§E}#qn*ïçï(V^¾Æü°×‘\×f;+ÇC~õ°Û&w´{±–UÔ‰]” ÍûÆ«ÚKk›,E½öÒ«]>óê’ÏùÇlóŸõunØê|ÿóÚ<»n¡®Îækq¥Y–x^›Ï_F„]Ö+W¾ö¨×#²ôT®‰è‹KéÎ61˯6±8Ÿ-«8'6qN¬¼M`````````````````````¼°!½´¶ÉR›k/½ÚeËÖ%Ûüc>ç¿P›§¹ýz‹<¾P›WÙµ=Æ`O›wÙŽfe¹çµy“jäu¥1Í+Wî×o ôH=Öž9+çäÚý NM¬®—]ìò'+«¨—¨š÷ Œ×5¤—Ö6Yjsí¥W»|æÕ%ŸóÙæ¿=Tß¹y9S—š}¬ %ç‹eYÁÓÞ»WeK/o+•+WŽ®®Þ{ÉUþpMäÃ(è½Çùpìbt9Z61»¨š÷ ŒW5V/=Ûä÷·³—ÖvùÌ«K>ç³Íáqñîj¾ÅÒœïŸ?.>¾|•ÝØ‹ óy÷Vdde¹çµyv]÷†/úáhŽ]>½u=Rªëi›(AÞ¼ºÛ—˜ò.®Ë&ÊÄ%ê„æ}ãU í¥µM–Ú\{éÕ.ŸyuÉçüc¶ù‹y®\6LóõÖ¥Ó_YVð¼÷nR‡ZæÊn–ЬòŸGb®ôkB~mPâmbuGº›·¬âœ¸›·¼M`````````````````````¼¬1{éÕ&Kï={ik—-[—lóùœÿÐ{w/oE_úyË}|édYVð´÷îÁ%yÔ}scqËÝÅÕê¯#ò†Îc›¨‡ ýξÖRã½Ø»eubeBó>ñº†ôÒÚ&¿¿Y/½Úå3¯.ùœÌ6ÿ±÷N.Êî$^v[‘®E˲/KùiïÝ!O³{—ç“â–óÙê¯#ã~sØ&RGÐwûZK÷»8n¢YVQ'.Q'4ï¯kH/­m²ôÞÚK¯vùÌ«K>ç³Í¾=xîð·P¼<‰þ°=ø.ŒGw©~¾¯x/.—[ôÕúö.¿ XYîüyß^]M· ;«ûyåÌÑgÙ"F?3=’Æ=ï±È‡»Û—èû.6¹…•M›8'VÞ&000000000000000000000^ÖX½ôl“¥o_½´¶Ëg^]ò9ÿ˜mþ }û!{»„R\Œ_¨Í»+Ç-†ê’Ô懲ŸùÊrÏjóÃ{×ã-Œ[-q^©¹\ç•ëH<¢ÙšSu!ßÙ—Û.wË&Ž‰Mœ+o/j¬^zµÉïo«—¶vÙ²uÉgýÏùÇíYäú&¯ß,ÂÈ —ºeYÁóÞ;»n¡´¹qÌÍrë.öuwóH<ý¥€åØdI»mbr¹Þ‹ãüÊ*Ή;qL¬¼M`````````````````````¼¬1{éÕ&Kï={ik—-[—lóùœÿØ{çeòõBÐÃ7w4˲‚ç½wuÉO¹óʙ㒭]¤äRØ&â!oåÜík-¥Ü‹=[VqN܉=[Þ&000000000000000000000^Ø^ZÛdé½µ—^íò™W—|Î?f›ÿt›)ÒãXt-â~¶MÊịºK‹>°‚«Õ²ÜÀÓÚ\6gï·Ð›ëm^9s ]¶QŸ·¾ŽI¶;¿&RvþçÛ¤ÌÏuƒ ɲ~˜sbçÄÊÛÆ ÒKk›üþ&ÒK¯vùÌ«K>ç³Íx\ƒÒ\A“}ÅW–<ｋk² Œw¡Î+gŽ>Z«mGŽîªß&r'ÑwûcÚÅêŽ`ÙĘvqN¬¼M`````````````````````¼¬¡½´¶ÉÒ{k/½Úå3¯.ùœÌ6ÿ±÷n®ÊSÚã ÏÞ;)¼W–<í½£w^÷S sí+·*_p®]Äì¥î>'bˆ²ËnÛZºËñ^ÃÊúỉ;±zËÛÆ˳—^mòûÛꥭ]¶l]²Í?æsþCï=®ïõúÜpEVÝ-Ë üÓÞ;‰ŽæbœWj®ó=síz$¦¶^¹¹²l½òÓÞ[ÞÅîÅÖ-«8'îÄÖ-o/kh/­m²ôÞÚK¯vÙ²uÉ6ÿ˜Ïù½÷áB¹ÉžãU»ùêj°,+xÞ{y„<Œµê+=W.]ÞékŸGdÏ•T¯ Ù;¼†;ûZKó»X¯–Uœ›8'VÞ&000000000000000000000^Ø^ZÛdé½µ—^í²eë’mþ1Ÿó{ï±y;¹2»ùä]÷–eÏ{ïîJšzÔ'Å5·âŽc­}™wÓ®‰¹³¯µô¾‰c-±XVqN\¢N¬¼M`````````````````````¼°‘k“ßßd¢û³]¶l]²Í?æsþCï‚;äÝ›E¶Þ–dç»eYÁÓÞ;E×â-Ⱥv͵¹ºžU×#²×xM×DŒQ–´Û—Ú.&—Ë&Ž‰Mœ+o/kh/­m²ôÞyöÒ«]¶l]²Í?æsþcïåKÉþßE×X]l–eÏ{ïâ|¸…\å+É•škrm=Í®Gd-òNN›wf¶ûwfšxÈcîwbÉ–UœwbÉ–· Œ—5´—Ö6Yzoí¥W»lÙºd›ÌçüÇÞ»º.ï±LRdïrµ,+xÞ{wú¼›ª;¤h.M@×µÏ#1Ø>/+ûèúÏŸ÷nî(÷⸉•UœwbK–· Œ—5´—Ö6ùýÍzéÕ.[¶.Ùæó9ÿ¡÷ÎÁye–V{¬ 'WŠeYÁÓÞ;GÙŒ%´ÃUmÌWî²™É\»‘^÷P±ì]¿ßCåZK=v1ɦä+«8'6qN¬¼M`````````````````````¼°!½´¶ÉÒ{k/½ÚeËÖ%Ûüc>翽ýáûÛïÿ4ŽÜ¾ÿEþçÇípá¸ù›×?‘"=å[8º,òûûí7¿üööýoãô÷_¹P:û/¼/Ûÿø}¬+»äo¡z×æ=sqí°,wþ¼o?Ü!o.觪¹Dç×SôëH/òÑ\ãs÷÷ïê41»žw±¸,«¨—¨š÷ Œ×5¤—Ö6Yúví¥W»|æÕ%ŸóÙæ?>g^]–]Q’+]Vpxç³eYÁóÞ»¹*OÑ7¹Y»æ2… k×#­¸P·‰Ze›•Ý¾Äw±»ì-›(—¨š÷ ŒW5V/=Ûä÷·³—ÖvùÌ«K>ç³Íá9ó"ω‡ØäAõO_¾È6çÑ¥YÜÉÅdYnàim~Ù&Yö\‘+5—vÖæz$†";È\=ºvÿªÏKLa£;ºeÇÄ&Ή•· Œ—5´—Ö6Yjsí¥W»|æÕ%ŸóÙæ¿P›w×Ç’Ru>}¡6O²Jè‡|9¹±òhYnàym~ÈCó¡t×´p×ÜÒõáÌ#²ÃzËÛDïò ÓÝ61»Ãß‹µYVqN܉µYÞ&000000000000000000000^ÖÐ^ZÛd©Íµ—^íò™W—|Î?f›ÿ¼6„±$yîýîæ?©Íçþëa¬â˜/=º+Á²ÜÀóÚ¼¹¤¯6-zåÊÍeûpôHµm¢ïŸd7±ÊÎ0wb¯–UÔ‰]” ÍûÆë^·ÏeÖæÚK¯vùÌ«K>ç³Í¡6/.Ž%/ÏÁÿš—ê…Ñ'×üç}{ñ.êÃæ%Ê—(—­,wþ´o/ÁñŽ([¶È•+wç×úz$Ƽ>Õ5Ñç/vÛD/ÏÆobt¡XVqNlâœXy›ÀÀÀÀÀÀÀÀÀÀÀÀÀÀÀÀÀÀÀÀÀxaCzim“ßßdBzéÕ.ŸyuÉçüc¶ù/-Éå4Ï÷ù(}9\ï–eÏ{ïìª4ÿÑ¥ùÑ+Û®æëH¯òÈü5QŠKùξD_wñpé°l¢L\¢NhÞ'000000000000000000000^ÕÐ^:ǵ«yÑ^Zó/g^]ò9ÿ˜mþóçÌKte,É'y~ý×ÀÝ]“‡ªƒës5ºT,Ë žÊ58߇œe7¹Rs®îX…¯#½˜­Y¾äOåê]>îÅš,«¨»(š÷ Œ5V/­mòû›õÒ«]>óê’¯û!Ûü‡Þ»Î:[6×­SêáŽÃ²¬àyï]lcr×»;s´»Ó#%ËÞ*ׄîÖ²Û×ZJ¾{´¬¢Nì¢LhÞ'000000000000000000000^ט½ôl“¥÷^½´¶Ëg^]ò9ÿ˜mþóKj”gÍ{t½}¾ýH-.ô[HÁ%mÍ›«Ù²¬ÿyk^]–ýQ¢TáråÊç 2íÈáåW×Dʲ·Ën›X\K»Ø\–UÔ‰KÔ ÍûÆëÒKk›,­¹öÒ«]>óê’ÏùÇlóŸïRpËÓäÚÛ·àz²,+ðÏzïæ])·×æ“î+§èrе¯#ã)mÕËMïö%ú¸‹ÁÅnÙD™¸DмO`````````````````````¼¨±zim“ßßdBzéÕ.ŸyuÉçüc¶ù½w‹î¨óirÙ…e¬àp!Z–<ｓkǼ›æ•šSqy½ÄséA:øk¢fÙØe·/1†]Ì.7Ë&ÊÄ%ê„æ}ãU í¥µM–Þ[{éÕ.ŸyuÉçüc¶ù½÷ájÑž[WÐdq+Ë ž÷ÞÕù<î¦ÚÝiŽÝõºÖ®Gzq>m¥¹òÓ}NÚX¤¿Kµ¬¢Nì¢LhÞ'000000000000000000000^ÕÐ^ZÛdé½µ—^íò™W—|Î?f›ÿØ{7×e·ñîÊ\Aîð–eO{ïîeÃñ‚ì#.WjNQ@Ÿk_GÆZÚ&ªwG¾³m-Ýý^lŲ~:±‹2¡yŸÀÀÀÀÀÀÀÀÀÀÀÀÀÀÀÀÀÀÀÀÀxUC{im“ß߬—^íò™W—|Î?f›ÿÐ{÷è‚<ƒ¯sÙ•nYVð¼÷N²uJðeî<~³œƒKë½–ëˆÜTÞ&ds—Ÿîs2ÖRÛ.fçË*êÄ%ê„æ}ãu 饵M–Þ[{éÕ.ŸyuÉçüc¶ù½÷![‰‡£¯woöêZ³,+xÞ{Ûð·×ú¼RsÎòP·®]”CÊük¢—ã}­¥×]¬.fË*êÄ%ê„æ}ãu 饵M–Þ[{éÕ.ŸyuÉçüc¶ùooøþöû?…[¸}ÿ‹üÏÛáÂqó7¯"EúQn½¸oßßo¿ùå··ïg¿ÿx¨Ìÿø}èÍñ|wEvr)>Ècâ+ËúŸ·æÝÕ~“'ÀµðŸ1ŒëÏMRôHì²?ù5ƒ= ÿXšwÊ ï]N+š7ΟÞXy›@@@@@@@@@@@@@@@@@@@@ø;Vátû‘³ðÕÞöÌ«¥=ç³Í¡Ž®RóÊ(Gþ´–/_õš^o »X,Ë <룋² ‹ì;®Îü!Ï¡Ë…v¤ŒlÛDÊÒàïô¦¼Iv×h^Ê»7VÞ&þN-|µ§•>Z ßÕÛžyµ´çüc¶ùOûh©Ä{¾Å±’ö•>z|yÓetU÷êòaYnày]\¨·:ß )Î(û€[=Èþ$5\R´§Ÿ=]ÆéÎkaEõæù;o ¬¼M ü ³ïÕšVÚèY÷®ÖÖ¢u´6ý˜müq‹êâ›óÝÆÇ—Þ•ü³Ê÷c™Ü]*~®{FÙ»ÛºõH ÎÇm⨮ÿìÍŒ²”/p¬ÄûÕ›ç/OVÞ&þn)|µ§}“ )|Wo{æÕÒžóÙæ?Ê!¸Xçöé˜+H²KÇʲ‚§…rˆî8n=ËõráŒ!d××ãÉëHkvsšïr¿£¯¥Œ5^`r¡¯¨Þ<¿ys`åmáïV맕BY ßÕÛžyµ´çüc¶ùÏwËÞá¥êþÂnÃͲYG”-6䪬|e¹ç}ô!_`Üi)óÂC,®¬}BÖ‘’]Ý'Žg»7óÁo`q©­hÞ8¿ys`åmáïTÐÂW{Z飵ð]½í™Û­yÍ?f›ÿB\Kê‡\ðy]]9n!kç~Ñ»,Ë <¼ß0øæòܸzå˜åÐ[×#ããZm¾æRŸ= =û.Îw2®lb컨š÷ ŒW5´—Ö6ùýÍzéÕ.ŸyuÉçüc¶ùû‚Œë[¾…ÔÜ1$É¥nYVð´÷ŽÑ…xëöºÆ0þPt‡j;RÆÒ6qtyT}§ ò¹ì`-+ª7Ïßyc`åmáïTÐÂ7ΞV e-|5ÿræÕÒžóÙæ?{\àÛ-†êRüü9ì˜O·P³l˜-7PÜÑ,Ë <Âl×Ù¶¯œ³ìs¢·>DåK\Ý»òÓAÆZJ½ûaYÅ9q'Ž‰•· Œ6ÆŸ®6Yjsí¥W»|æÕ%ŸóÙæ?>‡]]”‡£›kóýŒ±»Z-Ë ž÷ÞM^3Â᪮]sjçÛ'õHôÙu¿M´C ßík-­ìbw![VqNlâœXy›ÀÀÀÀÀÀÀÀÀÀÀÀÀÀÀÀÀÀÀÀÀxaCzim“¥÷Ö^zµËg^]ò9ÿ˜mþCï¼Ëa¾vRß)/€,–eO{ï\‘ÝU¢Ü‚\©9uÙe®]ÈÝh«¿&J“'Üwûý±‹QY_ÙDìâœXy›ÀÀÀÀÀÀÀÀÀÀÀÀÀÀÀÀÀÀÀÀÀxYC{im“ßßdBzéÕ.ŸyuÉçüc¶ù½wrEßN™tE6KYYVð¼÷ήÉ/Avv‘+5·ù]ûüÓ(ï®l×DôÕ•ûNýcÞÅÃÁ²‰cbçÄÊÛÆËÚKk›,½·öÒ«]¶l]²Í?æsþcï]\•7^Ž1]cw)[–<ï½›ó²#J–­YäJ͵»fÏ{ë‘^\¯Û„æÝ6±ºœîÅê-«¨»(—ˆñê†öÒÚ&Kï­½ôj—ϼ®8ç³ÍþZÌqO³zÏþñµ˜_¹0&ÙpüÓ÷iŽuwyfriîð’£;’e¹ó§}{.Êïã«Î¦~åâåÑùù™é‘8þ°l;Z#ÿзgïJ¼[·¬âœ¸ÇÄÊÛÆ««—Ö6ùýÍzéÕ.[¶.Ùæó9ÿá9óœ\g¿³­ñp5Z–<ï½³Ëò}”GãåJÍ­¹Ø×ÝÍ#Q6Ù&¢oÏzï$'6q|µfYÅ9±‰sbåmã… é¥µM–Þ[{éÕ.[¶.Ùæó9ÿ±÷..•[èÕåù,|n®˲‚ç½wu%Í7w¦ù&Ε{tm퉮GbHòÊÏs"úär»³/q|‰M‹®–M›8'VÞ&000000000000000000000^ÖX½ôl“¥÷^½´¶Ë–­K¶ùÇ|ξ¿J>Ü!²D)Î?Ý&%wwcÁ‡‹scô#ºà-Ë <­ÍïšøÕ¥Y¸k–[÷Qo}©^žX¿&ruîö%†¾‹ÁåbÙÄ1±‰sbåmãU ë¥g›üþvöÒÚ.ŸyuÉçüc¶ù‹ëk维벂Cž+_YVð¼÷ÎÎËÃðEz{¹Rsò½»y$sWðs"Æh½wr©Þ‹å°¬âœ¸ÇÄÊÛÆ˳—^m²ôÞ³—¶vÙ²uÉ6ÿ˜Ïù½÷áºle’eLVÐä9ó•eÏ{ïꢿ…dg¹ræè½Ü‚®]Œ/qäm"vçï;u‹;ʽزeçÄ8&VÞ&000000000000000000000^ÖÐ^ZÛdé½µ—^íò™W—|Î?f›ÿØ{wdw¿ºù\©–eO{ïâ]’ÞKu/WÎ}¶]LìH:\mÛD ²Ën_kyÇ•,ë§1'6qN¬¼M`````````````````````¼°!½´¶Éïo2!½ôj—ϼºäsþ1Ûü‡Þ»Ì7pÆX]HsÙµbYVð¼÷NîhºƒÊ|VÝrwÞzo=’úz-¨åf;€?ôÞc-=ïbv1ZVQ'.Q'4ï¯kH/­m²ôÞÚK¯vùÌ«K>ç³ÍþZÌäYóPºìIòðZÌ/\kv©}þ>Ír¸Üo1÷õ,|i²À•åΟ÷íEÞC‘“+5Çñ¿>3=" JÛÄ‘$ïö%†´‹Õå`ÙD™¸DмO`````````````````````¼ª¡½´¶ÉÒ·k/½Úå3¯.ùœÌ6ÿñ9óæŠ>ûÝ»¬ ³eYÁóÞ»»^tç–yåÊò„ûzF~©eî‰~NŒ»ë÷{–_bŠ›XÇŸxË&ÊÄ)® ÍûÆ«ÚKk›üþf½ôj—ϼºäsþ1Ûü‡Þ{\ßêMÞ·â\Av9Z–<í½k’ÇËC+òD»\¹ç¹v="{»Ä}¢7Û_æ¡÷®Q¯¿k¿çÄX»åmãe í¥—ñf½ô2μºäsþ1Ûüçû«Œ B¸õäRúUÛ«èuò`yÿ‰ ÃqÈB?Ý—¥²9MhÝ5ýĪ+Á²|bÏ{úâR¾õCž¨— g e|>ëñt=S¶Gß5‡äJ¹£ <äúìmEõæù;¯7ËÛÂß­ …¯ö´R€ká»z[ËÖÒÚüc>ç?>øÝ$‡V×›wÍ[–†Ý‚Ëyèmme>.ëݲ¬ß?kÍ[t5ÞÆàªÛg Å»¸n\Ä8Ö_®‰Æ*ý}¾m`r©¬hžo»7VÞ&þN-|µ§•:Z ßÕÛZ¶–Öæó9ÿíínÿïÍ;ïýmÿßÿù_êÝpûoãÿþöäüí¿¿å\è·áô­Ö( }8¿â‘å±r=Ÿ]jÛù÷óÕu;?ãv>Íõç5îç“ì8~Ÿq?_åŸ×ù·óÙËVèçyûùYÐ_çgÜÏÏ7{^çgÜÏwùÍÃu~Æíü‘äo…ó¼Æý|qqû|4îç»Ëe;?ãv¾DW¶ÏOã~þpmûü4îç»óÛç§q;_£ìsž×¸ŸŸÿñÁu~Æý|“­ç¯ó3nçÇßÈuûü4îçÇŸnŸÆý|saû|4nç{piû|4îç³;¶ÏGã~¾ºº}>¯ó‡÷ò ";¿â~>Ë/©®ó3îç«K~;?ãv~ü?j¾>Ÿ÷óI6â¿ÎϸŸ/²Ðu~Æí|ôòßsœç5îç“ü²ê:?ã~¾È7¢ëüŒûùîJÚÎϸOQv:ÏkÜÏygÁu~Æý|ß¿?­¸Ïqÿþ´â~þØ¿?­¸Ÿo®nŸÆíü\ß>ûùñ{û|4îç›ì uŸq;/ï*Þ>ûùìêöùhÜÏW×·¿¿4nçkp~ûü4îç³üæ:?ã~¾º¼}~·óÍË8tž×¸ŸO®mŸŸÆý|u~ûü4nç»ß¿?­¸ŸOû÷§÷óeÿþ´â~¾ïߟV¼ÎåOÏ~jÜÏùMôu~Æý|— ®ó3nçCtÇõù¬¸Ÿ?\ÍÛù÷óÍõ´Ÿq;£¼‘ã<¯q?ȯ ¯ó3îç›;üv~Æí| òV’ó¼Æý|–ÿì:?ã~¾9¿}~·ó9ÈÏTçyûù¼Zq?_÷ïO+nç¿Zq?Ÿ÷ïO+îç«‹Ûç£q;?ΤíóѸŸOò†•ëüŒûùñÉöùhÜÏÏ—‚_çgÜÎ×ùò”ó¼Æýüܵí:?ã~~þ7x×ù·óû”çÏ—×ù?|ÿÉO­?þúöû?ÅõhÅüAuþÏïæÿê,¨·ñí~üå”Ç+BAž°øüÊñ·ÈXË~eúâ•u~sÚ¯Ì_»òð‡üX»_y|ñÊñmÉwW–/^9þR÷WÖ¯]Yüø—‹tweûâ•ãÇóñãõ~eÿâ•ãÓ¸_½×+ÿøý-ïŠcfül6~ö?„i”×Uý¯Z? $ù)Vãv>ŒŸþcºVÞ'Úü uMhÞ&âø6~B?'VÞ't]×ĵÎùÌÐú ¶›ÿÝö¡¤±êñOÈ6)ù ~—ï?Àÿøº0¾ºÓäï⯺püf× #ý¯¹Nþ¥vüD:ÿ’ýªëêüN)×eûë<þö–Ê6Zö?üøóíÝþå?øŽ‘Ößaññï°y@~´ÿèMiþ¬|ÝüÍ?ÿãooÿ対ýõÇ¿þûÿ½µð»Ö¿öÿ[ãÇÈñcÒø‡í¡·þï?ÿÓ¯Ý}õòdŸ#õîÿǸôßþõ_ÆíýÛÿùçÿ§_¶ã?õ/ôú7áú†ûVÆ]Ÿ?ÿýb1… C¿Øôc¶ñooÿðöÿq½[J endstream endobj 761 0 obj << /Length 359 /Filter /FlateDecode >> stream xÚRMKÄ0¼÷WäØ‚Íæå;ÝuAñ"ô¦â6»[è¶UØoÚ¤Z/hó&“—™"~.±ˆ‚“èÐ$o J:VåÜ‚x,›»(ÚuÉ£_K/šùJô¦H6{ÍÁFR‰Š#KÃká‘=¥ûêôÞ»,gÒ¤@¯³\pnmkK{åQcÒÆÖn„.‘~Lׄ‹Èzp.ßÚ~\MWº:Êõ®¶c•Ñål` ®ªWã%Þp {Ye/Åýfoäj~FÖþE$ŒþL€ÖÏWb£ôÂq½kÇ iOË3Æ.쇳mOqšª]z£­'U”3XËÉ\ÀÆ›; 6]ï Ó¼R¤x"&·ÅW1à+õ¯hÙŸÑ®DG 0ç€amÄÏh© )0Ÿòí„ÎÑÆÆÎÙñ> /ExtGState << >>>> /Length 5932 >> stream q Q q 59.04 73.44 378.72 335.52 re W n 0.000 0.000 0.000 RG 0.75 w [] 0 d 1 J 1 j 10.00 M BT /F1 1 Tf 1 Tr 7.48 0 0 7.48 70.10 359.09 Tm (l) Tj 0 Tr /F1 1 Tf 1 Tr 7.48 0 0 7.48 78.07 339.04 Tm (l) Tj 0 Tr /F1 1 Tf 1 Tr 7.48 0 0 7.48 86.04 340.06 Tm (l) Tj 0 Tr /F1 1 Tf 1 Tr 7.48 0 0 7.48 94.01 340.98 Tm (l) Tj 0 Tr /F1 1 Tf 1 Tr 7.48 0 0 7.48 101.98 336.39 Tm (l) Tj 0 Tr /F1 1 Tf 1 Tr 7.48 0 0 7.48 109.95 346.70 Tm (l) Tj 0 Tr /F1 1 Tf 1 Tr 7.48 0 0 7.48 117.92 313.89 Tm (l) Tj 0 Tr /F1 1 Tf 1 Tr 7.48 0 0 7.48 125.89 337.12 Tm (l) Tj 0 Tr /F1 1 Tf 1 Tr 7.48 0 0 7.48 133.86 326.43 Tm (l) Tj 0 Tr /F1 1 Tf 1 Tr 7.48 0 0 7.48 141.83 318.49 Tm (l) Tj 0 Tr /F1 1 Tf 1 Tr 7.48 0 0 7.48 149.80 317.46 Tm (l) Tj 0 Tr /F1 1 Tf 1 Tr 7.48 0 0 7.48 157.77 310.69 Tm (l) Tj 0 Tr /F1 1 Tf 1 Tr 7.48 0 0 7.48 165.74 302.80 Tm (l) Tj 0 Tr /F1 1 Tf 1 Tr 7.48 0 0 7.48 173.71 314.78 Tm (l) Tj 0 Tr /F1 1 Tf 1 Tr 7.48 0 0 7.48 181.68 303.23 Tm (l) Tj 0 Tr /F1 1 Tf 1 Tr 7.48 0 0 7.48 189.65 307.08 Tm (l) Tj 0 Tr /F1 1 Tf 1 Tr 7.48 0 0 7.48 197.62 268.32 Tm (l) Tj 0 Tr /F1 1 Tf 1 Tr 7.48 0 0 7.48 205.59 294.42 Tm (l) Tj 0 Tr /F1 1 Tf 1 Tr 7.48 0 0 7.48 213.56 293.54 Tm (l) Tj 0 Tr /F1 1 Tf 1 Tr 7.48 0 0 7.48 221.53 285.33 Tm (l) Tj 0 Tr /F1 1 Tf 1 Tr 7.48 0 0 7.48 229.50 292.74 Tm (l) Tj 0 Tr /F1 1 Tf 1 Tr 7.48 0 0 7.48 237.47 251.85 Tm (l) Tj 0 Tr /F1 1 Tf 1 Tr 7.48 0 0 7.48 245.44 259.33 Tm (l) Tj 0 Tr /F1 1 Tf 1 Tr 7.48 0 0 7.48 253.41 271.84 Tm (l) Tj 0 Tr /F1 1 Tf 1 Tr 7.48 0 0 7.48 261.38 261.43 Tm (l) Tj 0 Tr /F1 1 Tf 1 Tr 7.48 0 0 7.48 269.35 256.54 Tm (l) Tj 0 Tr /F1 1 Tf 1 Tr 7.48 0 0 7.48 277.32 221.27 Tm (l) Tj 0 Tr /F1 1 Tf 1 Tr 7.48 0 0 7.48 285.29 259.30 Tm (l) Tj 0 Tr /F1 1 Tf 1 Tr 7.48 0 0 7.48 293.26 252.63 Tm (l) Tj 0 Tr /F1 1 Tf 1 Tr 7.48 0 0 7.48 301.23 242.81 Tm (l) Tj 0 Tr /F1 1 Tf 1 Tr 7.48 0 0 7.48 309.20 241.76 Tm (l) Tj 0 Tr /F1 1 Tf 1 Tr 7.48 0 0 7.48 317.17 237.95 Tm (l) Tj 0 Tr /F1 1 Tf 1 Tr 7.48 0 0 7.48 325.14 223.36 Tm (l) Tj 0 Tr /F1 1 Tf 1 Tr 7.48 0 0 7.48 333.10 236.22 Tm (l) Tj 0 Tr /F1 1 Tf 1 Tr 7.48 0 0 7.48 341.07 220.15 Tm (l) Tj 0 Tr /F1 1 Tf 1 Tr 7.48 0 0 7.48 349.04 200.67 Tm (l) Tj 0 Tr /F1 1 Tf 1 Tr 7.48 0 0 7.48 357.01 156.14 Tm (l) Tj 0 Tr /F1 1 Tf 1 Tr 7.48 0 0 7.48 364.98 191.07 Tm (l) Tj 0 Tr /F1 1 Tf 1 Tr 7.48 0 0 7.48 372.95 188.09 Tm (l) Tj 0 Tr /F1 1 Tf 1 Tr 7.48 0 0 7.48 380.92 192.73 Tm (l) Tj 0 Tr /F1 1 Tf 1 Tr 7.48 0 0 7.48 388.89 165.53 Tm (l) Tj 0 Tr /F1 1 Tf 1 Tr 7.48 0 0 7.48 396.86 171.70 Tm (l) Tj 0 Tr /F1 1 Tf 1 Tr 7.48 0 0 7.48 404.83 162.30 Tm (l) Tj 0 Tr /F1 1 Tf 1 Tr 7.48 0 0 7.48 412.80 134.21 Tm (l) Tj 0 Tr /F1 1 Tf 1 Tr 7.48 0 0 7.48 420.77 131.07 Tm (l) Tj 0 Tr ET Q q 0.000 0.000 0.000 RG 0.75 w [] 0 d 1 J 1 j 10.00 M 112.92 73.44 m 431.70 73.44 l S 112.92 73.44 m 112.92 66.24 l S 192.61 73.44 m 192.61 66.24 l S 272.31 73.44 m 272.31 66.24 l S 352.01 73.44 m 352.01 66.24 l S 431.70 73.44 m 431.70 66.24 l S BT 0.000 0.000 0.000 rg /F2 1 Tf 12.00 0.00 -0.00 12.00 106.24 47.52 Tm (50) Tj /F2 1 Tf 12.00 0.00 -0.00 12.00 185.94 47.52 Tm (60) Tj /F2 1 Tf 12.00 0.00 -0.00 12.00 265.64 47.52 Tm (70) Tj /F2 1 Tf 12.00 0.00 -0.00 12.00 345.33 47.52 Tm (80) Tj /F2 1 Tf 12.00 0.00 -0.00 12.00 425.03 47.52 Tm (90) Tj ET 59.04 98.78 m 59.04 361.69 l S 59.04 98.78 m 51.84 98.78 l S 59.04 186.42 m 51.84 186.42 l S 59.04 274.05 m 51.84 274.05 l S 59.04 361.69 m 51.84 361.69 l S BT /F2 1 Tf 0.00 12.00 -12.00 0.00 41.76 86.94 Tm (-1.5) Tj /F2 1 Tf 0.00 12.00 -12.00 0.00 41.76 174.57 Tm (-1.0) Tj /F2 1 Tf 0.00 12.00 -12.00 0.00 41.76 262.21 Tm (-0.5) Tj /F2 1 Tf 0.00 12.00 -12.00 0.00 41.76 353.35 Tm (0.0) Tj ET 59.04 73.44 m 437.76 73.44 l 437.76 408.96 l 59.04 408.96 l 59.04 73.44 l S Q q BT 0.000 0.000 0.000 rg /F3 1 Tf 14.00 0.00 -0.00 14.00 44.73 442.09 Tm (Canada, males over 45, Lee-Carter model: relative sensitivity) Tj /F3 1 Tf 14.00 0.00 -0.00 14.00 102.47 424.81 Tm (of different ages to change in total mortality) Tj /F2 1 Tf 12.00 0.00 -0.00 12.00 237.73 18.72 Tm (Age) Tj /F2 1 Tf 0.00 12.00 -12.00 0.00 12.96 217.52 Tm (log\(beta\)) Tj ET Q q 59.04 73.44 378.72 335.52 re W n 0.000 0.000 0.000 RG 0.75 w [] 0 d 1 J 1 j 10.00 M 73.07 352.73 m 73.07 370.64 l S 81.04 331.98 m 81.04 351.29 l S 89.01 333.36 m 89.01 351.94 l S 96.98 334.69 m 96.98 352.45 l S 104.95 330.15 m 104.95 347.82 l S 112.92 341.22 m 112.92 357.37 l S 120.88 307.06 m 120.88 325.91 l S 128.85 331.73 m 128.85 347.69 l S 136.82 320.80 m 136.82 337.27 l S 144.79 312.80 m 144.79 329.37 l S 152.76 311.90 m 152.76 328.21 l S 160.73 305.08 m 160.73 321.49 l S 168.70 297.04 m 168.70 313.75 l S 176.67 309.76 m 176.67 324.99 l S 184.64 297.95 m 184.64 313.70 l S 192.61 302.22 m 192.61 317.14 l S 200.58 261.85 m 200.58 279.97 l S 208.55 289.47 m 208.55 304.56 l S 216.52 288.73 m 216.52 303.53 l S 224.49 280.41 m 224.49 295.45 l S 232.46 288.37 m 232.46 302.29 l S 240.43 245.83 m 240.43 263.07 l S 248.40 253.92 m 248.40 269.92 l S 256.37 267.16 m 256.37 281.72 l S 264.34 256.50 m 264.34 271.56 l S 272.31 251.58 m 272.31 266.68 l S 280.28 214.69 m 280.28 233.05 l S 288.25 254.70 m 288.25 269.09 l S 296.22 247.83 m 296.22 262.62 l S 304.19 237.67 m 304.19 253.14 l S 312.16 236.61 m 312.16 252.10 l S 320.13 232.67 m 320.13 248.43 l S 328.10 217.35 m 328.10 234.56 l S 336.07 230.82 m 336.07 246.81 l S 344.04 213.92 m 344.04 231.58 l S 352.01 193.33 m 352.01 213.20 l S 359.98 145.60 m 359.98 171.87 l S 367.95 182.70 m 367.95 204.63 l S 375.92 179.22 m 375.92 202.15 l S 383.88 183.77 m 383.88 206.88 l S 391.85 154.03 m 391.85 182.23 l S 399.82 160.09 m 399.82 188.51 l S 407.79 149.06 m 407.79 180.73 l S 415.76 116.95 m 415.76 156.66 l S 423.73 111.76 m 423.73 155.58 l S BT 0.000 0.000 0.000 rg /F2 1 Tf 12.00 0.00 -0.00 12.00 340.96 130.85 Tm (Age 81) Tj /F2 1 Tf 12.00 0.00 -0.00 12.00 261.26 200.96 Tm (Age 71) Tj /F2 1 Tf 12.00 0.00 -0.00 12.00 181.57 248.28 Tm (Age 61) Tj /F2 1 Tf 12.00 0.00 -0.00 12.00 101.87 292.10 Tm (Age 51) Tj ET Q endstream endobj 732 0 obj << /Type /XObject /Subtype /Form /FormType 1 /PTEX.FileName (./fig-deaths1921-1940.pdf) /PTEX.PageNumber 1 /PTEX.InfoDict 770 0 R /BBox [0 0 468 468] /Resources << /ProcSet [ /PDF /Text ] /Font << /F1 771 0 R/F2 772 0 R/F3 773 0 R>> /ExtGState << >>>> /Length 6282 >> stream q Q q 59.04 73.44 378.72 335.52 re W n 0.000 0.000 0.000 RG 0.75 w [] 0 d 1 J 1 j 10.00 M BT /F1 1 Tf 1 Tr 7.48 0 0 7.48 70.10 164.14 Tm (l) Tj 0 Tr /F1 1 Tf 1 Tr 7.48 0 0 7.48 74.54 166.20 Tm (l) Tj 0 Tr /F1 1 Tf 1 Tr 7.48 0 0 7.48 78.98 164.41 Tm (l) Tj 0 Tr /F1 1 Tf 1 Tr 7.48 0 0 7.48 83.42 165.75 Tm (l) Tj 0 Tr /F1 1 Tf 1 Tr 7.48 0 0 7.48 87.86 161.13 Tm (l) Tj 0 Tr /F1 1 Tf 1 Tr 7.48 0 0 7.48 92.30 160.36 Tm (l) Tj 0 Tr /F1 1 Tf 1 Tr 7.48 0 0 7.48 96.74 160.11 Tm (l) Tj 0 Tr /F1 1 Tf 1 Tr 7.48 0 0 7.48 101.18 160.08 Tm (l) Tj 0 Tr /F1 1 Tf 1 Tr 7.48 0 0 7.48 105.62 163.97 Tm (l) Tj 0 Tr /F1 1 Tf 1 Tr 7.48 0 0 7.48 110.05 157.73 Tm (l) Tj 0 Tr /F1 1 Tf 1 Tr 7.48 0 0 7.48 114.49 161.26 Tm (l) Tj 0 Tr 1.000 0.000 0.000 rg 1.000 0.000 0.000 RG /F1 1 Tf 2 Tr 7.48 0 0 7.48 118.93 150.88 Tm (l) Tj 0 Tr 0.000 0.000 0.000 RG /F1 1 Tf 1 Tr 7.48 0 0 7.48 123.37 162.44 Tm (l) Tj 0 Tr /F1 1 Tf 1 Tr 7.48 0 0 7.48 127.81 159.00 Tm (l) Tj 0 Tr /F1 1 Tf 1 Tr 7.48 0 0 7.48 132.25 158.34 Tm (l) Tj 0 Tr /F1 1 Tf 1 Tr 7.48 0 0 7.48 136.69 173.81 Tm (l) Tj 0 Tr /F1 1 Tf 1 Tr 7.48 0 0 7.48 141.13 168.11 Tm (l) Tj 0 Tr /F1 1 Tf 1 Tr 7.48 0 0 7.48 145.56 168.35 Tm (l) Tj 0 Tr /F1 1 Tf 1 Tr 7.48 0 0 7.48 150.00 181.93 Tm (l) Tj 0 Tr /F1 1 Tf 1 Tr 7.48 0 0 7.48 154.44 175.15 Tm (l) Tj 0 Tr /F1 1 Tf 1 Tr 7.48 0 0 7.48 158.88 194.09 Tm (l) Tj 0 Tr 1.000 0.000 0.000 RG /F1 1 Tf 2 Tr 7.48 0 0 7.48 163.32 171.99 Tm (l) Tj 0 Tr 0.000 0.000 0.000 RG /F1 1 Tf 1 Tr 7.48 0 0 7.48 167.76 201.84 Tm (l) Tj 0 Tr /F1 1 Tf 1 Tr 7.48 0 0 7.48 172.20 191.00 Tm (l) Tj 0 Tr /F1 1 Tf 1 Tr 7.48 0 0 7.48 176.64 190.95 Tm (l) Tj 0 Tr /F1 1 Tf 1 Tr 7.48 0 0 7.48 181.08 214.42 Tm (l) Tj 0 Tr /F1 1 Tf 1 Tr 7.48 0 0 7.48 185.51 201.68 Tm (l) Tj 0 Tr /F1 1 Tf 1 Tr 7.48 0 0 7.48 189.95 209.12 Tm (l) Tj 0 Tr /F1 1 Tf 1 Tr 7.48 0 0 7.48 194.39 223.31 Tm (l) Tj 0 Tr /F1 1 Tf 1 Tr 7.48 0 0 7.48 198.83 224.41 Tm (l) Tj 0 Tr /F1 1 Tf 1 Tr 7.48 0 0 7.48 203.27 248.61 Tm (l) Tj 0 Tr 1.000 0.000 0.000 RG /F1 1 Tf 2 Tr 7.48 0 0 7.48 207.71 221.00 Tm (l) Tj 0 Tr 0.000 0.000 0.000 RG /F1 1 Tf 1 Tr 7.48 0 0 7.48 212.15 255.09 Tm (l) Tj 0 Tr /F1 1 Tf 1 Tr 7.48 0 0 7.48 216.59 248.91 Tm (l) Tj 0 Tr /F1 1 Tf 1 Tr 7.48 0 0 7.48 221.02 254.08 Tm (l) Tj 0 Tr /F1 1 Tf 1 Tr 7.48 0 0 7.48 225.46 265.68 Tm (l) Tj 0 Tr /F1 1 Tf 1 Tr 7.48 0 0 7.48 229.90 268.95 Tm (l) Tj 0 Tr /F1 1 Tf 1 Tr 7.48 0 0 7.48 234.34 265.89 Tm (l) Tj 0 Tr /F1 1 Tf 1 Tr 7.48 0 0 7.48 238.78 287.94 Tm (l) Tj 0 Tr /F1 1 Tf 1 Tr 7.48 0 0 7.48 243.22 283.84 Tm (l) Tj 0 Tr /F1 1 Tf 1 Tr 7.48 0 0 7.48 247.66 311.59 Tm (l) Tj 0 Tr 1.000 0.000 0.000 RG /F1 1 Tf 2 Tr 7.48 0 0 7.48 252.10 277.42 Tm (l) Tj 0 Tr 0.000 0.000 0.000 RG /F1 1 Tf 1 Tr 7.48 0 0 7.48 256.54 316.29 Tm (l) Tj 0 Tr /F1 1 Tf 1 Tr 7.48 0 0 7.48 260.97 326.30 Tm (l) Tj 0 Tr /F1 1 Tf 1 Tr 7.48 0 0 7.48 265.41 329.90 Tm (l) Tj 0 Tr /F1 1 Tf 1 Tr 7.48 0 0 7.48 269.85 356.85 Tm (l) Tj 0 Tr /F1 1 Tf 1 Tr 7.48 0 0 7.48 274.29 326.34 Tm (l) Tj 0 Tr /F1 1 Tf 1 Tr 7.48 0 0 7.48 278.73 342.85 Tm (l) Tj 0 Tr /F1 1 Tf 1 Tr 7.48 0 0 7.48 283.17 372.90 Tm (l) Tj 0 Tr /F1 1 Tf 1 Tr 7.48 0 0 7.48 287.61 363.43 Tm (l) Tj 0 Tr /F1 1 Tf 1 Tr 7.48 0 0 7.48 292.05 386.06 Tm (l) Tj 0 Tr 1.000 0.000 0.000 RG /F1 1 Tf 2 Tr 7.48 0 0 7.48 296.48 343.44 Tm (l) Tj 0 Tr 0.000 0.000 0.000 RG /F1 1 Tf 1 Tr 7.48 0 0 7.48 300.92 393.94 Tm (l) Tj 0 Tr /F1 1 Tf 1 Tr 7.48 0 0 7.48 305.36 387.99 Tm (l) Tj 0 Tr /F1 1 Tf 1 Tr 7.48 0 0 7.48 309.80 386.23 Tm (l) Tj 0 Tr /F1 1 Tf 1 Tr 7.48 0 0 7.48 314.24 385.71 Tm (l) Tj 0 Tr /F1 1 Tf 1 Tr 7.48 0 0 7.48 318.68 382.43 Tm (l) Tj 0 Tr /F1 1 Tf 1 Tr 7.48 0 0 7.48 323.12 355.25 Tm (l) Tj 0 Tr /F1 1 Tf 1 Tr 7.48 0 0 7.48 327.56 366.53 Tm (l) Tj 0 Tr /F1 1 Tf 1 Tr 7.48 0 0 7.48 332.00 339.33 Tm (l) Tj 0 Tr /F1 1 Tf 1 Tr 7.48 0 0 7.48 336.43 335.97 Tm (l) Tj 0 Tr 1.000 0.000 0.000 RG /F1 1 Tf 2 Tr 7.48 0 0 7.48 340.87 288.44 Tm (l) Tj 0 Tr 0.000 0.000 0.000 RG /F1 1 Tf 1 Tr 7.48 0 0 7.48 345.31 299.00 Tm (l) Tj 0 Tr /F1 1 Tf 1 Tr 7.48 0 0 7.48 349.75 279.67 Tm (l) Tj 0 Tr /F1 1 Tf 1 Tr 7.48 0 0 7.48 354.19 265.61 Tm (l) Tj 0 Tr /F1 1 Tf 1 Tr 7.48 0 0 7.48 358.63 240.83 Tm (l) Tj 0 Tr /F1 1 Tf 1 Tr 7.48 0 0 7.48 363.07 220.97 Tm (l) Tj 0 Tr /F1 1 Tf 1 Tr 7.48 0 0 7.48 367.51 198.96 Tm (l) Tj 0 Tr /F1 1 Tf 1 Tr 7.48 0 0 7.48 371.94 175.64 Tm (l) Tj 0 Tr /F1 1 Tf 1 Tr 7.48 0 0 7.48 376.38 157.71 Tm (l) Tj 0 Tr /F1 1 Tf 1 Tr 7.48 0 0 7.48 380.82 144.98 Tm (l) Tj 0 Tr /F1 1 Tf 1 Tr 7.48 0 0 7.48 385.26 123.93 Tm (l) Tj 0 Tr /F1 1 Tf 1 Tr 7.48 0 0 7.48 389.70 118.38 Tm (l) Tj 0 Tr /F1 1 Tf 1 Tr 7.48 0 0 7.48 394.14 108.48 Tm (l) Tj 0 Tr /F1 1 Tf 1 Tr 7.48 0 0 7.48 398.58 101.92 Tm (l) Tj 0 Tr /F1 1 Tf 1 Tr 7.48 0 0 7.48 403.02 96.13 Tm (l) Tj 0 Tr /F1 1 Tf 1 Tr 7.48 0 0 7.48 407.46 91.49 Tm (l) Tj 0 Tr /F1 1 Tf 1 Tr 7.48 0 0 7.48 411.89 89.19 Tm (l) Tj 0 Tr /F1 1 Tf 1 Tr 7.48 0 0 7.48 416.33 86.12 Tm (l) Tj 0 Tr /F1 1 Tf 1 Tr 7.48 0 0 7.48 420.77 83.27 Tm (l) Tj 0 Tr ET Q q 0.000 0.000 0.000 RG 0.75 w [] 0 d 1 J 1 j 10.00 M 73.07 73.44 m 428.17 73.44 l S 73.07 73.44 m 73.07 66.24 l S 161.84 73.44 m 161.84 66.24 l S 250.62 73.44 m 250.62 66.24 l S 339.40 73.44 m 339.40 66.24 l S 428.17 73.44 m 428.17 66.24 l S BT 0.000 0.000 0.000 rg /F2 1 Tf 12.00 0.00 -0.00 12.00 66.39 47.52 Tm (20) Tj /F2 1 Tf 12.00 0.00 -0.00 12.00 155.17 47.52 Tm (40) Tj /F2 1 Tf 12.00 0.00 -0.00 12.00 243.95 47.52 Tm (60) Tj /F2 1 Tf 12.00 0.00 -0.00 12.00 332.72 47.52 Tm (80) Tj /F2 1 Tf 12.00 0.00 -0.00 12.00 418.16 47.52 Tm (100) Tj ET 59.04 82.29 m 59.04 388.86 l S 59.04 82.29 m 51.84 82.29 l S 59.04 158.93 m 51.84 158.93 l S 59.04 235.58 m 51.84 235.58 l S 59.04 312.22 m 51.84 312.22 l S 59.04 388.86 m 51.84 388.86 l S BT /F2 1 Tf 0.00 12.00 -12.00 0.00 41.76 78.95 Tm (0) Tj /F2 1 Tf 0.00 12.00 -12.00 0.00 41.76 145.59 Tm (5000) Tj /F2 1 Tf 0.00 12.00 -12.00 0.00 41.76 218.90 Tm (10000) Tj /F2 1 Tf 0.00 12.00 -12.00 0.00 41.76 295.54 Tm (15000) Tj /F2 1 Tf 0.00 12.00 -12.00 0.00 41.76 372.18 Tm (20000) Tj ET 59.04 73.44 m 437.76 73.44 l 437.76 408.96 l 59.04 408.96 l 59.04 73.44 l S Q q BT 0.000 0.000 0.000 rg /F3 1 Tf 14.00 0.00 -0.00 14.00 93.98 433.45 Tm (Canada, males: Total deaths 1921-1940 by age) Tj /F2 1 Tf 12.00 0.00 -0.00 12.00 237.73 18.72 Tm (Age) Tj /F2 1 Tf 0.00 12.00 -12.00 0.00 12.96 208.90 Tm (Total deaths) Tj ET Q endstream endobj 781 0 obj << /Length 2703 /Filter /FlateDecode >> stream xÚÝZKܸ¾ûWôm5ð´Ì‡DI“Ý›ÍØ`Ùr° ¬FbOk­–zõ˜‡ùí©bQjQÍñØ›ì%O—ÈëÁbÕW”ÿróêÍÛŒo8³8››Ý&V¡Êä&QïyÑ~5‡ª@[ufȼ~ËJ£ÉP‚" ôŒÐG‚m}j‘Y9Àb\+Ñç¶^L^äB„±J@ž‘£q)gCå´+ç;D!ËÒéÕAw‡>¼ØJ&ƒ·¨Ê$™Ü“Žµ¾Är LôŽ‘¾ÓÎKD>TÞ¨¼,«¡º0{‚fÚ¥»ÎHªÞcÝ–g`Úlø’ÛP±¦djiJÈ!æ­%ß—ñ‚# 9‹&Ž÷ŒKÏ*iÈżÊkÇÒ6tv5ºåøžÅÌ''À/F^µE­ºË D¾X%öx)“tÞÆGÚQÅò´HÌ=ªÄa,ÅÚöôYÛÏ…p4S 1LL²6ÛH&¡[D"Ã… c!Ë·"ãÁ·>Ñ°ýjöC©—”… =˜Å´6å®Î­ _ÜE@u~¤GO&ÿÄSøŸRà7ƨS´du·ˆôjš0¶g©Òœ¼+˳4Œ’ìÿ)úa=f‰L)ä°mye1úSqéfÁôUçÄ•Ç•Oœãs8+ðÒ‚é¹}ÁúŸº3újwŠgÜ™}‘;¿À›^q¿ÏD~rm)b–råK#’…L&Ë &ŸË` ñ\÷Cu˜àD¢,vJbH;Ýé¦Ð4n !Pýhð °Ì( ßë,ãðt|R×OÄØþ6t9¦˜]Û™,’w%¦¸TǼˆ]³EÑŽÍ@k˜Dƒ‡±**ØðT· Æ*ŠL:ä`à™%e {MuõѺºÚ·mI“è'Þ…—Äò³¾ÕÍoàØ¿0#VuyIãp 3$ ~؇ÆÇ<âDcŠ—„×õˆfz¥DYyâCJénWÞ3'd6ñè\¸è¶ uŠ3»YÆšg¢ßLȈCÑV.¼C##lEõž1Qì…# ðÔãàf¯{}>K#`5Q‡œ\mU¨ Frö¡Ûz4Ù¬¼h )Ò>jp Ý ,+gh­wØÒ‡© #>Ÿ‘»æà­ópg¸a·‘%ÌXên)–Á»1w6¤ ÿ^Bí<ÕÊQ#†D4kGmÀóÉ ƒ¨kõÓöN7º#ãQœ“ü ¡ÖVŸò©€a:±@ôîžÁôð4ûvÍ6ìM/ñÐo×âÎ5ÞL[¢ØçpúíÛæÂ/:#ÃxFdàíÔuÌmk¢‚½… wu{;õfæpNenÔòÚNÓ¡>qæ.”™—tqh·€»Ý™€jnC>7Œ&`¯`›96Íݱôº•\œ»ÀŸô€ ˆÐÇø‹‡©ëí ôjÆÂ>öíAUæCNÔ®kDYlF‹X( 0RêÔÝàd;ZbN7N—‚3DZ;¶½¶*M‰æ,ƒ=´Fo_“&Ó$LN`¶i¡QÝîº È0Ú¥#UsâFã.§ÄBÁ„1U¨Ã íê È2ñ¸Úò©Ω&QTxúIÝ>ôœ goí¢Ÿt×þÉÞ L·³Ûڎͧ±wU'W"5»r…2ÊH½ˆeU¯±c„ˆÙ¥3õb—Šì%9òe9òl­ìaq²’#Òç>cÞ“rvóC¤,埃H˜§Nƒ‡ž²•) 1ìZ wÏ„ Ò?Cé¦ÑŸo·ô˯8)º¾.±ìO.»&€¼uÖzc—b3Ró¯Õë!ìµ.q~Æ«\Þü^—á®Â+¡¥uÕøþ ¢Là#¼‘X g“©ƒt%½{W}øàŠƒú‹+X/ü›~®É öý×Ö¶åEú^w·-–W|øŽ~Þ~ÿÓÏ×/hm.¹péOù±?@IAºÇ ³'¢&M­œRß›d‚ËNÿlL1˜©P¦pšXFKH˜*Ž(»Å€ÍV0!/3E©'!‡2îÒ€®•~ ðÖ0¯š~·É«Ó=@J;JE—âô¡/ó†}ñïãÔW¢/\N禧jì Á™Ó`©(ŒÙïCX°¤aYh…ÚÔÕÍ竤ˆÈ*$'L #Tw‹aš0 Æ—õǟKs,ƒjÄUb¸!•ÛÐ.r8e™ð©)‘<øû8¤LD[ŠXhÍîL°Â,`î®-ÇÂZdM‹&”CöØ·©8°ïc)E¾ìfÕØvÊýyɇönìòW1 €ž*R&ôá¸ÐàÑINW&m‰6 §! t¢*¹üˆFø5ц±¡ò *œfÁ-^qß(¾Fwù5«öÐP„ikËݱ‹®qÙÌ^{òœ®ª Zb} ‚Î ü6‚Ôž0³ @KÅp‚Ž(Ò±UGËæë.xÜÙÖÝ®/cü.À½¸>TBzoK °ÄLçG’¬Y]o8Þ)ÞÜÛÚFYæ$†dê$¬,t^"\š-C7sm–@mgÂ^klâpÀªÔ¢Ì¨‘…>„1¡c“:á¹:¡ç%ó[šWmÒ€`åî–e*R½eP rÍÝ°Ô¯Y) ¸ÞÁg{£ª”ÅŽ¡µ òÊêá ÄÕvüCç]ßZ\ %sKbã\®^ýç\J ’ÝÁ+QT–ù¯(ÖsnP=]¶™ÏKÜcÏÀ=P(IøWÃ=ÈLvY{ao€ZÌ=ýú¦çt#4_](峚N/¯ó¾ªíõM ­”Ž±ê÷Ú{a´å"…„žÙ{£øto$ü÷FëïlÎ×1çÿ¼tk4]“×›éë^SŽÅà|ûë«CUçÝò»•ü¢šöò6ï§ l›•/?óépñ%pþl@U@NR€H·1T2&m_cî¦_]ß¼úFhc endstream endobj 658 0 obj << /Type /ObjStm /N 100 /First 898 /Length 2488 /Filter /FlateDecode >> stream xÚíZYo9~ׯàcü°lÞEŒœxs³“ Éî̎ᇶÔvzGV;­V’ù÷ûU«%ë°YRœE°€²y«Šõ‹GpA(‚ÚX¤^Ïi6r9 ø; ï5Ò$‚3"дOðIÄ‘WB+ãz!8¡5¾‚'PTU‡~„§¹3ó½È ’G½¢å*/tâA( £U@CÎxÕ I 062N€0G ®Ø+&Œƒyi LGÆtŽÑò$`oqÚüd4ª@ñ¬õÌû Nç<«‹¼)«ÑiÞâÉéOFïÎþ6q„«Áƺ÷e3DÍ[ñ¢Îo>”ý±x=in&ÍQ+÷`Ò/j®µÒHuÔUqÑѲxzÙ»ÉEÓ~r¡îe¿æ×\gzÙÓ|\´³”½,†ŸŠ¦ìç½ìï£~5(GW­?di·'iï$ù·§Õp°]w7Ý×Ãò㤸›ôYöêÙ3î5@i;Ëç ÃÍzL Ïûg¿•£“Ѹ¼­>-// L>[ËÜ]v]Ž&c +û8©šbX\6X\6¨0eãq™]Õù§"Ëû“¦ÈúeÝŸ\_‹/YSEv÷ëj”]ÔÚ KÞïî²A‰!Æå8“0«Aq™Õ;ëÃ&‡Ã|^øa2ºÊëÉõ0Ÿ4YuUŠ?³~ÎôÆ7y¿X1þÖD4úK9úÒVõ ¨[#×çÙËìUö PÏ9ÃX;3ÚJ^RŒ’=¦qÀwˆS,œ´x|'²ÕûJd§âI¿l ù"¿¨Sƪ,½„Þ{ÂÍã›4ûšNÎì ­ÕîN&ìì¸áÄ"œåÆ-]ç­r—õ¹ü-Gx;ɘÖe¤=WœòÇ«C:+í%v8XB >Á°…®]ÒàpÓY)©oÙˆ˜=ì°v`£µ‡øP{Ød+ˆZÞ=ðZ2¯E@íjRi6éa°I~6éÁfIi; ¼¤°¯À´§À‹2.H¿£Œ¤ÃªŒ45“‡Êxë'øx„YâÓ‘ijºÔv©ëRߥ¡K©Kc—N>™¦=mé,é1 Î8ˆ Ö’dwÚùÍÀÿ¥(žåu“Ìá\ƒr£[^'Ùà¿/cƒ(œ6Òj/°ù–ËŠ3Vê{¶¿•ÃëdÇG‹H)I•‡0;°P§µ4Ö=ØÚ´ñZdžVV>Úû'ÕR‡ãî@†[6L”^ÇïÀ†5°}ËHÂÆÎä«Œ<­«Éi1šü«¨¯ÕÍU'‡õzÎGÉÄïÇÑI††,oa·o‹Ñøe~Q¨Ú®d•u·¼8'“7Å »íEÅ †`0Å ü‹0‰ ÙÜÉÊË˪j0X!í9ù<ײNàr± DPÚ+¸=ýø㈷¨nÎW;B½C8øÒbÛârÛéÑy„Ãé+mIc&“Ðœ&³wä¶e„»¼ÑÜ{[HÖ¯…BÖî é.äé$¥ŽEêö½ÔÉIðç‡\!" f_¦Ä— @RâCJºc—åÕ¤.¤V‡º¶±=Úð&Èö†Œž,—‹ ©`…¾¬ÐùèVóJ®°‚߯ }H·A/ò'†÷lâR@€É‰ñ~mÐûºÄÑ-ß¹b¹ p2V¶7ÀàÊ[û÷ë7ò±É½é¥õÐôZ“ô8)/-P« ʇ¶,­(ÙÉñq;BvÒç ¢ì]öÏ·¯øÿäCÓÜü”eŸ?–×UÝäòùKVõUöfr1,ûÙëOEý©,>Ë›7G˾кí}ábÛ™ÔE°þà ü¶gÝ«'Ú»x1·¶Kç;×Ý7­ëWt*hmø‡tå^mù¾n¥á⥜‘®½”Û|‡× {£õG~syZŽ®.òf<¿óÒë·a 7ek4îºÀãÓØV=·7~;‘\½À[¥k·¥ûƒÜ³­Þ!.`ËmÏ´ÔV#!F |üÈ@\¿Òüb¯ƒ0òf b^bX/t0ØA| b‹ W!¦ÿG æã·Ø ÝÿCl¯·‡Øb[ Så m/ˆ!ôìîkv…ØìÅÉ"ÄÒC FÁm1,ŽÖn@Õ¬nˇ&3ÀíöÐDoâ«ð»§oñXåNº?òë‘Mö„p‡ŒºÛžæu?=‘û6ötÝÙžØ wã¶rÙ‹mµrÒc'éÛøÞ{ó¹(ûFm¼Œ^oµ—Y¹J[qöü0´ÛÍàËòëн\\wý´“ë_8†éž×uÇ/ÝSš>ñ9ÔM’ŽûDÁð+MIËOªm”ÞÜs‚÷®¸(b:àÙ³wÒðà#¤æW¾ßž‘•C›¤v$¼ 0Ql‚‘Âò‹f¹á:«ø8i=«$9?Ëü/$U endstream endobj 790 0 obj << /Length 1970 /Filter /FlateDecode >> stream xÚÍX[oÜÊ ~ϯ0òR¹Í*£»ä^€´M€ÁiѺèC ³Ò¬wp$ŽF²½-Úß^rHi%G¶cà<ôi¨É!9$õQ¿¿~õöC.. ¿HÃôâúp‘¤~ZDYø" .®«‹ÏÞï.wI˜x÷´üfGkpqùõúOâb~‘$+öÓš]Ýw_D"v+]oY•€WÑ¿¢%¤å—Ï '$ü¸!V ¾UªB Á7¼éšWÞªÊ?èÁ®­¯µPþ™£¦'y5‘º]Äëìç¿·Ô€óaº2ãógýõëÚ;v]¯¬ý‡ì[ÝÞX<ë¦mpá˜ÿ—–÷µu`ïßLf٨߲}Ë;xĶ(}±ÚÕ¾yRá­ê÷ƪ•Ø‡wŸþöE‹8ŸûŸ'.sûZ¡©½[wýc;]¤•]WŸˆšbÍ.TêV˶TÓé¼/­yû!é¢l¢Àò¬€­¨ +Äå.ýžðƒ$/XIà§qT8£w‰Fk§B‰Ò,ñ"^c:,gAŠ€Æ‡}l1u o8jKÞºæ‡Y䙶>Ñ‹0Ã5÷ÌaQD`=;¢”uÍZÃ,›DE>t»;’oÃL°. ’‰ç(/CáÝ^Pَ̤e Y|ëÉ°Ú”rPY'Ï{5‘¼×ÍØПŸ¯õ—ðT­ÆTW¿8ÇÚs‘ŠÐ/²tlÍ3¯WÔXJô˜E´–Ò*KäÌ !퇙óö®TÖ#±åÞjUÏ.Àë;ÕObcyœ4ɨvlT¯É;xìzÙÚØÉ$;Ö /¯H=ÙVtÂlÈ!ÂÁY²Q²¾1½Ž«$÷î005'…'÷ Ì´ÎFx皇F?nÉ-dÙºò8ò ù:øû§Owƾˆ‹‰‰œÀ’ñá:¢Ä{טé r pÙÿ<çÃ/&ÓÀE,Ê0 <=Цf7™\©–_p`EqGOz5íõJq.¤™dɺcTÐt[¢†HxÖÔã Mkéctá¤'ðªSå@1›ò™öu5ºëD£»Ê( LÅŸF0‡UªŸfN¨0·R›@w£˜úÓtù(d-¸«% lüKõ,ü%tq„ã(°°UKèÄÉöDÂå´Åæ…,„› º—|tWÀì×üÌZÉA²8·K |µWª%ꀗH×ÛP‘n0 Hš‘•/l"Bî Z2Å×ÉgœIqx×Äæ¸ä6ÄÒšvg:8c‹Ô=6‚àz‰¤ÔâtIC¯4=Þ®q‡7îjÒ)¦@ì~Ý•#ƒ›œ@±B ¨Y%TÒ̓K¨£.YìèjˆE ¨¥^Ç„zç ‚Júf#7fɆ£Îøq½‰Ùp Ùp ZæªT§Z¨©’™%sBÊaZ®ú|Ùè®éÜî7Cêq<1¸OòÙÕD°«¨ÿÀ³«ðàŠ¿ÄÎÞ¿c+KÙ²”âý‰ß˜¦“=76røj£¹í‚Xs?ŠÃ5"t— ˜g Â^¸Â}>V8§DÞñû™ØÁõ.äžDHOÁÙÙø Ì"ÖâpÉW[)Ä®¶mÆZ®lør?Œð É£]j¨9K.%™ŸDñÚ/<úc;¨¾TÝ@…äð»á놿!’FBl{î;W ¡ð‹0Xß‘ð!'´í¼n`ˆÒ4†§BD§ }²²ÜÁꥋdzÅP;ˆ‹tëÿ¢do Ï!åQ屮%%îè'dÿ:ªÃƉ3 Šl£¤ga£dûg?Ÿ­ —'|²}Ôãùÿ\éþóe§x.1] M¨Á I“ôzž ÜIdÉrüxYŽíÊ#´2gÏjØ–Õ°+xƒu´ f6¢Ýÿ/÷3}8±tæA/tW³Kg\¾ "Ÿæè1xžsòÌ“~þ`Òÿ5þÞÈøNàm¯$´1¹§T™xZ¹“>pÿ7ô¬¾Uªš~Bèóoj³'|ž/ÿGp4Óкg7ù[ýåcû™ãO=g¿)VØù@:˜|7ç£+QúÎEɪ~çg´$sR@²dQî~ôúAŠ€Ä 6(ØÄí1ŒÊ &3¾93õó,{ñ‘á÷iÜ°vÍ0W1cþfNæºl]ôn.á‚—sú›õØÐè›#G»SýQv\÷~Ua¤{­hº ?M3øÈBˆarö% ½zýêô¹ý endstream endobj 794 0 obj << /Length 640 /Filter /FlateDecode >> stream xÚ½T;oÛ0Þý+4JhÈð(’ ´@Ú$@Š¢èÃ[ÓA–)[€,9’œ8K{"Z­ã  <Þó»Oǃ€áE¥ $*â8Èד» U\098ˆƒ |˜Wœß¬A—Íä3~{ñ9ÉAÒ·³ÉùušjÅU0+’B¢‘h*A³Eð-¼.—ÛÖD„K‚˜F$΢”…J³tÛ9i6}¹Î*§¸eŒ÷“{ï”í}פ)ˆ‰@†»MS›º/³ÊûæÛöÞLô}öñ±|D§T bsÊ„Ç÷&"’ËpY¯o™dîöÃeÝõY›Îš®v{€3½rÇUpæd~ c\Û;ñµ;rüñÂÛŸÂ(Ó¹\ØòO¨–/æ îM;o:3ªq}ñáë•tÍSÝcŽy uyÞeU5=VÀ3Q4íz[e£üÿ“‹Œ3ª9Œá&l¸¾H 2á3óñ„4¦(ʼÄqê1À4MãÔ×WÎͽ©{ÓæfÓ  *ôÐ1g¿¨ÁàÒ?s¹²T¨qÓ8¯R‹qÅXH7ƒfèG%Tk1:úKèiD»ꆟèFqš¨ß Œ1C,6ˆ3(ªc®#A×÷í< {ž@I™JýDø‘¸4÷¥pÜZ*FÖ9’ΓTÂű§óÉdm×Ô~¤W%éî¶YkSé…ð/¦+[»­÷¢À0‘ ¤¥_fñóÏùaUæ+·[S•Ù¼zt·ÇÒT‹ý"]™ýv•lY5sÜÞÍ/_7ÐV±1maòþ`—ñS®EÖgÔ=WÐT D˘‚J2™Zãäj6ù v endstream endobj 787 0 obj << /Type /XObject /Subtype /Form /FormType 1 /PTEX.FileName (./gnmOverview-doubleExp2.pdf) /PTEX.PageNumber 1 /PTEX.InfoDict 797 0 R /BBox [0 0 432 432] /Resources << /ProcSet [ /PDF /Text ] /Font << /F1 798 0 R/F2 799 0 R/F3 800 0 R>> /ExtGState << >>/ColorSpace << /sRGB 801 0 R >>>> /Length 2488 /Filter /FlateDecode >> stream xœµZËŽ]·Ü߯àrf!šd󹵑Î …á…¢H€„‘diäØùûTuŸ{xFlYh¤o]>ºÙÕݜ螻èÞº—ïÝGW†Ù5ñ9;ÉÉ·äÒ¾$÷é•û‡{ùæñïùÖ}ww >„àŽ?ï¾ûþ·÷Ëå‡]pÿºD÷Þ^"?àþzùöþòÍŸ#¦»ÍŸ\ó¹ã“ÁþQ»oÍ Öš»çnnÝý[ŒÞºüéþ÷Ø-úѱèâGZ`Áá^ϳ»—á$%îó4»G_ÀŽqiß=ûœ„Evç’ÓÀáåólÐRt© ÿ:ÏÎ>ƒ£²Àn¾Â;kóRϳc¾ èÁDzBÏ\t*²8{ó)»”« F‹¸P™ôàÇÂÉÅ(¾—¤ø´D¯¾ƒž†ï w%&Dƒ zöu¬Ðál ÇÕÙ«Ï tÜ÷¥Ù‡¯¤†ªótI¾w—BõeÅm¤ú@zòeÅp‚È:\Ý×¾@‡”zf ] ßHôÝz÷à X+ô’|$=¯®/ ôÈ ³@ïÔ•(+GW£ozá½] gƉØÒÚ•.Ç z ó§C˜tê%»·ìKÑ=ôõƔ$ÖD…>Oï±ök>w‹ thÕÊÞ;Ä­™ó­ì}º[„HÅ•h3Ä7Ò£+n3 nÃîíB°JèH:¾e!§J9,¢MÎKé #t òÆ‚Û¤uÂpŽ :çéè-^/Ø=A ¢(Z ŸIG¼^pÚ„,¸"ÚH^JëR*¾gUÊŸgVHk‹Ç´©¨Ì¯¤…L©2éÞ³@‡2WÍQV"m‚»uÒóÚìèÐ4ÁZÚ;.z"]µê<afËeÅp%3©‰±/ê„Ýmb£ÎÓ!Ðahf¼R$ˆ› Ú )]²;¤µDMëûÊÑAØéi)·IMb#Ê×¼Do>&-iÒÊÑA Ñw)Ò¢t-¢õØRœï•a&B-ú}°ìŽ¡,%fièMe-ÚVŽ-¤ËÚ…e_Ñ&h—è4]È#}-\Î ¥3Ëø•½ ý¥»¡2}ž«"`ws ìÂ~ÏhKµ¿ Â5e‡¶µ¨`ù´ mo”¥-©³ßƒò±¬Ì-(½À–¥ÚUàŠ´TDt9({%< d9 {e²²od°l¤ò+gžµò@œé+s—ÀNØK¾I…FYH'Š A@´RòJ >ƒ]Y½-°…]°ÓBT—Zée½0 >϶2»¯5Ybœ”½tæ­ú<\_iƒ=°ãŠôÄB?Wð[ÅÀNúWÛØß»ÿÛËEíLì]äfÓaðÁÝ}1l^l”µLÍûðãC³ˆ}Üàg“×ð:¾Á9.Høáæûê ÆËÝW?Çy¶z8/¿r8/ßãèÓvô*aÿa¸j3¡f3]àáïVûMj„¤`±“›Î±“IÎ'ȬLr=A¦ §ã–û2.J8îYÌ•mpLgmeªÝEmõåxÔbÙàa¼&^ä}Üàgû{È>¾ÁÃxLpöqƒs|[Ðuü°¾ãî[~v8 È¡&~í`u_¾vx¿M§õ,Ä¡ô:MOHøËØèçg×zÏMZ˜]b¦=Œ^žš~»¥—ô£[Œ¹Bi‰×åáj‡§ðúá; xäZËæ§ù ?U\´"H5³žô7÷·®wóË-ÐÝ|p?ÿóÙ‡Ÿ>¿y÷âÁ½~óùÑ}þà^àß=ûðúÙ«[XÉÝüúÓ‡÷¯Þ~óâáѽüùÓ-ÎÜ¿ºýÑÝ?ÿƒW55mæõk¥ñë‰#zD)…kÊûŸ+ùÿÿàÝ´¸á›4l÷îÒ2åI¤2ƒ¸4í«3-†`>\ ^m‰ƒ_ ¨}o¾ #ljkGäCõ %\$Ô>3ëÉ „÷5Q9ïZéúÒ W T5)ÖVmB˜@òõ°R©¼cÄÚLMH“(®ÌÅRÓkÅö•Í—„ùr}g$Ö†dÂN‘úæ´)ékûÃU  ¬ß'ÅÂRfx'Ì$ÙU-º¾¬ž™P&ý~p楎Œ«ÄÖGc‘¤ßÏ0Õ´¡õû‹uºp\’wnU̲€ØJyF'ý<îLkÚ®]qa6ʆrÅZ=ofNúº¢-ߤX+\š5Uű)[˦ýÛ&.\(Û¦¢ß/Z)27°õe}vccSÏ—Ž×‰Í@Üé÷l=ª?²—ˆÜ†½Ä¨û…Xà~³9Øu?E« íöéú-£¢[gÙŽÄ~\R~ÕZ$Æl÷³Ü´á¦þÉXmÚA³ý#î][bA×Ó5§çµÒë©¿01L¶uý]óîºùï1ŸJ!R¢ûGåÉ·O˜-ëü°73ƒ–ÄC_'CV¾°·R ëúØ,áû!Ü@l\ßdغ©\û|áÃýQ{°C!†Õ™íð aÄ>Ïoa÷D¯û–pK.ЪvG¸xîÒ™n^X‰ýíf£w Pw.ü•†Áþ„^¿Qá6ª:0Š¹)ËÓªP½LŠ0 !vm<}@ÑEÂƈA¸9C¿ŠwOa1Xy”€IW…›‹“i;OØ*˜uU0/Î0Úè YPßkTg!ƒàÃgný0‚JWu‘–|ãBk¿Ûęؙ”˜tD«‚¯Ò±Á«tìФcB{2Ýa³Œiƒ&îO”U:v¸IÇ«t°½òíؤã€ínb•Ž¶—²›t°½eM¬Ò1±IÇÛóÑÄ*›t°JÇ›tìx“Ž‰M:&6éØñ&›tLlÒ±ãM:&6é˜Ø¤cÇ›tLlÒ1ñ&W|•ŽoÒ±ãM:vlÒqÅWé˜Ø¤cb“ŽoÒ1±IÇÄ&;Þ¤cb“Ž‰M:v¼IÇÄ&›tìx“Ž‰M:&6éØñ&›tLlÒ±ãM:&6é˜Ø¤cÇ›tLlÒ1±IÇŽ7é˜Ø¤cb“Ž‰M:v¼IÇÄ&›t\ñU:&.‡84¥cÇ›tLlÒ1q‹6é8àl¿`±c‰M=8?‰G›~LlrÀª ܞģMCXžÄ£ME&69`Õ‘V!9`U’‰MJ¸>‰G›˜LljrÀ*'<žÄ#”~—ÿi#þà endstream endobj 803 0 obj << /Alternate /DeviceRGB /N 3 /Length 2596 /Filter /FlateDecode >> stream xœ–wTSهϽ7½P’Š”ÐkhRH ½H‘.*1 JÀ"6DTpDQ‘¦2(à€£C‘±"Š…Q±ëDÔqp–Id­ß¼yïÍ›ß÷~kŸ½ÏÝgï}ÖºüƒÂLX € ¡Xáçň‹g` ðlàp³³BøF™|ØŒl™ø½º ùû*Ó?ŒÁÿŸ”¹Y"1P˜ŒçòøÙ\É8=Wœ%·Oɘ¶4MÎ0JÎ"Y‚2V“sò,[|ö™e9ó2„<ËsÎâeðäÜ'ã9¾Œ‘`çø¹2¾&cƒtI†@Æoä±|N6(’Ü.æsSdl-c’(2‚-ãyàHÉ_ðÒ/XÌÏËÅÎÌZ.$§ˆ&\S†“‹áÏÏMç‹ÅÌ07#â1Ø™YárfÏüYym²";Ø8980m-m¾(Ô]ü›’÷v–^„îDøÃöW~™ °¦eµÙú‡mi]ëP»ý‡Í`/Š²¾u}qº|^RÄâ,g+«ÜÜ\KŸk)/èïúŸC_|ÏR¾Ýïåaxó“8’t1C^7nfz¦DÄÈÎâpù 柇øþuü$¾ˆ/”ED˦L L–µ[Ȉ™B†@øŸšøÃþ¤Ù¹–‰ÚøЖX¥!@~(* {d+Ðï} ÆGùÍ‹Ñ™˜ûÏ‚þ}W¸LþÈ$ŽcGD2¸QÎìšüZ4 E@ê@èÀ¶À¸àA(ˆq`1à‚D €µ ”‚­`'¨u 4ƒ6ptcà48.Ë`ÜR0ž€)ð Ì@„…ÈR‡t CȲ…XäCP”%CBH@ë R¨ª†ê¡fè[è(tº C· Qhúz#0 ¦ÁZ°l³`O8Ž„ÁÉð28.‚·À•p|î„O×àX ?§€:¢‹0ÂFB‘x$ !«¤i@Ú¤¹ŠH‘§È[EE1PL” Ê…⢖¡V¡6£ªQP¨>ÔUÔ(j õMFk¢ÍÑÎèt,:‹.FW ›Ðè³èô8úƒ¡cŒ1ŽL&³³³ÓŽ9…ÆŒa¦±X¬:ÖëŠ År°bl1¶ {{{;Ž}ƒ#âtp¶8_\¡8áú"ãEy‹.,ÖXœ¾øøÅ%œ%Gщ1‰-‰ï9¡œÎôÒ€¥µK§¸lî.îžoo’ïÊ/çO$¹&•'=JvMÞž<™âžR‘òTÀT ž§ú§Ö¥¾N MÛŸö)=&½=—‘˜qTH¦ û2µ3ó2‡³Ì³Š³¤Ëœ—í\6% 5eCÙ‹²»Å4ÙÏÔ€ÄD²^2šã–S“ó&7:÷Hžrž0o`¹ÙòMË'ò}ó¿^ZÁ]Ñ[ [°¶`t¥çÊúUЪ¥«zWë¯.Z=¾Æo͵„µik(´.,/|¹.f]O‘VÑš¢±õ~ë[‹ŠEÅ76¸l¨ÛˆÚ(Ø8¸iMKx%K­K+Jßoæn¾ø•ÍW•_}Ú’´e°Ì¡lÏVÌVáÖëÛÜ·(W.Ï/Û²½scGÉŽ—;—ì¼PaWQ·‹°K²KZ\Ù]ePµµê}uJõHWM{­fí¦Ú×»y»¯ìñØÓV§UWZ÷n¯`ïÍz¿úΣ†Š}˜}9û6F7öÍúº¹I£©´éÃ~á~éˆ}ÍŽÍÍ-š-e­p«¤uò`ÂÁËßxÓÝÆl«o§·—‡$‡›øíõÃA‡{°Ž´}gø]mµ£¤ê\Þ9Õ•Ò%íŽë>x´·Ç¥§ã{Ëï÷Ó=Vs\åx٠‰¢ŸN柜>•uêééäÓc½Kz=s­/¼oðlÐÙóç|Ïé÷ì?yÞõü± ÎŽ^d]ìºäp©sÀ~ ãû:;‡‡º/;]îž7|âŠû•ÓW½¯ž»píÒÈü‘áëQ×oÞH¸!½É»ùèVú­ç·snÏÜYs}·äžÒ½Šûš÷~4ý±]ê =>ê=:ð`Áƒ;cܱ'?eÿô~¼è!ùaÅ„ÎDó#ÛGÇ&}'/?^øxüIÖ“™§Å?+ÿ\ûÌäÙw¿xü20;5þ\ôüÓ¯›_¨¿ØÿÒîeïtØôýW¯f^—¼Qsà-ëmÿ»˜w3¹ï±ï+?˜~èùôñOŸ~÷„óû endstream endobj 806 0 obj << /Length 2120 /Filter /FlateDecode >> stream xÚÍYKã6¾÷¯Ðmm fDJ¤D`/ÉÌtv40‹îMI€eK´ÍŒ=fº÷×oQõ²Ü– ÈɲÄGU±ê«¯Š?>Ý}ϱƒ}äùŒ8O[‡2ĸç #—aç)v~[ý°Þ`×uWÿ)e±ö‚Õ&‘kLW_×Ø]Éd½!Ô]më,ªTž•ë?ž~þþ>ôŽ8#L¯é:BaÌ,÷ë:„y0ÍóV‰*+ó´—…üN?²Õ6/Ì;)ÊWóTÈ-|Ï";D$‰ùoÍoµoV$«Z IÜFHÒI0¨²ù«²~®‘š‚Ô˜ÑFj¸`¤o¤ÞeéŒjÄEØ옃ˆ>‹Dëçú«{½µUe[[‰ã<ªS™UBKb^•RŽ${qŒ ƈS6ZÞ|/“ƒy:À»éywžî¾ÜaÇupw”˜ p'Jï~ûÃubøö³ã¢€ηfdêø>G0Âuçñîßw?¶nÑ«Ébvæ‘æ!íáß«ªRÙSDH8Áçá›"„î@„ÐC®: ~CÂÞ:ìRD)µ‡ò»ë’Êȹ“™,D¢þ'có"˳DeRæoªU*O«„ˆ€úª1gIÆ zdÉǃŒ”–82îqRˆOЛXÆqâQÞ+±;oÒ¨¢jýp«Q@DUÞÚ0Vf>gƒù”#Ö‡Éï.öf6¡ˆÝí•[µëŠ]ž‰ÖN2‘:†,Ú ÕÚà#ð6vZ?¾¦éeŠ•0CV…ŠÌ_•Uà;Q´ð¢Až‘æ¥iwIó”òˤùoS’|§"‘üã„D×Hòа:/J©ýrûjöæ'ÒSEVYiìG³X8XŒC2i\±YËݘ\í0 Ö-È2V—ÍhFˆ¨79ñ÷3®Slìj]ê1Áæé¼ÆC‘0%ÈïU>TÌÚ1iÓ:³ªr1'†‡:¹ò E×Q5öèÎÖåE}¢pž-ÞRŽ@Ê£Vî_yšï®ÓöªÔ!QP{Bp"‚¾©jߦW½#¤‹¼n g`aè<®u’QµÀªEøfVõÂ…0udÕ/‡ íiVc °àË!Ï„Nƾ#¦ázÞƒXˆÜàVÄ!¿Ò‰}̾^«k/!F¢[¨‰=G'—ë2 žáXÏo2Žeñ\ÙµêjÞ¬¢9´^©yÿ¥¥Ú$곡üjŸçñ0ç¤*y=ÉU¨Ç€5†· M”Päu´IVûàî&½¹ í{ÃÒÖ);ö`æ#Ÿ…—xð±jÆ Ò‹ýÏÍyËòˆ®øœ· T8ü›X%X`î¢à(ô¸Ö&; {!¶±ª‡yCÿ7«Ÿf1*7JÙÙæ.Ÿ"êwë}ÈJQRÇMsjnój/Z ¿ÁI˜'¦iîbpcêçWÙ|xÍBÍAEŸßåKz¹L.ò Ë“xwt祀µßåÈ^V œRcnOäÚ‰c ^®Z6 L±(Ú²(º¦ÄU„Þ#Æn‰  .7¼h/£Ï%Ÿ¹€×êñŽƒÞ…¥Þ­‰ò®3]ȉºÖ†é³Ê„½Õ8§?aˆr|9a2+Ûb]ùë¼}"P eȤGUÊ õô±wùpж_b’9[`äùûÂlhx@{¨O®i@C‚\¨ÝÞDh‚ƒFà9@uܶÃùÈBáÀBÜ7=ŠpÔÍäãn&7ÝL JîXö7Yw€0e7!ÝØ'Š¢ñÞZ¦O} rÁugÄô×'Å]Ó¸È4ÍšzÞ4(úõð°Ó5d%]ó·Á.eº$O«ò™võlß„b!¹EW„@ºC{W »×/*Q@ç—ö@ôU§NÈ7è€S!Ï #ÊŸ–U÷²° GX6nØÃ(æ# ÁåÜõ$îœ-™Ý‘ Ÿv¼§»E—]Z¿Ì¯ ¿‡ {Ö§ÄòšAÚI7?DxZ‹É—Äö;À„]nÊ‚…ôùÐeP{‘!*a+N@ȶÃój‹›M; °¥ìîÙ"EðÚî=YÈÖQ^Ÿº~;Aš[ÎûFŽå‰ ë= oGz],LÕe¯ÜL-¤‰ž¥úÚˆ!?˜ öÃ'µ¤]üCFÞ|’Y‘Û½¥¬ã\ K&ûiÒ6.¤H6m¿ª¶Ý4H…zYlÑ/Åcž|]`Ë•©´N7‰Ìvöv£Ì“ú¸òÇ©µ|-+iÌx4ußEk#œr *ÿ'žÉ1 endstream endobj 810 0 obj << /Length 3069 /Filter /FlateDecode >> stream xÚY[wœÈ~ß_¡§„9Çbé 7Ýl¯-yI›MNœ4Ó±f`Œwµ¿>U]Õ ƒ˜D›'š¾Ñ]—¯¾*Îï¿ûþ­'BÇJyrÿx’šØXu’'FœÜ¯NþݺG×.D¹zéºÅ¿î?|ÿ6W'6¶F\”œœJKchþÙºu]_¾YœÊ4‰Îbz~IÒD&‰„§ˆy“|´Ifc!ìå÷¸(z·ÐI´nÚrYT´ÅeѼi]TÏ]9w‘ä±2:lä?[¯h™[ÅôùSe’è“[È4ú•†þ±"‰šöë_`Z?ã×ËÊ-D=/Lù3ŸœŠ<‹užA>)맆¯û¯{sxídôÝ˲[ljšÒ<ÒóÚ¹¹5/E…gPI®ø©iû'–ËÆ¡¼xã³e¿+Ú2ˆï^·Ùµ5uˆHωO§±’iØ›oeÅ)Á0XJÉØ(ÅBHYõʵ]SO1Ö¿°¹ âÖK­e/Œ"èªia;Ç/K4œ;6‡ox¼ßCåº99É\À1Sð÷çÓÜòó®/ú²ëÃ[³äÖ9=´™QžÇRÙ± ¼ßÛKJ%ARÆÆy6‘ÔyÛìžêH}Š}C‘Ɉ,ºtõ®ìy”ä½xô#?ÿ+"×nvuO=‡v*ó<:£‘mSv—0¼TÍúô¡¬ÊÚ-õ´S=@_±Ý¶M±|¢·¾á瓣Ʋ©»¾Ý-{\€7Eñà5ElS¾f°kØè·ìƒ.«òÑõGuf%l ‚p¨»]‹‡+zر 8Ä´¶ì&s‡j6C¿sæ úÈ´ž˜·Ê”×Ú`à2OóC/LZ>m›Êñª«xr€óàû‡ËõÏ=¶~›aÀõ •²^³”Š¶Ø@WK¯`ú»z×> Žx=xëyÙÀmù•a2Ÿóp€øLf ¤ „5ÁpµŠµ”‡†{Qìz°X£´…ëbƒ]\I½]äIÄí÷ôÄã^@Øy‰ƒ*‰Þ–±°‹“³`ŒÐÑEUt/m]U®½Qâäa#p}@À¿]ÉÃçmÙþE>ú¶E㙓žOVÊQ\´àÚõ}»PYÄÛ¯õW‹&ƒeÙÄØoŠ² öv‘¦›í¥3bãÍgô/Ý[<‹?‡N¡'Ò¯JZnÍØ¢e,òQ¸jNç±Í†xÈŠÒ•"úì1•’d,0hœqÇ Nh{uR“5¦Xc©VÑ=ÂŽw”=!4¾"J¸ïˆ7øÄ|vͲ ³6 ÂQÔÈᤠŒ| ã¸ô¶yFYó‚yhz$÷·¡‡Hç Ü¦qš¨(÷JÐ2½  žRïF°žìÔtþ·ƒ·Sqñ=?ß5ÕªjÚ­;@Š£+¤†P%§@í•بåa”òØŠ“µöx+µÌ ¸,`{ôã7Ä'×.!bú¤%Gwh´Èø…\x±¯gê|C/^òð•aùŒÆÕ½<õø»Ã³#ºÅ:6ÍÊU³D@×{ò6–ZyÊ@laJ&µ6Ð÷z¤­˜´ÈGxœNðx4¦&S7~ÿw~®}qµûÝuD̉ÿº+ºòt ¶ˆœ³_ÚØh{,ð[ñÚÀoR¼gø* •™Ã‹¾+ÚÒU¼âã„BO"8^â\²-*{,ßôA}Ì7ç¯'¨ÄQ^“¾úz¹ÅûeÉpÁ,‹s;avïšfµ)<ˆ¥™G$¥RŠƒð¤ f6P¥¢»r³­ “ S€”Þ°Q„LÞ …?ìí|$ÁŽ’ŸË¶éºÓ%2 N. ½hð©@ÚóÍû«ßS¤#ø5Nšæ­Hƒ²…™Ë0‘ äàežt†Ççv6›E&rÌKY©ò$3Me€¯4b2q«÷Eÿ;\¾8®Ìr²µL²{Aãœ;|œÉ!9J;'n¥ }Ù3Õ†ñmQRf ÿ€y3wËä Ml—ÜwË›Ü/²$r˧šñ –ߺ-Ä-š¦ø°—vì7€€ÿ+3º p8MeFIËŽêmß-ŸŠÇ¾ÛÕ%j‹A¼+{à™ºOå€uHQØÞ¯ÃÖø«)cˆ™›@¡ãš;.à*˜“`ß!59ÍÀnP\Yäx98‚[]?ŒC`£Ü8šÓ9´KêõŽ}?ÝÑû¤ZüWRhØnv`»!ÚŽ7ÄïR°Èa‡ð‘"ë%wÑ)ØypS<ÙlÉDÆé‹’‰I=ò˜L Æ.cç‡Æ~³¼ØUU±qäóØé^pÆPKùäªUH™&Ù£|1µ€èðÐìS“ªü=Àóõœo‚ðÞwX|È°n,ðSsrÀ^j'ìòâ©Øn‚„ÿÄÜ·¨XÂ×ýj¨)åqªÄ¡xn]Ýê ,Ó! Ñ&,ä;îx_<¸–>f“)ëÙ£áŸ.Ù¢qÊÞÌüëÈj‘ã'h“óªr½­9¼‘ü,–4X~~ËàSçgçUVÒ„"†—-1Ó‰æƬ¨aŸw_YÁq,ŠŽHÂÙv[•AåSèIå\¥ Ž-_0Pá [¨,hN(ÐÜÅïÜÃÁ1Í™˜pH.n{Ú;aW¯µtžîË¢õ˜tÜR=™{*tÍ\ØÄ:1 >V –ÊÆ™™Xì]óà™’â‚žJr_(Â'_@ ™3ànY¬-˜0dÂü†žˆ‰‹4”ÆÚ¨Û=ø,o„þÕ0=ô"wôå6ìö\‚š”|çÀE»r]s'QhBCcDYà‹°èHÛŸÖ×Ögj0š<û#µ °æySqjö–çõ+B|ÌÓ ß,2Vœ6Ò.L¦P Ýjõ£þÖÿ•øVBLw˜”kȬî3arþYàÔyŠÀ<…a5Šæ)˜yŠ@Lýb>ÍíÎê3ÖRýQu’pŸà<Û>GšÓ°1±Qÿ‚Óäµ 6–’1ðO“Äy’z0ùÆ\FfsPAU~}C?@tÿ¼°’Ú2Änœ•Ûè÷^Æa'Zñ®-ŠÇ7ãxÒ-?oxûw®m=«ÍV+yÿ݉µÐÿ†|–ùr}ŽD·Å<¼¢ÞvqãÈî¬{Þø”jyLÿeí+,\»01¡F‚Úþ‹oSîºi¦C•¡Õd3·ÚQŽs4 âÏÕ °¼í+ŒBÌVZ³8³vbš2DmgIçù¤òM[Š}]ߨšbRB/]`Wðè`*6Ù̺r»_Ë_¾ò´<å’÷¸+ž[ÇcɹáÁ=E´áæÏuS­:f3 AFäj <ÛÖq=}ë ÊPüš‡gzºCÝ7®¯Ìp—¶©yUŠewضXîÙqþá­[üÃÁ×âºcÔÔî}~Mêݯ|ñÃÒªù˜¥/~Xj¯Uk†¼ÿlë ûöìÊ­ÀÉ0³†/¡Ð1­‹éGpo¥sB{þXdÁÿL–rÑ=ÃUåSÓ¬hÑã®&~ÆPGÜ-PœLxþ†<”âƒq¦¾+v]wÊ­ûãÿÝ6¾ð5›48›£‹Ì«‹FZyéj( Ï÷áûézÞ„xà¦ö>I‰w˜û_ Ê ÈÇwT:¤¸iø>-å8‡¿!É9=¾r8ÅŸ6à¿õm±mªb?Îq/ó)ÀÔŠŒ„NMŠ‘ãäóˆ@^¯ÓÃiUÀ¸°*JƒûrÀò]ºM³†cúÈÈÙ öÿ4MíŸ×]ÀÑ@ uYð‚sײáM)ªL€¡"ƒQIœhV‹ññò»«ûïþÀCš{ endstream endobj 815 0 obj << /Length 574 /Filter /FlateDecode >> stream xÚ}“MoÛ0 †ïý>Ê@£êÃúÚmËZ`C³›lXvp% Øí4ë¿eÉi‚¥;É¢H¾$úCyus§yf°‘Lfå*KÃ3%)&’få2û…~8{O˜ègN™A.J¡9„Ãà¤`* ƒÊ oùf=Ùî}ïvÞ-ªÞåL çœ 4¾W/¶ilþ»ü|sgäITs,¸ÈH,`N(^çµb£ôèc}L½m–ÖÇÔ«¦¶E³ÝU­«×£Ë“󮉷¾zò¶ÃI@Ÿ–!5’ï·¶…nêö­Y¸Æ79%h Vâ³ /4¦”eJ±"~µ¡õg7 u¡£‚`_Iï:zrMæŒInŽ‚`­ H@ K@†T¹nQÊ›.ØHIGJÿö¨ f¯->ìvMÛïk×;Û¥L÷®ï½Ù¦› †-­ M}Õuñ}vœlxpu´~êÚà^Yï¢!Ì΂—(‰P£ŠÐ%”J/ †•>N„¥Öèq“šü˜Î~c;uGJ`¢EB!cä—ý¥uc˜ªâdÛä…4fÜ·Í/#œiãý0‹µM¸¾×'ÛÞvqÍÀÞ¬âùðÖr™Pr‚¹’ÿA"Êp”ŒâM”+rl#Q¼î;ÜÊ0ö6çðëÕËäH…óŒT0œ“ –@jò&˜(=¡Ä%A÷6Ìã຾©ß ¶çvyg˜¾[Ûuã’L$Æ4±“ƒäÕmyõ›+¢ endstream endobj 834 0 obj << /Length1 739 /Length2 922 /Length3 0 /Length 1452 /Filter /FlateDecode >> stream xÚuR{8Ti®ØUÓZtùä²LÆœf4“K£ÂTj\Rf(cæ ‡™s8s†Q¬%”¨X¡-Kµ¥Õ=E©V[ZÉÓe++ÔJŠ'ÙÝ\º {}¶zö9ÿ|ïûþ¾ï÷þÞß±4øÒxR,vÇP‚Æ°ƒ¸'Ô!ÀÞ¢XZ.Ãa1`èr1sƒIg8Й1A".yx‘eþ^*¥8t¤b:ÓžAaB@ŠH‡"(…>ò<•a€Á㥪Èq-Æ•d`M6µdK)†Êc–‘÷Trù±Ö#¶>SÅ Dûÿúz #€µ,ETŠOU>!–#*‡4F!JwD K! 2±\ ñëP)ŒË`Jd$@c@Ð'š_"‰@a¥°>Hk0‘Þü?LHh <«?u+ÈèÆß+%0J€ÅãX ÌŒJ? w4új¡ç2÷E£+•V LŠ ¡€Éb1Ž‹c)`ˆ¶0BZWXMK·C1‚¼"UD;QnÁÒ“_×ç §Þ¾gJí¼¥Ôä'¶æzVÔ´»õÒÞxó·&ž/ÅLBóÒÚ ^eyÏ›,Ò}“½þþûÇJûÎSÏk+”BN7P€¢~}³\Sb'¯ ¢{Rz€™Åâ×€@ý]_ᢽ*?þºFí¤dGs-ß°kOc`*›«1u5>p;+˜3Ïû‹z£.ûc‘ª’kÑ“#«£{Þí.n)¾ÜÖöâæDn7¼ÁÅ7ÎÕª`íp®ÑéAæxÊQvº00sý.žeu/ô®¶G×ÙKUšP½ªï©á*^ëݱ.Alph¿ÐÖˆjòÆøEsêsÓ±®‘u[—.„ëzê3sÜÞw]¦»õ•ÚÚmëL´v€M^>¾•ÿzuÞ_¨;gmÆ©B4¸£=§Öãú³çʃ"ê vÕýU[Éiº- 6þ•¤õa³w;Å5wøë~{²ÁnÀëÌƪêÝŒ[¼‰Ë ·qÎ9ïÜPfŒÿÖ»yç=õùò)ÚÝ·XK­muÌ/ g£61ufþ/û&=í³#—{)öðÏVä^åµÖü9°§ªéGƒ¡9C•Ã³Ê²³ +ëväƒýé/}4ÝxÇpKê„Óü §«Ê¶N㬠qãîL þa蘽0ý2õ”gþ$øYÝÝÖùŽ®û›ô^r]SBÎõ¿“7Æ B #j\6=Ž3ôL»žY´$nåñA+û#NíñiúIæϺ󦢔žˆj/å+ÒËûëovV¥Ìµ8ž^”Ã=*êŠPw¯›mÔk3±PÏ]4äþpaÆ–æ·ÍH«ÖÇ¢"c‘Àaáõp\'*~÷½²éñvïkYÍ1×4‘5ùÍ‹¯@S-¤žš'–æQüÚb ÿç{N¼dÇ©0kŒ1SgàÁ”Z-ÅöðÚ'eê>+oŽ>¿\t*æš^ßpÜ´r~Iyf´¦›RâMÃYS)´3âûvâôÉ.+Sй·KçšéÝœ‘Ý¢ccÀ¬ÞüüêëÇS—}WÚÞš&Kœ¾=ܤ±>B8ë÷Êþm¾:¿¼®Ë{`¼Íðþ¦³³û·^Yue—uKKÓ\véˆú{ç$‰õrÕ[Žw®Uœ‹¯\ü­ÎpTÈZyi¯26vŠ×ÅzOÖΫ~ä´ b¸,)Ó$·fF6ˆ. FLL9‘œtñ,|³ endstream endobj 836 0 obj << /Length1 766 /Length2 1059 /Length3 0 /Length 1602 /Filter /FlateDecode >> stream xÚ}R}TLiÎ×8²}­ …}CTÜùÊÔÌÔÙÝ&I1™UMRè6÷6s×Ìܹ“•>¤E%ª¡¶,­È–Úíƒ6Ãú*)l„„I»‰²{“öìÚsöÜÞçy~ïû{~ÏïÚÛ‰ü!DŽ.Qâ$Ä¢3ù€ 5á 8ÓYÆööž “˜_ “(°Ø ‡Áf2™&)ħ"!U&ªU°t¸‚É^X\.ÇŘÍ&!A8*ÅpcÆpGPBŽá¨H©Â†S‹ÉüH a’õ8ªRÎÉOIbÊøÃ”Ô ·hùÇNPßè{°J‚â$àŽb!ØG>‚ ø} ßåÂ¥>‹|Xä{Ñ —( —6ÇÀk™€E!ˆfŒ2¯¨†—AÇ•$uDªÉX¡$Œ‡÷ÄbF8JÂÃìá £øß6D0†“ÚÈ¿Óöõ³F°?I(×£ABÊþY"„IÓ„0éL&‹â©oô´æ¿s JM4ÄâòÄfsÉ®¼E±ÿ*”¨ ‚Jìý¿F¥5Š#0jÑ(ªA%Æm-J‰[Ò×'’ª^—yØn„:¯ÒîvÔ¤¤ë·qÈy7 ¢eË'ê¼›6ÂcÓ.g‹Z‹“lž¿®×„ÖîOhß»ô[ÒÈKÿ@Ð^xU+Eg!JiƒkvÊ‹_Í4 2wPËàV×_œ»~xtp¨FÕµš×  ès»½¶dšG~M÷ØÁvs;ÈKÁ¡æ;?%BrÔ> Cc“\ç ùX>ÍjMvá×g7‹øé™ô-ƒØ§§é¢e¦tÀƒgw8æEµµïõSzgÖ”8øS«6wî Á ñÏõDÛºæÔœóäøǕšeu!Ø–« í*õìdyV “Ëîœe;^d>U¬¬¦EöšÖè¬÷\Ô~QT÷ìm\žÍuëI“¦Ï8sqnOé%çsWnæèjZVêŸ$Ç$Þßf‘3Sk›['ˆÛé89ž½Áºª"ýÇÆyDAâJ÷£úiÄç§i >ÝWö[Ÿa)bcÀ»ÑQå‘ï¿Ú]èÛÈ­Ùãî_[î7¦¤¸Íˆ›æëûvÇ 'tž×W9á“çÓÉÖ"ïôÓlSݔҊºNï!gZ½§XðÅšÌòk•Ì:}Úé°Ääš“û­öEKΣ+ñ¼Ó<øµò±OÀ¹GâWnõά+s¡èlÁ¥Êr;ÎÍx ÐO–𺉻öhFßåwêßÍÚLz—Õ›½‹®ÚÑ6ºyiŸ:4ö>ÙøVøeQxqQ›ùÊŒ£•®ùŽQNQ;{2cv ›ß¹Õ»<®˜Rõ&óéü4]gPÆ”€¥ºï~Ÿï2¥ÙeGžÖ &ÔôOŽù¦;ü6âúý ÿÖU3Ÿ¼> ]¦á›¡˜Š¤-4±åÚÔ®ëÜYŒ`çêË»«m­Ø²Øüéå‰tßв©o¢ƒÙVC[.» MË}ÈuwŸiy»0uúÜï?›0Éõ¾`< —•ØøG»BôîMªÄAq÷íôê gü½Éîƒ]Úü'Áþ½Aì ½=†„7óž«£ÖåZ}3îMáÉ]sWÔ;[<èpâ~-¾(••Î?e)3!Âbä×y‡·¦ÖÚš§;­JŒ¿e½fûg¦‡²‚mÛ¥¥÷‰© ¹|s¶:…·ÍsAUh“U˜%¹®6³1Ý·üWël·w;M§ß¼Í·2‰#Œ+š3Ç.>-ϳ4e™¬ÿ²59n?äöƒ™»|S,c†Qäá±k²²® xܳÚr¼<ð€e=óÈ î™UoßÝÍþ­T2ä¹{B?ö­Lº}+æÌMN ½TÀ»q?/^ç¡ûdƆ_Um«~vóZ±Ñ²IÚA;<{™ïŸãKºçܼi´þ»²øÖ†g¦U¿8­–™Võ¿¼£ˆµÙiÄk8²'üÞ™q}‰êù·Ìú£«7Ñ®pz|¡·ƒŠ‡Õ_ ‡† endstream endobj 838 0 obj << /Length1 1023 /Length2 2837 /Length3 0 /Length 3520 /Filter /FlateDecode >> stream xÚ}Sy<”k–,ciA*KÎ+ Yfa¬É–,Fö%5ͼÆ4f0;•&BY³d9YB"K–5T–,ƒ(KÔ±¥EÙòMuúêœï÷û~ï?ÏuÝ×s?×½¼ò²h;U#õ xŒJ¡«ÂÕ`º€/A&êjpAyy_C'P)G1tP€# p$ƒÁ ƒt94Š#sDùÑ0ø¯ B€kkÁ´0GÀÒ3 ž@„~ÍoAñ¤pÍï<ÎÏûGÌô¥q9*œ'qT é€=9÷üH$k ¿ùúŸ0†L û?'€÷¢Š(Gð#ÿ;jAÇX# ž°ïvŒÀqhëxbH4ð;ï@Á¾$DSi„¯=Tá0Ø¿bö^,‘Òhòï5•NÀrÌ9þ]#§DÀ mõo' ™Ó¼ù04,H¡Ú?0üN þ&@ îß}à´÷[ h”™“½ò÷)~‹™R°T‚HMãë‹9'à„á〠NµP5 •Î¹xûуOª¯à×!ÁáCòöÂ|¥¿3z¤ÿ$:K%“‘hפ_4p-úë¸:Åc~½£ÉI‹ÿºp ïO@I_ûùƒ@pDd¿Ÿ“…òªs • ⱦ @½A_÷SÄÉéíEø™ƒãŒFÀÿê„S2„¡yýtËIC÷úÕ?gÑ¡ŒïIþ9 4†@¡ÛŸóþïF}Í7 ÿŽíè¾T"èDÀѽ~• 0t_à ¦ƒÁ9<çûq:ù¿³66¦2UáªN¯u`@Kü!ÖÏ×—³5ßþ6ÎÆüÀžβƒ Ä  P±z‘g+"ë?—I¨¾zÂ÷b¼ñB¼KÛy$ýà`v —šY¯ñiÜ­½ÉhvatÏ@Ô¡…ÏOîÍac×̳è\¦m“ÆoUsX+ï›ñ  Ž*‰ïÒJ¾0)²xÅvߎÓnÌ­Wµ|¾¡ð|Üý>>lrz‹‹éˆóÖþÙBªõg?+^d÷µÃ<ólÄSM±ÜëÁÍïÒ>?Æ~@úÿ~_ ¹ Æ6Údµë·®7…ûƒÍíFl]å.\e¦ÝùPX`乧` ìÅV‚@“Ê´“>wgÔaÕ¢\¯rÛÌçµµgDÑœæg]Gë-N\¨Ý#5¯/›°†‰ßëÑŽñˆÉ×›£vV'  .cÚutû b¢>‹ÃO1©G´£#Ý;²uú;f‘yw„ˆ¬Œø²£imÅK½6¯‡ì\süj/¦¿ã˪Øm4õ~›.ÿ­lŸ¨sš®®z5F°®÷¼¸N2”fß)¶•wuÕz|h¡¤/nUŽgk^~PvÌB“;™Ö 3…‡ÍÅŒ C æ[ÝÛzhpp¥ú$îË{Ó˜ªý‹"¡Gû©m—ü/¹””–‡x‚Yà˜QÿÂ^è},ž â[ÞK%ë>b»“ðö†óWN®4„ƒ=Lvp¥ÊËñù`ä`[ObÊ[fgfçäXZø0iÏؾ´u'ý jûa Dæ°\Xã²F„—Ÿ”fEÔìØïhèë¹ÌKì`í4ÑXÍFû¸O駄K–ÇdÊÀTÛiÉÕ^Gˆc^½@¾””œï#ìe’¡B­qƒÜü±††Þ>ûÿ¨-F®B™÷zžÆr ¨ÜhæÏÏM4jM¼NÊÀ6VKuKÝzêÕxg·XÓun+uíš*VGOeÂË£µ3îÔíÜìâ2n'–`ÕÞLX²ª tȨ¿I‚"LÉùy°» )}®o',[6ŠÝU'Z£&d¦¯ÍÚowR^ߨÓäí’ µs OÙ¢-ÖÕ5QòÌ5]Íñ6’õéC˜Þy%ÙH V®Juk¶¬Ç肬-+7¸6úŠx6Á²ÚñlöSÂåµvšïä±dÌiC‹€ÑöŠš™YÅ?7sùjc ˜ûhGý‘&ºrlâOn7t9TáxÝûòsp~PöE{¡‘ÎsQiÊsÙ”„~íË’Ï> Ç‘Ã#?î>ßéÒOØ}s¿WÒoh Á5SAòåù1j5«°ðpÙ™P-"WXo}§hCoî998Ùû‘i¡'9êDÝèGí}¼3Dß߈ÿŒ±cõÝFSo•‹{úÖÙ¡Öè«áIÁ~òBñyq-ÍÛñ7ë;žTÔÞ?qÃçïeÞÕQOÌu(.öfð’îêJ¨9–.嫹 G\êÑ,Šäæ Ôœç=|¦{‰(²­‚½¡[Çç§%Áö[¹ÇŠRÒ<–.?dòð’¡]æñšú¿®ZÙ!grt/½ÞLÛ3¶]`Âztüß%æ#Äú8£÷¦!rqs³¼w)¬žwÇÃèñâþBwöi;ß“ž[6êÛÿ‘j”ßÖ=\=,|²c™ç}âaÙ»Aûµ[¥—çuˆ:ÏUÿbñ ?Õ|fà•gÆ©´úyºšÏž”•‘µ¼\ÕÎæ|®GFð1ï3:Žy Í´ö®(ñ±?#™mBõäƬJ«AÅHµ$HVùµGqñ}ŒcìÜ›¼åFO>Öð¥ÛMšÓõˆšr8%ZHYt‰ßÊJ”ü¬ñ²qènà“u…»ÄZÑÿF±ðÓÀ8ümXTÿSñÕ’“Õú„)wK-@4-w>i˦Âù¢mÎj™TÓ#Â…ciSo6ŠäT‡RohÚ`የV,O‡§¹xÉÎÌõìíjÎY!5<.ðAñ·jþŠ˜€B{~«:ñ;gƇ…Ýþ«“kc~oì®ÒË,kÂ×ΧڵCZNr¾^¬i4…¼å:ý´boè+Åœ@¥Íól­Ž{šw<ÜnGE}jëzF÷åz÷6TÞ†«±Ë„wêÅš¦;õ­p™ÒkSX=Ñqi,žeþؼSgÓW|úÊʇ_bäó×·»@ÂTÄI'i¨!k±ñôøÔ#Ø0N,µJ‚v»S8Æ‹"(½œ±ÑkÙmºçËg7çZÅN#;*,wD±m‚%}¼gr×0ËVÁ"îr,P2Mç—9?¤YÏã¥ü'Æ<°›%—\‡ ¡Å†enñÈÆ8/íã ¹76÷þÅŽoÞ®-ÄíÓ­k¹꾆ÐZØu¯‹+rõ#ÉýñŒ]¿ËØ[ ßÏeÄ=Ó®–ØiZ1Ó!\ëu£Ó¹dñBÕ*îLJ&„}/Ç|’¢’Òó8EÃ0B4õã½(™ÖÂÈÇiâJ63J-7Õb.– ÌȪf¹¥#¥ ‰S§NJ_iÍKèB†øHØ {ˆ³ýGã¥ËÉ­«G,'Áǧ_ïNÑZ8åh¢rò`*™^h6e¿Â‚Ì×iYì÷˜øx–³‹º´Jüå]SÒ=ä «A Ñh=½ïÜÇ@"öñaQ 7x¥„×€÷v¯Ú¦4|ªÅòšírå€/k^9»è·ÅÐz·4Ϭ»Á«^vh”/,–LNN!ãÜìíêõ3⾸*µIôpwOžÁ–%UÝÊ09¾«‹h)Õn_¶2K€, úljçžÃ0ð$HfQÛsnÌêî+Ù/ÓYù&RM WÊ&V#&šCe—4ʕѳ éh ÝT4›–äZÊØø jÝlÖ¿[F‰ý!Ü\Ùä–ÐÒ¸J¥eOâ-]†§…Îo»yíˆÜë¥O”7἟ ®n¯7ŸŸ‡¸V>ñ€ëIxZfU¼ìJ{Îü(ÞÿÖ¿K©BZÄÒx!#bù€ôð؆•ì’Açô}ªúoh½ÀÊë:NX¯­nþü¨WÚ¨ ›ƒÏhM£eøoøèˆv Üq¿YëºúN&ñ,æ/xIgd‘|i™4éÇEˆ6ò·Hmã½M¥†E<Ÿ®—³M÷2ֆй†¨kꨲ÷g H7O”K·¹:ˆ8„®˜3/ìm<ìÈZÊÜH¬)“îÌbÎö>ß¹Q#Ô’µ.cÙ6Ón`EŠJ8ÌÙîóRn9SZJêF»1L&=±Áå$j±%fðiA.¥ºs‡ããR)•O0::±yR í¶iåyê'îã¹×Y:ßTE²{‡y$_Ÿn›„Õ£"¯‘@>Ÿ.tK1>î>ªùÆV-§½:~›íeý±  O±Nãõ¥ðOEÒÆ endstream endobj 840 0 obj << /Length1 808 /Length2 1347 /Length3 0 /Length 1915 /Filter /FlateDecode >> stream xÚuR{<”韕Óe–PQérÉ\ Ã8 åÓ™¢,Ãf̼3ÞÌ…™w˜É%•×r)É¥h“Ë&ÑGÇ’[…2±¹¶«ìäiÑ9n-öEv϶Ÿóyþy¾ßïïy~ßß÷yôuhn[–Àtða ‹§BX"L±x´¾þA!È€!ÿ)ˆ#pD<0`Q Š” ŠE ÎržHB¢)MÄ,ˆ ~ â£qË×;óÙ€`¾Ê³ÄkZ0(!MC¤©€´d ø\)ÀÙÈ91—{„ÁÃe[Q<ˆ+ýÿºqüaÀ ² 1ïsÕfp!¦-ŸÃü*‰ È¢A0Ó`3¸"p•?ÎgB.Äi´œ€!àñŸiîþ3€ŠD€Ù'鈆˜ˆ·Ÿ&D4i®Ÿ;yHtk÷1DLk˜®ÄOÈg}îJ8GW/cûVžpE²ç3,ˆÏˆfæC(dHÑx€€ 3 ”@ˆu JaqX¾FŽb8` „èå2'80HÌà.Ó+ aØì? iB¿c’)€ äŠEd'â2Dþ+ÌŸ}Óv—þÿò +˜°ŠÝ`¡ ô€X°ÿÿ–P°’ÐñX<ž€ðÈZÛùü5;;$CÆ"Á [s Óð?Õ1ÅB!’ðÊ¿DÒ]Ãlù (™èÞ.Ó*êÔݨêÙ;¶š!˜Ágë_ ÔÄ&y6ŘÁ{»sCý]7d:vÙù²Jv´¦Óz £Û»Â4ß϶I¼ë²Îö§9åÀ(û&¹Ý&¯yîCÔf vrdäôX¹Ú/ÉGw©øÒé©]¿^ ×›—å/Ôˆ†½,ÇpsZU'MN´Í®Sˆ÷ÔÑ€[<½U/nÒ3ÄîÎÇe QdÝçm£—eÞÑ攧”³9p{ÚåÅ)Ò‹€MŒëë=¾Øxñ:/çû°ôº¨0Òÿ¦‰™þþ^Ìxæ효˜oRêZâŒ3ô4þ èO‘$J6)Ô[ú/!VÔŒ>°š²ì™øúËô'¶é”¨ .Næª9½©†‘¶Šãw܆uÐOÆM¼kK¨\0" \mþéÃ^LÀ¬‚Sùøõ¯N4l}GÄ$7ùn(VôL¿ jWyäÕÊý-®‘ïÏ/ÄiAS¦Ú˜Ûé/ͼŸ}[Ûæ„JŸÜXCÑÐ0‰zãÐi›ýç…¤æRTzúçY—‚¹K Á²–¢wß.éÕGµ¼<7?cq ÆŠd00ÝÅÒåŽÑ½ ÷¿•çDXï…ž{\¬Òàë„ žÙTñ0ÉxŽ¸ø¯”}˜üûéäHõI£gÎ<ØÓ ÞÛeíª¼uÞ³êTEß9"~ª¯½<ø£tjZžeµÍóÁŽj+–×É­Cã]©:šÙñ”Þ]’m•ÞDjZ}×dm7_[ïcŠç l-Ò¨,%-6$÷¸öM-¼¯§€¤â,“è-Od™¨Â„ykæìR÷ßì§?”)Ãoª Œù‹Õû”2›þA9Ö´á…““n/ú¨§´Ëfé^‡r˜cìÏ}qÙç¥~wÛ¾Ò±ynsÆ@“᫨viW°¹#Ê UFìM÷5>ºœ{¥1æF›87l-Ü1¿ð08ÑÇÄÿô«A^ÐÑ·b\-´sýÕÆæ23Ò­˜@í'¶W4Ñë°s åjOjÜÛì°-°õ­*ìÏ.è7ħþÇ¢w·4LéTö\Fš1‰¶óöÒF]×_`˜ø½“Ó²ßI¦× endstream endobj 842 0 obj << /Length1 1283 /Length2 4498 /Length3 0 /Length 5271 /Filter /FlateDecode >> stream xÚ}”gX“I×€)RD)¡È#½!¡éE¤÷’BI i‚4¥7¥.E‘é©‚tQDŠô"E/®»+û¾×õýJîsÎÌÜsfæá½j`,¢‚À:"5±¼DTLÀC|ñx@\TŒŠ—W ‡„áÑXŒ: ” P0D `xÉþèÊÌô¼½`¨ŸbP *-++Ch8pD¢Ð*ðÏ´1NX"õ+Žðöø;çƒÄy‹ „%XŒ›€@:Æy»¹Ý‚¹#_bÿ“‡¹£Ýüþ¿ s$åŒô´·ûgµñ074\ƒrC"²¢2b²¿h/M´/a€ÆÃ<Îù+lŠA qnh Òë…þÙ@Bhÿs&Îh¸+éåHþ•º…Å£áE³¿¶JØé5@Ë@÷¿uî„þ=Ì ŽÄà™¿ü€þ@bÿÝ B—ÿìØB×ÂÒPEøWKDŒÝ`¼=DJúW™ŽE 1(*)Àp8˜•!$àЄmøH_¾Á¢,ž0ððÆNXÕÏc#LVûúE2Xý7É`H €oü&q¬ý›³èýC2bØà7*“6ùMRØì’%ÌûMÇßDpÿC? Œ8ƒŒ<ƒU§3H0@A‚‚ó”Àè3Hp=ƒ ·3HÐpÿ‚æ 4°g áñJj=`8$Æ é„ÿ…üÅýy³%HâÎ AÒë úv¶˜àì} Î>gà|û73Ø÷ üþÄ_?ƒ7ñó@b¿ï㟠ùÅÆxÖiŽFàÏ–èÁð8´¯µ˜¨áPÄ€Ÿ¿b?/ë?Àö﹪*Ö×_*ˆ@!âD’°?hà¿ áÞ8B›ð~p¯åovB^;é‹„SMŽcáòa.aMe*l·EÈ?|l¾Ÿ`Ù%‰ç›ÈõwÖ¥ÈÐñTu@”°ô=2xW1<À&´~0èkÓ–2ûðFžH£s^uS$¯çp§ …¼‚À^F I?º?O·‘hÈqÑÁšÔš?Ù|üG¸t»øâ‹/OŽ›½­d¿î@Á·KW²ácUþhþJmy•ç#¾×ÒæR=Î:ÍÛDÛtè˜$LšûX›q%eÈ&BJ®×6¼ +ÒÄÊ0KôÇ]7Å¥“žmvzò½Ï &­·÷m0å1ìpå€jÖf®§–®ŽRÙˆÃ+ptB`M‘PS঱|åR©Ñz1ž¿}½”èD æmáÐ-[æÚ9¬|80ƒ›þDSz`­âÙþì %$ú:üÑÎnõv2'¤`ºoN€<šÜ·ˆ?_ÉSJÉÐ!¶X¬»©/n“$ï4½ó¦Hj_5¬K½FÄn'èDbwÜíI†VãƒÜ=kå±wDæÙ„,_‚Eåñâf»’“?T‘Ÿä¦bß¹Ì'}`Záà0fzÛ\EâÓ¥Çú2.2W¶ÕMÖ'ä †ÞßË{xùÙWì‘„í!ÄÍ^t"}+÷3ã^ó¸R¢4`”¦†¾§‚ìbÄlµî^Æ+÷7nn@Êí­,—W*/x’7}çÔ_üÞ+¹A/üˆÜh–ÞP)+nòÄÏr]t¡ˆ«¤`DãÅ j#2hã@KÀБTR{"N8{kµëù,›ÛÉu —’h3¸ÛV6ãÝY\ôÈu(õ&ÊÈX?¬©T¦tˆž+â P–á›`0¶Æ³Ò‚Gà¯o3.ìïí!¼™®)*60ù²féËÊÔþKHÅ&q[=Å»ž°ÚÔïZ¬jHb3µ5_'Y›t‡ŸnUo Ug8{‹î d }Û]ŠÐSÎÂjS²Î>ý@rññB®#†”ظy_ÀgÁ?$u˜ap,øèöD™Ï³µ$;í¼–O%Æ‘+ݬ. ̧T[ê„ÚfÞÊ.ÊA§~žú8+ï,±ÄÑþê(%8;˜8GtßΧ;?_,ÙI«õ˜¬´…œ^˜¤H”$/££dû¼j¿É³Ó£(Î\ç!)mýVZv”D^eÑw÷¶›Ö²0¥ÁöÕ'ãéÉ:cm’^ÐÞ˜¯e[ž#ùýVöð û>«®€jÎon+ÞèWX?ùÑ’8{}†™ñ ¾0%›û&G®>†I¦+=–ÞS‚`3½Ù&®œYíªQ^më8~â{Ë’,ÏG}S^„’M¨¢8>9Ç­—×’0Éùåíe¡´ëÎëêÒj7ÞT<)¯øDË?79ÛöÍä{¹«`ÄRÑìšCõLË;E‹S›™î£ÂŽ*UzÞŽÞ!Ï„ð—V<‰Uòã–Fïð—Sl>Ê;Eºñßê;KB0ƒ‘Ë¢ºE‚Å}»Uo˜·ï¥Ö0©¨m'8)þ!@©ïçÌDéâ¹lhïÄAÅ–û¨$Ž6³†ž<|†égêÜ¢OtcøžÊ;³Kù¨ÑÁ­ÞæSÉ1ùâî,ç–¶#U¡l`â@% µ;ЦªÎõ/ìA±Í6ÌJbwbûP_UV[C%¹^@âܽF´¯MBƒÁn) ×AvýþX9«báöáè^™[tfJ0(qÿ¹®ÕÇr.Yê/nYC @*þ_eçCâÕÉô( öωjŠ»€Ó]÷ö]Úî…\É°m^Ë+z¤ÿTqLaçhã|š©£åŠŸ\yVŸÝÞ¹ù¨þžm îvÅŒƒv Y;ç™ ò„îdË€ZÁÁ PµÄŠ,yž÷msÌmâñçY/YÎ{=µiÏjZ*}•×åéÇbT:Úòs>xb¨6æõ~^ã‰$ 3MŽ‘ؘOÒL¾¼š ðŸ‡ç­0ƒŒÝ¦/Aî®Ãòõh×(‰#¼N×>1ø0ØîÄ”ÍGáY˜xOÕÛ˜²_·e[f.y_IÁÖy&=Qi¸×3ÙE9¹Ê­Ã,ž¹øˆô=0wùÆñ1l¦<¤8„T:[N3;£šŒdÓ ºh$WËfªJë¹m·>­¾$L¬œÃwHcbáõ\éò•õ×»vûº92s¥ØÓÈÅ«gáw¾Ìºy‡f5–0œW!ëw1Ø<•}ê2£2–‹xy0Y¯ªyÁEú©¾}¶‘µÝl>£©e¼ÔÂm§'×÷Lé¡ž3Ál0A^· §ÙkgWŸ˜)Z ëUs, XHÞkŽ­úcd¨Þé „ª¥9™ÿè Ç|ªn]+ïSÁ»·4?¬a›áLˆ5W?žÑ´“f‰¾ wÕwˆ“–¥£ âµ3 „Ce§TÖêuˆ-Àjí­æ¡ÞÿÖÇFŠI!¢YNHÛën£˜!Ý[6Ÿ‡æhÖ{MÄ®{‹Õ7üÁkÅ#æÊR9¡$9-d÷LßÓmH™S[‰®¾§•±J C®²±ÔWüA;Ÿ¶à”ºÖLyTÙܱ­í*ßr3{ޗᬯ‰Ø\<7®xÆ6+`ï|µØg·Ü×?ؔ̊¬ù"¾š´¨äc´Ï'K5×ZùïžjP§æ=õ{{Øh é¦Â\¼ªw¢a}ÅXýÚº§WÆ÷ˆcõ7WÜÖØhuš>ˆžÓ£»ìGz¿¬J-*]O¸ý  XèNWÔpp4\Üï]Zr5¸ôÓ/EU<\X¤vÚ^·U54VzãY"—„*u@—Ü4"6Îa ×}`ﱬÌjËÃÂÄèUEÇÛ(¯:RñÀ< Ž»GØTšUe~.äs&º'èŸ1_y®1%í:Ÿ¨qÕ‡J2Ê\TA|daܳöWÙ7?d‹ÌE Ž#]YE?éñ°Nkèš\uGííƒ#¹Ž2{ÅêúâðiR2Ê1¶é~ŒÏ#ñ¶ä¢Búƒû“É/II¬:êûS]•R_dÃØ%Ùùxû.kŸ@>=¿ŸT7T¡ô¡–G½¬e~_¶Vt´}ÕI2~͹¢`ßÒ5ÞÌþëHÓÛ„…v¨Ãcø{=!Ž1N>ÇP÷ dg4$Ù’95÷ã,·Æ3ÜNƒ„.Rw©eéì1zQXgÉxlgºÔµú«šu‹l]¿Ý÷‚ÿ®w˜Óµ³>´—_ç9”Û‹yÙ§iï°õðóekÓA\a-ÕTœ¸ñMðÛ¯Å,篛´l%ü‡–¾c'XWjë”9E îþ‘­Édµü륗ù÷«7¶æï)°| ð姷ŽäEí3Á¥|ÜZ‰³^Ã7H‚”dº‚Öv{áwêO.ZÛ8 Óåu9˜*(xíÍ gT uŠäïdz$ìjªÄÜ´H|©Sk^L뾫>U•E9c¨äÑÒw^‰×.ñž×º-„|ůǘpq³‡C—+dÅÅ›‡qu«Ë¶ÇÖñqäacZeZR~•ýŠµrÔärÖ÷u·}"sÚB‰;»’õš>ÎÈÂI¦ÏìžyÙ",í¹ÍÒ—PÖàä1ñ~g÷~fÿ…JJqnõw”Ѿ㥜Çñ1ò0íÌlä‡öTG}ßKo•¿œLêƒPG¾‰Û¿¯_~q ›»?íù>bÒBS­…‰›+ÃÛb fñ/Ï ÞØO¾ë’¶·Ô’5Ÿ"Ž6ÍþˆânYϪüQŸÙB}øEÓ:ÇÔZr¢ÊüöNÞ'Úæzw{ Ea`Z·Ïž*t°âŽnª ~^}Ážn.”~ÍHèª#Ç=ñÕïà‚eÕ>½×ÀZç@³‰Œ ³/³>Ï£Kís%|2è0Õ¢9#AE…ïêÛ;œêýª)ûbé(|{µÍÐé"éT¸Ñðc¦Ùs+¬;–/<¸Hä_;×qNeðÕ«r™ÕùýÙs¼BT¸;~¼üÁ#œ/‚ì îã ëiÂu3¬Œ%3câùz½ŒN®½GË$sx¡ÿ†Á:·‡ín+‚º wêP¥ôÙ/öDò®ï8â²ÍˆÆòr2Ñ™¨*SxKt—ÐÆKÎ5!.¸ÀÅŽ˜þ΋ó*öaŒ/637Žd틳¤YÝõ&R¡ñéu4Ímªb7ãࣙp´…‰ù’aÚ Û{ XýnãßåØ@ˆøó§­õ.i> stream xÚ}•uXÔ[÷·) Iiº†nîFr††!†Rn‘FºS:E¤‘îixÇsžsðy~×õ^óÏÜk­½Ög¯µ÷þ2ÑiépJ[A-! Pg'Pãñ†Á|\À‡LL²nÌê,‚AD<¼Ü<üܼ@ ‚!HñGK¦¯îá²ùäåð ñó?ä¬ìÀ0€%ÄÆÎù!÷ïüÊÎÖPàßv+—|ž7wD+¢(QÒ êìè°‚X#Öy8:j€œ Ö¿tý7ÈÉÎÑçÿ`±³±…XÕ!VvNÿëU†íÀÒÎ6Žðo“»‚7ÄJ˶ÀÜ< ›õœ­ nŽvÎ-¨»Ýï–8y€ÀÿñéÚÚœ!îîÿ¸4 0;0B›þ¶ˆØဢ–Úÿ 8!z÷O>;â ÿÃV¿ ¼ÿ1@œ­þ· ˆîþÕn m%}yMŽ¿‡ø—OÞ µ²s¶ð @nn Ÿ‡@‚/yvíÞˆ7b³Ü\ÎPb ÀÅæ°†º=ü=#A·ôoÓß$à–¹'!·ì= ¸åîIÀ-ÿ/ Ü ÷ÄàV¼'^·Ò=ñ¸•ï Q]õžÕÕî Q]ýžÕ5î Q]ó_FT׺'Dõg÷„¨®}Oˆê:÷ÄàÖ½'„½{BhÑ¿'„ƒ{Bh1ü—D>пÄËXˆ˜°â9ZAî£k,ÿˆB”¶¹Ý»¢-Ý@`Ìb »·óýkwûë¤ÿë@tü/ ²ƒ¡Žˆø…Ÿÿ·ÅÉéeˆãÅmõ/ò!tZAÿÁD¨¸—Œ8OÜÿ**øÛïêrüc ¢¿Ö÷Kr­í<ÿÈñÛ õø³"Äæ>#Âoóû‚ü‚Ðn{¿Ä\l}\l!ÎD lv b ¢÷Ûvü}iïýˆÞ9Ý#"à>·"—3â²ÿáGìz/±ú_nÄ\î݈d. 7ˆó‘Ÿç?Öÿ"¢– Äí¯Ûÿo¨à_6;èý¨ø-sqôøc<‹ë}D¿\= 0ˆ•å“áAXÿh*¢e÷~Äóž ÂÝí¼ï‹"´¹;‚ÜmÿHv/T±'˜­äa#öó‚þ±‘ÃãDÌÅóDtÒë+Xíý"Òûüˆ.¿¸‡Èôâöw©ÿ~+µ@vÎ0]—_üßç_Ìó7ëÀÜ ;+˜íŸ!ê Ä ó6r·øû÷Ï?ÓÿûËÈ@½_r >ˆ…G'ÌËë÷_q`7ļa} /ú?lm‡øA ÞðÙ)(X,ؾ&¸õ¼JšÒ‹sm}q¹="Þ¨7\Æ<ýÒV #MqÜUƪœl0Eë{qèØ”/%ûþùWo“®ôÀ…·JY0$ùÞU™CÎœ/']6+(…͈PJÄ*ÁAÂ3êGÆhÆ,o ¦®C„ºù6*Ö?Ü´»o<Ù8 ~áÓ½¥„ÅHg¶ï DÑ1.ÃŒLðc ÝŒS=t•õFnP‚…n”‰·“FLBELCZ¡œ¯È‰2%7On¿,wåº UÑá?‰t;½ÎLœ«£©ÀnXÃ2é‡Ù®ŠjªRU㟸£ÅýŠÙ[[Žò[4ø•Úû¥0–îýJ¤[Y‡¡E¿0h‹þä¢öíð¼ÛÜ Nå¹±j kwÑyG W_Hʸ›Ù®×Zž‚¹Á%~- ¨ÿÄ›Á (wÉvŽTRš*²áF£Š&ŸI¢˜õÜÉh±à™Lðg¹N³“W·ü§SŽÒ["³áÆO“€Ë/8‡ÅW)ÙÚ¸¹Ä`|ú§3×2ÑÙ˜ïö«‰‹·©©uHCFJ^wí|syB“ƒiHk»gšâsÙ‚"à9?.(Šv —üRx8/CpàQ „ E:±•$V•¾8 ,`¸4ä3±óQç)ìéЕN{3D¾½U]+嫳æŠÞzE«¹q5 p@È‘‚®½@øL*=væÖÇhŸk­˜¾¼`\¾â+½ÍAØ«ƒsEÖg–hÊÓ±G»ŸK®ÓM¼Þ4Æã'â¥1t7oƦ¼ØØgsùÐø™èòB³ŒùŽ/ÐÛ<5½g"ߟ÷·nv•l!¤X è ß3—Bº&.Z=}Ý5üõíîOoÁ:ŠG£YdC7ÁÓÑ̺¯Ymwµc‹=°‡9\cR‚¬÷zÔØ>j÷º“Íñ½P$]ÒbPüL/.å9*£mnxL€õí»mè1 ò¦ÆËÿ Þ Í:‰ùãn€<äSÓW÷ð²–}Ýäwø¥å Í^£“ÄÕt›êt×3ðôtC^Á¦×’Ú§'1"yq> J¼ðQy,Œ¾ÔŠ.RÃ;ÝETÀ_™HZü+‡ZL$“M!owû)C+廀r¿€à{ (n¾• œ6ÀŠA.½ äÖ6Àñ0‹®û~•×Òºf²]êçw /oÈËH‹g­)w~gQiwjr†ÑoG!†àžÍ3dc…® ‘]5OYqQ»QÊ’sá…”qR0s²ÅÉÈW˶bÄ"æëÎá8…™ LÒ~˜qS gãC‚1µÈ¦Ô?TH>Ì ÑÇÖ/œ‡6¦1.ÑuN„>k¬Œ{Û•ØS“s#*å+…Vº·nËf—Td$e~½ã\ jÌä 7Ñ“´mY0.Nïë“È—IL@ n[VïC_gù°9#yÅ€ FO0qvücwsËéR²bk¯»- àÚ%75U¾Ãû±%éB0µ!xLq̧8Ùà—A¾ì†JZ3®€µÿºP`a]¿Ž ¿)Œc«èý@¨¦ì$ˆ[“Ê›Œèç®f’‹JÑkcÜ 'æsÇaëÞ©ù˜«ÎÌq@ Yù²7`WKf&çSw$ZÚf3‚ÁJ§“V‡7É~dÜPŽ©…Ön™¬û,ã¸ìÌ*MH³Qénxû kÒ—5ð*E x0‹PHªÑzL»H¶ÓAzcÔeŽHÍr(næsÊùXl„ÑáëÚ[jSYô•ÈøS¨]¢[†Ðþn_I…6%—«axT99K:»Á–ÁúK®›ñÞÁDp‹Ä´ò”6©J8q€’æ×£Ï;ú]3PŽ\|Ú;,¬7E%>»{ÑÝ°’(áØôwAß©Q±[F+½n:cÝÇè”fÕŽ*–ªeËsº4̦ÓX£i=Å`»™ªY-Â=^TƒoøÂ1’9‡§zŠRðÇ à±xsmÓ²'í?7mH®prØ/6“tü°trbuktrJ’EÐü®Ú{‚¦èÂ0’èO¤ÞÇ«~ ;¾Xh¡T_5 ÜŠØw=ð 5Û¬Õì—Qb\ÑaÛÌïƒÕµ%=Ú ™SáÌRpç•Å¸DTªr¼ _ça{Ë÷ ü¨¤~Ç¡Å°ýí\–è‘k8ŽP–¼Y`D¢Ã÷2W¿ ³ú¬Ê¥µžwT#ô:Ò%öÌ£Î"`ùªVÒw'å1¦uÒ˜Ö\ÛM¦Wå„ ‡*=j×ÏZ&úÛÇ¿(\!Ét9ò(CñqWŠXYnÎ13¿Ð‚ýéŸÃ’“†½Jö\û™÷ç²²† ‡-—Þy'ªð-€>[³ˆêÑ¿æîI·M[m|P!N%®KÁhòý±úL`ɇ%Mí“pG5ß¡$y/¶"UŒ5¤'xƺæÃ?uÃç:˜ `¥”–9¿V¹¾V¸AS¾ûܹ’¦ž½=X«Ÿ%KQÔPú­¹|‹Kmëkß[ž¿Œ+6e`°FôŠ87Ÿ‘jg¨ÁzŽ7E¥çK“€*"t­b¦wövƒÖõøx†?¼G*‰º?"ðsFJÇ EcŠUW!“ÓãMB•[¸"®§¬#Þ‰¡Ôftü+Ny_Û¡¸ ¾Õ–÷¹u_P®hêÈA½0™þ™³LO€Ã¨Þqë%?Ôƒµ¢$MeAÑ…W¨·ë‘…ÛãÞ¦!ó-Pìouè>ÃZ›ý³´ªÑ[…sKR²ºá;¬TVZoÛÝçtºË«h’_Øù†.aØDƒˆ“údöÑ<D˜¶ÆéI3:ÈjŠEKé3ïú—ñΪõ=žcKljü,Ùܱ½‚%t±¾AikjâÍ{ÔÕõ'çôo½ h ¿$b^ž,‘> Þ&”—\®ï©a¥à¹IiÇh™_Æyññq”üšSKñ€ó{sMÑzCÈ>=M/Z4ÀŸŸÆœ¿–£±`U>›q¦I'Â>»)ïújÄø8)Fg29èó˜Öô#ËTïú`˜7Wä톣„g@i Á+å œ²FB ‰¶¤¡½j‰ïóXÇi2¹ãXü$ËsûŽ>T6#u;Ж/g+ÆGhe­qjÒaè)ïuÉÉt„ßÌM#*òÁ@‘¾ˆi–9‰ÿX©ú¦éWCfÿ9œrà·Ù2²$X›†|8wÌѾ’† ùE÷žÎoØn_Ö [1.>xÏ¥¾zvéfû^ª±ÂÆ«G®·þ€{“L]ŸG³ûfæÞ=œFÜ+b[ ·°Uó˜ÿ•æï¥Â|NŽìWÙÀYà ¡Œ ‘Ú`F.D¢çÉäÝŽ†E'¨ªýxò@}§;`ñd–‰ÚR&¿nv ‡iWZžÉôuh[Ú¦÷¡¢ÔÚ¸hª¿},uçC¹ä2nª©¨ag²³5 J—%ÇÕªT—ót¸¯ýSœ»ß~§y'¬ãËÜü0ÎÂœ&õÖÝn’ÃâY ln—¹®æì"˜l±Mq}¥ƒ¬·ygyVT#Dé,ue¢ª_g¢4ÿD4à\bxküØ6iÁ…²n•hBÈÅ›ÙCz£<ß9L-ÑAS¹UÜDÌßn{ð±/ðÌ °DŒ{Ý4 ØšÖû}Ê”æw™ÿ@%§Š'ÛɪóÔ¹E/›k6©µÝ“2ÚÇú”™¨«}þäñ2“@€öæ‰}¶rùòyþÙ` “_ÑÉFp:÷øÛ¡¬´ý$ËËr0ŒQ~pÇlížkt¥ Ù¾ykñãÕ1ÎÛ¯¢Ô²‚¨( ;wyOQ½H%Eœ$—gåÝVTëRXf-ÙaåD6cƒ®¬7Ó¸ÅÀƒ~8ž™>Ÿû]­…Þ¥¤Õ΃ÉZ}#Ù™ÕEÓpù\lûWAó0×6Çf¢8ê~ú `)AÚwÛ=NÝó¾8Î 6Nu{T¾l9ŒCƨ^óôäq(ðb…Îà­±ÂJlkgM Rs¿ˆí^›ñ ì8Ûâ\m?Þ§7!‹ÁÈ™ÃF¶³ƒb××?»(Õé‡xÈê¥R4 ñ“[—b¯Ú§„T²ô‡„¨ÒH~¸$“Óu»Yn êh³.+‹cºÞ}Öc{ÏŸï/žìb“<=Es§À" |(b6–*RÞ¬»ÄûX„I·B?Y5¢¡½Dš£ —tf×Ü%ríÚ§}‡ˆÍ;d2E~¤LDnwò¬ÄÜÀ~D;W‡ë(¡ò-×LìL@ûƒ.£j ‰ £€ <Ë?¡/ð¾Cšo9gÕœk'aÜæÍ<'5ŸPßg;Ý$8ŸÊ31%=ýŒõ¼ÃH¿W– `9µÉÛup˜‰]_^„G„¶‘_ŸM!·rÃtüž{ÜL 7È%q8¤M2i°+ªŽÙly /^¡Ugº-ðK³$ƒMÃÈ°×Þ0Ÿ1Âøó&Cwåcþ†Š 2¯ˆUúrÊ&ù|©ÉËü8,=pš2vpàc³‘É4_Û¹7ïüäHý"BxF3Dq(xär>¥Ç¥îÅ ”À5œ×[§ôDZ¢WpæÖ#8Ÿ>“M;m:ÀtªH›°Ì훹,‰Mò¦ãõ¼'å}?Dz’s±>ŒûS :p\—qGtþÁGŸƒÕ~3›$1Ã,E2wÌ^ð±È.H›´‰†èB:Ô.yž72\{ÝÑ,*ºü"a²Ö?I$zØ8¶Ï ››møÄÒÿìÓ’RÔh»K÷W:¼_Á¾ÎLÆ°#ùzÛÆVœXÅk‚½S¢¦Ü=Éâá)ÜDY\–|û{ãå;SìŸH'M÷óeUÔÛ5£ŒDù„YÕºJ©¢¨b7°¶úÙž;z(“ß$³f#7ïs­&ñs5ÞÆì|ý’§€„¯¿MÅUÆo@émÃÕ$3A™†Ï>¬¿Þ×!L¹Aþ.R{{ø©úbúWáTy#{`ÛJm‡wÂ^jY³o½àU`Maé¸À‡o`Fƒ]4zžHñ‹;VåqÜóÔ_5j» X–‚æ2[qy®¢yÛt©ÜîÆU/ºŸE ¦­oã ²_ Ù©9Rγ{_ÅZ¬¾Ã¸ñ®=“,hcHØœöø¨O´ÁjžŸ/tQŒ_/ßšŸ™b|Ùû­‹ R>/im 0ñenØß’±È”ËmŠÈ…<ÿ9|<–”øC/T’ —xïà‡sÇèsjiO4í›úÍò+-‡Œ/kˆõiÆóœ+¿°`3oUì¼)%²å‡Ò²(Škl ¯§“°Èk(s¦ÜôÐ׳Å_˜âsÇš¿Ÿ©ûÑIYî; R ~¥þ—Ä]KÒÕô5;ŽêÔ‘lâ„ ^ö¡±­¯ì">Ö'äZðuÔõðC|;K'ºN©ÃÖûŸ3ŸYÄTDž÷²ËÂ:¹c/-~TN Š¦ü<§¬F¹™ÄÌâ-¿heûLÒ˜Aqô“Ë-\åZíC€ž¹æ³©Ã%‰lʸRUy?ì}êù®2ë&BáJ¨Á/»ÌÞWÉ°ÆX1©ð0jDkÿŒŸrŽeÉO¦“Yˆôc °xÅçÚGTâÉË—6´aehÈLyá?þrŠõmq=|õ‹ØìÎØ}¤P®Èx’QŸRôŽÚãkª„'ÅÕÙOK2ê4‚Ué´Ã­õ©÷Š³ETÙ×\µFö£Öžâ©`Zëï)Y °CÈbe_”%{}‚øñŸpÓoO‡è®w©E7t›ñÆO¼ŽíPìÌ ©°Þ!®’Õ®Æ:Öó¿ýÖóÙêA=<¨MŸªô”Ö–Û•°/óXF‰3Cò´Zµ~_²âgš_ÑîLHŽ—æí UÎOy¥È Ì길b­ÀÔÂÇ"Â~°¬Fî¢IÁ¤œ•~dÅ[ mëVÒÖ|’qÙL[w/Ї VәŠ£&¡’†ÔãY…g1 è|êƒ0´L?lvsWß'³£1ëŠT±y[5h& óÊôôÁzÍZÏ=Zr¶n6Vؽ®åG9¿÷ ›­ÅtÕv Ô nùÒ¦õ†T½ “Ãá(ª,lü@µÖõø‚r£ “˜VØffýaDé9,–¶¯«ðíÉäyóšo»CV ,(¥Ü¡ˆ4•2T×ðG$)õáÆ„ø™êÏÉÅ'ú×Úνý ½¬† ùÕ(ßBEY;"&«.Ž/U:VÄ9¥ýð$~Ù%øð4±{kþ=ŸGãeZÆ·»L3áÔ¦ßC½= ÑÕ[ºe-¬}òrRoª{¨˜ÎÆÉ BM)N¦–¾‘Ý_é=g¼Ò{vgvæ³Âe³ühHù“ñ÷hF#¶ˆAßçGúNwth#këªã–ŠcÐNl[†×ŸDúÊÞyÍå6$2@ ŸUó+›ÆBÆŠ_í[·²6-H`û&#£öˆ·1 .²Hˆ…õnÞM½³w{ÜÞ‹œa;͹¾f´ áÈ !b=9Ó&}ý«”ÏPpMó¶Q«ÅNëKazi}wc‹ß5P@·]Ÿ50NÏí¡š{ÝñÙg¬›m»Ë:áì[“¾ÏC¼ÈT0 lˆuÆnÞ°‰¬/LÉó‰kèâGÄ¿vp_Ûg©äÞµ]¾ÄÛ™éa›#èšù’‹?…ÂÜë±ï÷Ô(–Á”A§vqg³ONéÀÅ#Ó‚¢e¾Åæô&£ãsL_ŸÒ°Tµ[;F–ˆæ6£Žê52Ÿ©9ùš¢ðž°e¬([KâÖ¼À° ™gNýKhØ!—ÖNç7™0˜°ÐÎvN'º`mS®÷¦½j| Kh„S®uœt8w›Kõt<·å¾9püј@t H?‹2ù¼¬lÎâl%Pýbÿö'¯ü…1{ Ð%+Ã}gÑô —·~~åP¯XñÈʼ ú Û‰-ŸjJ³6ÎáêV ˜r¿©4µé|¹9Ôҫ{zLÇܶ²J™œS]-tî„Ä~ðÖ¨m+>¾m±—ÑN çÑË„IS¾Tá ™zŠÒÙt¼„cäÛÔÿͺ’È(¸^K]æn¥“2ʧt<;Ž”†`tܺ²°½—±{®]#`,!³Ó&,OéЭ+©gÐFM¿$ XÁôÁ$JàÀºt¤d6y±/h‰O…äÒ{“y÷Ü»F¨_—P$Þ±•A=iJ 4Õ2Aåâ?©w&ª0öþÜ„ ×ÚJTê»ÈîðCx4ÉÕüuê]åXãm^ÇÝÞWE›-~]›¥¼§ôDÜŠ…n [ºz¡ßåIкgî±Òè‚<áܤç¨_îUÜôh×£Zde榤ä€jû(ƒù¤2¯/ËÒ7uã 9ÞÄðXLNì&•@´åûVÈñn¸¥³uæœu°êuÎË…™~‹‚h27tËØ\+8‘KÄu³KlQ)}”©î2V(úŠ ³-ßòbZÈÙÕÊÚxKl.fŸÖ U·óëÒ Cçó˜5±‹DåæïžRàþQ.éœXDKרýµíêÔKˆØÅdÔOG¾çTuÀ™4„aØË\{î—›)?'\÷ŒH¤,ås>=0âêbûl)ø–%©VËRNõMEå—¥¤õ\ºÓÕÈÊVNõuc ï7N5×nÏCq’&Ú{ò%Ëû8öŸ=™ù<—:á¢BXø¸›øæÇGÛÑ//Íü¡³]¤8k!=y°xZY"ÇÏtÙéJW{íäóž\õ'£…MŸ_äÉÄ(’Øã »»ïî¿×³ï’IJŠ€!eÌŠ½d4Çä ã“ÜqÐNgg0%ToƒöªôëMYYY·x¼AÂ{dÓZö¤¨%ô°˜"¢®œÃù‡xÊåÊa)Ð…}–©S‘+‰ýÖ×]ÒQⶦôXáx(¿€5o#(›tR@ ·ÂឧÚ{ðÜ\¾ð#ly ¼Hô`Èôë²¹NûÞz1í1¿üÖ²žÙWûÉ ¸ß‘­Ðݾâ÷ÆŠ¾dÉT\â*å8öŸ'|.®ZCÆ15¡\Qa9î„YhbT ¹~ò`g³õµÅ›Aݦc´µ²ïeÇâIÄØ7Ã×6Éû(5‰iZš)X˜1~Ϻ¼irïêø`’e³OxØKδßÐá=;£ »U«Bâä±zák£ ·VïoÔ°)! #IÉ{ÑE==È› ²Ì4’=.®ÅRêàz´fBPÔ {mãë„ö£ƒ°#€4Ÿ>Æi¡ÐÝ&[…Ëd6²ÌZñ¾b:qݵÐwU+ÍêÂÜ6Ó.H75“d¹BéºÛÍ@œá†N_f¹%ßµJÌ|QN_>YRÝN ýtcf¹S q*¸AÊš `V0w·ŸMÐ¥Ùz1ÇòýÛ¹¥Rv›b’ÏÍâŠzèê릤z©3¥*3ÉËÞ‰Õ\ÜyŸåöw€š› ¡°g#¤ ]ÈÛ÷óã4ËúáƧƒ½($åñü‘­;„‚š«œõÉÐõÌkøøàå~cæ¬ó+d¿ ƒI@`ÝsäèØUQ¼ë1þ"þÚÓã­òe½³!ÛgÓŸ…j‚ZØM™«eP;§ä`èòa©µœ„ïÃrn@Ž¶éVJŒ|¢˜X½NEã‰Â/P:1†èðcÍ gê"§z»í çéôs ‹ÍÓÛy.¼¨Ÿ§/XŒúî?¯}y9:ï÷X ¥"oRê€;îÔšÚ×Ùpß–â\ÌÇnêD¥RaŸRyhëäê°š”¯¼rÝ'(äp|ËŒ:2oóÐÛ®'²fØŠDË•º¨˜;œTŒÊ²ù¬0Ë ùèéó­ðÿF~Ë4p}œpY{¸íPËYic1¸¯°B\ïaQÕ¼~Ù;Ó+'‰)1Ç•°YéÚŠs¢ïq`í¢þCzQãÂÛªŽÙMU5?Bo§>H¥ÏûîGâv5™/‹È+Ë¿G,,†Ê4oÚ’²º§ét Uªù4Ã-Õ¸Yµ¸ÍòL _ ²I*öÜÄ1”c ®šC ðífú¦Ñã»–ëÐu^â]³g‰©ó¤¸_ÙO¬›?Ô=?¼Ã»bŸiŒ ¢kIL\4 œžL.(møîdQzÕ*¨õ¾.ÐÀæÒÝô"óö§}½ãA*árTܪe2GzQàÑ"cqç;fµŠÄ½rÖƒ¾«ßôÚŽ_›X¿iPZë]ž¼qÇUÙ®‹¸Ùk\ÚPÍ•!X4ûVÓ×4Â#Ó095¶0„»yËŽ‰‚ÔùÈ?{0KˆÕJœ ¹42@Z¥0wñE&ªo-wŸš=²ÕÎ yÔ‡8[0ËRúŽUbÏ‹õ¾è˜†®,N¯Ø:x§¨÷ôG›Ç1^£òž>¦“õ)z¥èØó-Ôļ ujÅ…š¡Û$tN6U³O¨™XãFð#H|ìÆÝÌýy_Et–gUY CS EÍ©zýâ©–›Þ|H…!ÕKo¯å¯l‘Ç8´Xࢤ„M̤[‡XÅ°ª2è<ä#ñnðfLÃ~i`°K¢aðÕÞñŸ?Rênñº‡Pœîõ„¹ àÒɆU6WŽuúX…  8q<ù&늄É(ñ½K°‰s¹ã&’UÄÐXùUƒ Ÿ<òƒ"OÃV°;’Ío]~I.™]°•Pé׳ã@A²º·®œ3PÒ%Ã/a‘qjÏ–ðÆ—¤%t1grÚÇÐÔ0¹‹|²o½l#–I;c迲f­‘íÐÙź'cY)lÖLc8ƒúƪ¢Ây§Dð rNuÙ|‚#À¢ûeÑŪ¼8áUçw¥UrÃû*v6xª¿åjÊBjI¨WXÛS óòV¢ c´} ÕåX·çd„+®?Ñ;É.^E¶t¶ònü=u–¢f¸,žY™ˆŸ, 3¾J½FIý"ùÚ/NgÍAõcõ–‹²Äø ì2uËÊH þyEªÒO¶ß26Îo®¶è .ŽkL,½<ÂøU2¾3ÃV*«xÏ»—«ÀcÚÄ Òí.t«,2˜~‹ÏaIÔ(؇ýq%ö à]ߢ)žÚ6µç_ˆï¾Î«Ã«ã´ê-Bž>—äðÕH¼-ìتµÊ‹0i½Tè`{w.ÒÚ7_b¶§}Êò=T£„(.U ü†‡hæ+¯ª0ºøzo’x[—V#~g›Ë‰L?ʳ½ždj{‘‹}¾"@"F)f[³‘fÚCh%Û«7YÆÙCbt"Àõ@Æé_o!ý­ÂžÎ#›S+ÐKÜ{g¾†õâáO7 ¶âŠ×ËG%¯½FPŽø »Yá'•SÇ/‘ô©^À‰wøTñ¡-ßä-rWãòý/¶„%J¦,š=Àœhžp¾>Πo6±8}Fb7„Ú vúáh)TÌwuGçGœ\Çz-pIj_ñSziR‹Ö¹YبÛ"©/tñZ" ˆt˜¦Á¦%Óc"äOe6õn#%­³F—ül‚Y‡½Í#qÛo0‰'{ˆ}Lc—†i™°"2¼‚(s¦0¬ÓJÊÈÐúKŠ€Æ1µè[Ž!yHÁAXFY°V-$~¹`&RKö½šÕƒr•Ç´(´i}Ü ÕÝÃTm>,=6w•äüïf¢½žÐêT?ï ÀÃŒBŸT O³}—ñÔKd7#–Ã]ÚœMÄ-_0ÔxNS—Ž]PªI‡ì#‘e$ö‹Rr ?H2û•î#& yÐØ ¸¼…¿$›u^Ú1õ!å^oë˜Ä’æù¡‹·›B$— æœÜ!;¢ä°§IlIŸÐ»†È6)CúCæ}{Ã^2àH±OíBŠ;ÇdøFzÑY8=4dVHÕ<+:;ucÈæuÞ.…Ï?É¢ôôg‘U®L<9ºi<Ù &¶¾Ô²ýºüµÿÂirWèæ§ ûÈÿQöÎ: endstream endobj 846 0 obj << /Length1 2222 /Length2 9694 /Length3 0 /Length 10907 /Filter /FlateDecode >> stream xÚ}vuXÔ[×6 Ò-Ò1t3tƒRÒ-ÍCÃC·ÒÝÝ ]Ò%Ý-HÒ Ò"ßxÎó<Ï{]ß_3÷½Ö^÷Zk¯½÷–RYUÌb ~ ±ƒ²r°±  nP(€‹•–V ‚ZBì$AP° €ƒÈÁ ädgg€ 0$û£¬sÓTpv™ÿö`çäpòñqs£r²L-M c°¹¥*ðw|Y;3€ƒ÷oÞÔÙþ¿6°£LÀeÀ$M!v6îS°l³"È `ø+¯ÿcÙZÚ¸ÿ´À–æPƒØÔÒÙö­²P¥‰˜¹ À*ÀÆÍÏ.ð·ÁÒéµ¥ØTÙjb€::ƒÿ¦5ìLÁŽ6–v`eˆ“åïÆX9`Mø·MÝÂÒÄÚìäàùIµ4e¨ùŸBau²¤•åÿ7°-¬ƒÿr2ÛAüÿŦ࿠Îÿ`;Óÿm¬Çµ(-+%£ªÁüWGXÕl@vPC^¾¿½¤ìL ¦–væN^ÈÑäŽÊà€!€'ÀV…ì+Èf–ì¡Þ3ˆ#êï=ãåÅ~S#^Püñ€ˆ”|D Ô?ˆ|ýˆ8`)?"NPæq€²¦.÷ˆ`êò¦®ðˆ`ꊦ®ô⇩+?"˜žê#‚é©="nPýÁÔ5L]óÁÔµLýí#‚©kÿƒ`ž Ìdk› lCÿñéÂÀÒÒÄÒÑÄÙöžƒ“÷?¨¥)øž›ó7 ;Z:Y?…6þc)¬2cG Ølýƒæùíø×Iùg1Çß´5ú/®øÿY+ÒäËçïÔÍl`óô_––‘ ÄvþIœû7ckûØß øØ .XÁ¦ã°Ük‡1ü¯TxÛœA6Q`YÆÚôG+Ùa9š=Fy˜Yºüö·âü§,ÌÅüQf7ÿ}?‚ÿt•cñX¬µîö`»?<`œå–¼Õ6Ö@X¿‹à…5Ææ÷½òh‡õüŠ8`R<°Xv°ûè± 0i;g[ãß·ù)qÀÚyLòÇ*X¡öf˜†=Èl÷¯©àæøûï™à‚¥í¿î­\yÿâ,!{Ì k¬½ó…qÀ‡ÇÂ#g°Ó_7î?±¹“(ØÔø±E°×åòßcÎ÷høwŽ°@l!¬K™ðüF`—?væîdù8ÓÜ°ÈN6 '‹?BÀ²xàõjáþc´`=„ºBþX‹áü„m»Ë¶#®XØj·? ,¼û¶[ÉÁ"y€ÿ–ú÷k¡ ²´ƒª»ÛƒìÏÇ_˜ão¬u„Xƒµ,M¡º(€`ßM—v Ù¿Ù?.ÿýÿû,‰‹CÜ`oÛ±™%ìmƒÝÀ&¨_g!&BV5mWÕb$®¬[cˆ+kí¡1Ú}!Wït½Ô³ûAjâˆ2&žq’ã -/÷²zʨÞ$ìÝPÂÔÖ°‘Ú#Ç®R=*ƒÒwUÁ=H€œß òƒö¨Î¯k“Æ–¿­£W]éÊù;t_uø¿`Lžr4øîOÁQømd•[ â;?²–îç¾LH“ƒ»ÓôF‰K/NÈìÛùd ï¥x@¿d«Á¹Ï÷Y›‚4éÖ°œ ÝW ìk¬cÂ$LÚŸ€lBP.Í<_ïÄÁë‚ ‘óVq+øûddj¥> “ï–®où|3#iXµ]M_›còß…^ä.^@n¸_b^£{¢_„ÇâvR« kEU½0­òšFo 7¹÷ãÙvþ †¾½Uko ^ìï}¬}饶å€ØvK¡´s;ÌsŒËœŒ¨ºŒ«ò2#ê냻öÛV UEá”Tå8•ùq°Ïñ•4ƒŠñSÙ¹(æÌÓïý¥wz®ñM1Øq˜iÔÝ-û'QÉ;GŒöeèŠ7JSã.qç¨FÝj“ØP¹±Àø”pÒ‚*Pht’¾ÖE\ùAËóŸÝ¥DÉr­‹HŸfŠd‰šfëβž‡€éMt¾£-û9X!~µ“¿#šYC½åÞ8±ÛáN|+:e’x%,X%Xö «D(eIŠŽ¿¡œ _~§Ù‘HŒ 3ëy‹†¢gz§XÍdþD„i¿Ûk³É·Ø|\Ï4Oïö.áÖ†ëÆ‘ŒÔ5ju®ÆÂn‡-Ý7H’[ŸBœFRÈé-p!Ӳɮ—朜ítÝUÚ¬Þȉ2\¥n\Ì"€f áëE"Ò/.ÍÇ´yÐÖ`¨q¿†o/á Óež›µ‹90ˆ9A?µ0´¥ðf…ßn±â0ªZBé¹ òâáîð·%ÊYÄz2ö³¯q”yœi‹!¹x¿‰ë{U Bpœn’å+Ù$iÖKËðp”-¥íÕŒ0k¶ÙB3±SÉ6–fÑØ“Ä6¹Ÿì:.‡Toô#o^ïo—íf½27«(ˆ4}¯j\~Ÿ Š >»O]j6ìê0DâÄFHÌÄëÜï™o?ü žöÀøI®xŸ_z¢—ò¾!–ÙÔÃ9÷gæR¥‡,ªk5¶¬Âbñfû\ï7tÛmžbPó¶Êd'º:XÀÕ6¡:ï96+lŠ½0˜n ¸úxž\®)Ö9g˜ßc2áANõËQN ØÙc ov•"ƒ¯Ê;#0÷K&•î …3»|Sý)õZy:áN¿ï¡>Ú„[›[X¤¿ãÍî£ÿTä5ÏÓ¾Å/ŠAîPKbânQ—Dòwe1ýÙãÓ¡éþE}£ô >VÉIÝÅœLÉ@~ôÝÏw×ö }™ëi⪗K¯«þ°Ý"…äª!p¡ˆÇ¾È°¸ñ0.É"e‘‡¡M?n)o&ñtá*ƉîÑìרBÐW è#S”#Äìï¾&ý·M+_†`¥Fljk9ø³º]WÊ kžãû¡ê6z¹YòwÞäžRïô[UÍŸÝsŸJZ·ÐPXÌñYÐ]ÂJÞÍRïhV\ˆP¯TWïŠT t÷ ÐûäXÓɽÛRzM!•5›µ±óc¾æU« ± r³ÙµØP»Â6û—·öOÂ!•2\ÍW±ù-ä3?ÂvZ¬Ùó]°¶ßî½ÿÓÜ…Ÿ%ncë«&Æù5øI %Ùûª ¸¨,_Ü`ÿ8Ûcå1ò«-Òââƒi¤Î³uâã¦ÌQ ü:Ÿ­±í¢'¯Ñz·år-qN JŠÞÊë¶DÆ¡óx6z^ó#‘ݯ·¥À w±q¾‹œ“h³"–~¦•—ŒSÛÛ,|Á|•Fñ¢ëyºÂ¬¤þþN¹ÂÄX l“U¦§z|µAã;¾ÈŽXÁOºVåámŸW‰)'ˆoïf£Â8û‡ÕJ™ÕgE7Ö¡õCðp†;ì A)Ÿ°&çfâyK¼”Î}L~Õw_00Ѹ ××õݺ]ï¿GrŸ‰¤;“¶•NÌcÃ_ºXeµoÓHôF8Oo8ÃÍB^Ðŵ[’«5\Ži?9Ò@£HÛ@4­(pZ§ O<®i? ®i#hñ‹’|æw [pѳìB¦™«*h¹}·om¾N²~ºè5 ™A\åB Ø´'P¡kÞºˆ ¯{†¦'ªÚ/?^À{mô f›’Àµbr+¨Sä‹yÔfFÁŸJ•î{×bí ï{x‘0Þ_’W‡¿£k²[ªÕ˜…ðŒ×äçÕ puñìc´¹â¼º3~®#þùh¨Øž{ ?¯UwEO6€<çÚžE_.'û…®¾Ò TEý‡ ÕˆŒúÚúÏiž2jN'V_±÷^åœ5QÍÑ¥ä¬Ü´q21½4Æœ ©CÉÛÂù¥RY5*4#+g«†=Xx|3ÛU¤VGqÔ_ø“š|¦ðeLLí.J$`{ø‹ÜnAFKÈJ¤¡iæý„o´ûÕ2¢³[Ý' ÁIx„"Gø«[º#¹ÆµëŽ½õMÿÄÑK-¤±Ÿ9+¸‰„9/Ǿ˧§|ÛÁG¿a…?óΣ üzA(Ã2ƒj³'‹è,| ¯;CF£Óâس`6Òo'Ò㣔À'ª^¡rÙ“3J‘|dØÐn2½î…O8Ïqâ!‡@@vøküd½ /3W¨'àé-/Õ°O¸Ðb÷5ä¾9ÜM ³‡Õ=gV~¿Puýün—ÍÔl¡þöë"ƒ'9¸|2­o>×^àœ±@Ò…‘V4Z“jÑ,-y’v*üçV˜¶;Ñlñ@ºß‡×?3Áµ7ju`éÛ/Å ÆïRà¾:¢Œ¸`‡ª%Ê´ûä…Œhd[{-´…¨èÞz)]Å%>¹ÔVÎÝóZÀ9¼8†¹`2PBoö}µ§ W æÏÏ’îõ ÜP#J´Õ¦ïWç6À¯Î;àßöq b‚™/FÌ—Æž/:$£äÝaùg%3'zr«…TÝôÔ14ëåÐ1ÇMÕ4öetC&Df;?/—‰§,æB”ºáxêÐHø›dWHJ{‘î¨K³Éæ£.ôêMΙ8ç€AõˆA¡Ì’}½€fÖ2êŧœ_|˜Çm3¯FäH³.õ”žã £j˜þhëR‘Mzm¿¸ò¥I‹ÿ.e]”N­ùúÜ_ž¸¹3*2yHW¡ÁöÃÓüfÕ¥'­«.’ïð Jƒ§¦äTx«e¬Bb™"&äháíQ6>½zÒ|ds}?=X3qMœßxa`é‰~Šó3ìA³ù›j’Wåtû¨RüŠa¾âõGtÝÆ€¶4­Žö˜±Â¸h¸ jbo¸=pE|û¿è’­¬;'Z={u¶]€Ì@÷åiñà8À’뮘„Ž›ôµìü/&-bƒÉôá÷…ÌáQ‘œI³vÄ­o7ý*2’Ú?x@ÐzÙVW q‚k,•xú°¸dŠ¸÷¸_>•k­éˆG'ìËËŽÍçu—Êt:¾ÚXõÕÄcû i¬TI7åò[c”<Ç~E£ ÇŒ~¡Æ´kÿ¦ìì Ù°êû¥æ®¯³'’H©ƒ„K7GlKå´;Ô¿„gÈ#ˆ±.ååt{KÑ­ðSë±[›÷f³é*éÛ/)^)?L¨ñ%‹Š!=ÿÉJ½ò)ÆH•ŸcoÚA®©f¹bIZ¯ªÛˈÇúBØ;ø-«…Ù³ C¿/Ñ]q|!ƒ÷t¶Êtsç„E†Áø‘úh=_V‚@ý.üû7+›o.¢:Êx÷(êÂ'[Þ4y›*¦ò=gmGŽÀŒÔʤc jMcWUBŠ™ Ú^)´77a·X/Œ%éKÞJIéÉ*VWO*¶a=àå´1M„ jµö¾o¡Ó¢®ßî“dMKUi#{ÙfI໧€G˜ný–ú­ DÌ%¯¬/±°/ôÁ!gì©¥\®Õâßô/[έ~þšï÷¬²§º&ï¬ä¶)süž„zOª\Û0zlÝŒ’[WÐp}üB¼MÒ))'îH°ÈÍ÷9%žmZÜ%6T&(ö˜’òy¸fÿP…?°þÌ__û€ñ¬SZªªJvê—è~YSv+ýiǶ—›ø§Ë³ùÁ²I‰·íqáÃ8zºl»2Âcpþ÷L@YÃõl väœÇ«è¹ ×;xäâqVhayq$þA„4ÂU¶ê4£Î ð8(2)/"'q™íDúõ;¤ðÎ-NòÜÎλۄí©Pßùk5#*­=ÔžÓ©¥¼z¡ÚÌè»^Ÿg”;mé[Y½ýÊ–ëä hþZå_žÎîú˜çr’êÃsÒè³wçÑ©Ièm¼5vNYkŽ†}ÅHPyhbFW+ÞWXla¤¸´Ý&·ivoµÐ…ÒÎ,·¦,&_)ƒ [NS<5 U!ë²g’eÛ/øU¼}§Ž—t&N}‹ãëÓÍ+†¶mo¡RDæ3m¿ü(N!åLÈUþQ/ùƒý¸è¸Ÿã÷¨ÈqkG^ÍM{ͽÌÜQ\ÈußJsD ŽÝ\¿û‰‹§ìo]sФï°8?ðàÅà´M}ªå5Ï6õD>I¸@n¼ŠÓl³™u4kÚé9™9ËòLÏ®Eë…*M™FøÊ|‚ôµ¼×è†+hó”ßؤoEˆ¶yç!Þrç"Gôt­ «Ÿ ã¶}{/ûÞƒl²[«áœšý½DI$6¢®e¿ù¾&è硱lþŒÜXZ–o®~ó²!ntó9+¢+•‘“¤}¯ÛûçýRS6iåêÖilGš’+†¶ðÔ®Ô¯ÐðÑóÖ]¾òàqSQsJ΂)gì_¯ †±#obÔŸU|7a¾‡[é¢~ÚÿJ”ÞK>d¶¾„—(kP–úü&9uÉ¥Ú+E“ ìÉwëÛŸŠü¤Ž–§WEV¼ž5ûƒ@{Q#ÄÞ v™…8}‘`¤ÿ5Æt¨}˜%úúO¡i•`Š™1Ütá ­‡›­ôžë¤ìï´ˆ1‰i~êû¹5¯ Zž‹SÌå,+¸'.ÖçA]±C‹O¡ÉÚç¾ÀOäCÅ=füíÔˆFÎx×L¦2ê@WÌ'5Ëù$Ê~+ÖÁ‘ÕfLg•Ã¬Ca­aëÓA2ä[ ÓÈe3jà¾çH²•#=Wä²5]-:܃XåáŽä ˜imS¥‚¸TõÄ-)Ò#£w ¿³KÁU~Ï°YÔ3^пÿ†©5Aú>kŸdß±0ÔW3©ëª+÷C\;G)Á3 §’pæPGq&®Ë²{¡nþÿ²ìH¼Ïðû “©'¢ï‹’\¸éŽMÅ zvµZ&¸Ñqu"¼ü “:ÈÏ"n¥~ýàê%àõK«RêëìúÅ1éò—ò÷=Ù˱Þú w´07â]VHö£‘cî‰7…â´SA  ©jd›•eGÆSÑl´!e'kISjÎþÛáS¯PbWÙYCmEØ…vL=IѲËF‹ý™žËë{n_mÔ¹o“Yb‹åÚ‘GÒ<ŠSqß©»+¤-ÍùÏ4M{Ñéž2 ¶Wc囉–ôšXë(7sT«-~; MŸù‰!ËñnøƒÃ‚„èí îç—ý®¶äá›ðIÉÞJŒEt£îÏêl"WvýšÌ©Óê—±žÑ'u¾åÅæ\z!"xɳ¿›Ì’þ£9³+닆*JN|2|{™9ÑÔÒ‘‚áElV}BèGF¨'_ ¯÷¯„üÿÊ×€÷µ&³èM™?ËrÎ?B X›_Œá@„Ühˆf”2¸I#kËR‰Ó(1ÏÆû‚Y”˜°ðàcùÚ=…^°’¶‰ŽM¡G,gÖ¯VÞàãéZ_f¯„*(êÒÀÉ’‹Ò­‘qUÆù„ãðp L~×e8Ò4ÇÆv×]sÒ ïëÆœc^6„0rD>ŽÌ/Ú‚üB{RA÷Y•þ§6Ò›fúT5$Œ2ue‹ÀZÜyÚÂñ @ÆÿusŽo´ð‰ÅË…-Çè‘öµÅìbÞõ¯Æ#‘FûÈ›à´}Àµ=+Ù~{¢í†TÛ”ˆbotŽOçY¤F&]FÀX"‘KéÝŠÓÕaØWacTð(êb!§à^2Úa˾µ+bØØHÔÌÒ˜ á+¨Ú 7jvŠ´eߪt’vÊæV‹”‘óåÞ'3á[wn@üs~ƒ.°´ïs˜~ž€ ÖÁH»MEÚ‘Š›Î“.,£¢ËfyáûøIšZÿF?¶öȘ1³‰š3F¢Eì¿~ZLЗBpÂÆf•0/d0@Ê܉€±vÉ7Â…£4øYH€äÙ;Æ8ƒ@_C‚!¤b´‹¢q÷˜{Ý~k9†™áÕ-ߤ·î¢4rìÜ%zûyÍîßàH¨Pq¨ã÷Fƒ>á ,ZN=èÜ_©¹:œ¹Ý§ºÛI9ÆvdpÖ”Û]F±lÙ·u–6¨n*X1’_mæ—º‹ƒä.nZjøÉ0)ê_8 äßѱŠ›dí\ò®/b còQÕ)ý„|¦œ}žÛÛÙÙ^½EÆ¿¥žvWŸŠ8¯öìGª ½k`Xæ·Ÿ¾d«8¿,ßd N,˜;ÔUõÔcW€·Ò«˜˜–Âê_Š9+ŠÄ^b2¬ÚM£W&µé·w½5!Ôÿ‚wÎÝ…@[Îu_ñšƒ´Ñ%&ý_p©’r7wß>˜¿Lùhø@˜ü˜¤‘l»›¹™ë;më)ŠÄí‹‘!:*<¾>ÓEÄyùÐ¥|A¥nZå SÛN9êû½"jµ‘Ñfl1xÑéý&ÔdÿhU,~åÃN;šs€LUøD×}P*á$„ kE:†6¢‘@±»<‘þDesEü«QÜ’PªUíé‹2ôÛF®íùŠ:Py«Ã,~­œBrÆ*ògîJHã¢$:£ð¼EùLJŠœ¯ÅÖ WŒ³kÉCéµB›PS‘'…Ìf¸2æ× Ù+A>)óJg|qUÌ;bc)ÛÉ^è?Psi o¥Oõ÷Þ ÊÁgí‚99ý.}oÚ>„! ÙÚeÞ¼-¤À´!ŽŽV›;B†P‰‘-ô s(ºŒõÇt7Tttå•5ÃE(Åò[s4j"Ñ€4•q" ß‹dž>tblu€vÝgZ¦^ªÄ{|;¶F(¿ç©ˆ¥šJÌ› àæô¨I²ÿœ! ^¯µÙC[%‡ê*/Ò"v4"‹.õ‘Eh,_q5¬¥ùêбÔÇ !~⊿E±aŠÇò ìcÍÝ‘Š¥ÊÈ«òÛkT+:{5öÑ·Ól¨=r‚°hÁ³o;jú®Ä,ýÓ©[¯;Ö¨ˆ7ï.“<ÏÚYškM¿Ý°”9¦quù™ËLj5ðE–jI&RlÍ8ç,k1:ºpj¿f=Ï%µøŠ0˜üƒ!Ð!Í¡ Ó^fy,í”·a±±AÄHÁà^?$p=ÞL^Â×ñÃU¿nˆ]îZCRÊúæÁóûwï\M½µå¼—Cj}Elô1Ï;‡‚C·º2[8ªÂ*?<.B.Ž~©|$gË‘Û1Ô’Jr ͦ2 k >F¸ÚQ+è»9ÏúÑËÆTü=oÝ‘‡;Iî‹màZ+¢®¹ÏOÚ^Ún/Jæ—µ;§ )º{n/—PX?„”¦€ßìnväÌá7=ô ‰·)†ì0ö¿Oi/b©I`Û |9ŸUöNjN"Ca%ÀØìA]y˜1±ò2•ð åñ¦VoËSÉ å¶:»WíÝ{?e¹"!‘» Þ§]¤v^@Œ ÅÓå7T•ªSSBho¯¶5ÖÍà¶é¹=¢ÏïǦ§‰/Oõ[ÖðÙÛ u:¥A ˆí ñÆœãÙŽEßrB£!…‹z¯W‚ð‡’ù9+Z•^ª•¶l|‰Èr³y,ÕB7èœÑ‹&UÔgÒ±$µbH;ìËe”ÏV'Å(cE?Æâ—¢üÌu  8И.­›\ äôÞŸ¡=,ºòôê†\£±™*ð®T}Õ¬âº"æSEbQ¹¦æ8lEý `6 éw+Z½é„ç°p’È»äÞ‹ËJ†—£ãw+ÑÏz«"~'MWÝg÷U"åœP›%)$Ó½7ž¤¹N€‘áö}fq‰©Ö9ËlÒ„éÅ«,ñõqœPUZavñˆ™ uMŠa÷x÷¾Xcåã užFc;úñ,׊ïVn"2aï^;,b¯0p¤j°Ü(l?s~£LΡ–÷[Ö%=»Ã±'ÏðÄJaPx'ˆývl–ùدàÉMùjÚÅî^çÁ‰wP„¶c©H¨Vi üeÝ,2ê*™øŠ}ùWw§£”’ šæ¹PPáby_D0î¦ùÛƒÂù¸5kNºd,xÅó-=-Fe«Ñ‹a'æ(³}«ºì¥E.PÂØZªH•Ù7p§—¢ÎSĔ߮dåëãÍèûóøØcÙ G ÌÖàpÆ<5­P£¸w§+J bŽ1†¯Ì­Y\øV ´ žrøK€ó⨛¢ÚT–fJ[fØËN¢´1Ù¾Èc›Æþ2µ,¯Á .àÆÒäy—Þí¿£Æ“ùœõr< øVˆB¼¬†Õl9_˜íg½0áH^{xY姴šV¾Pñ£ÑÁ—ë³²%?#Ž9¯É¾BÚ³˜¼[-fÖ:b´³MÐiJ78_72+^õä kå ]èô'œî ¹Î} ÎïçcKi€n¼b›:vཕk¸VL!´™J$æšé¤öÏÖ™–Kuík¸¹?X0’D+D4O™ÁÚFñ˜ÈGg’çZ=ó\Jh¶ûæ;Ëõ--ýËã`?¼œˆ¤\•OÍyц–äàÓOÖüÂÙL‘žó‚©|+ûÉ懦㺂–êÄi®KÒ(ß™¤!Ýa­¤g_W-&"fÇ8IUèß°»xÖíÅ,GÃ¥Ò^k(0aȦg3*†­„~Q*™|æ>g¼~kÆØ&®&™Ey5Æf˜ ÚPh¯½“à=®²bNùq2#2u0â4Äg–OWb~=W.ã ×à›ÙµDÚ`5>,ËÊV:yÓ$ÏrÓ(™aϬ”¥ÈQ™nVÙ¿| 9%3œÛÉV–èD³H:—7ðoSÓ§ºúP|x‰‹¹•UnPoܱV<ýé´þäéPÌx5Ó€Ò¡•Å3–Ñ• ùmû;Ý”Ï –ÙËSìv5yš»7ø§>ÇYCÎ!r6Z8 žµFÈrm¼ks÷%-[=&,d¡Nxq„«Do΋|_÷óM#P•s‰W¹ÝYUë >±f{¸/Z§àÈ×Ñ‹oÈD&yÛ•†(÷›fi½(Yti-¿>ŒœW® ªI¯ú†Ÿ÷ ,šRóUãZÚW½j¸áAœ¿1>Ë«ÔQùágñ(f¶’¼¿‘_-°eÓÁÂÅn¿"\±yÜ2ƒ£? TÉn*vw.œé*OX¾PÅrýÆö¡ªMâT·*í­¶y—t<í4K}hÓ‹Î(¨ôl:®rCY®ö(Sõî8"#WÝ”•è;O³PÞk¥ac¼d-ÙņuŸ z ÙÞWø¶ÕÊw¬«M¸ÑçDÀšüîO(¹]¦QÖ]"áú×.*¡¤ˆö.Œ$ ÐzPZåû‰Sç ,%‡c¸¦º¤ƒiæ¤C…N‡#Üy{ÍQÌ.‚;Ñúp9Õ8Ë-räDBWÜ3îÊK­ÆFà=ì«wQ·ÃOƒ×s±—í­»ìN%:FƒŒš* cõE?éyÔµ!ÖV€O©§¡ïâBñçU ‰—¯èÉ-ÖE¨µ3¢äÉ×*m²M¡âß8]%$n8Shš®üúl0—`;Ö°i…%þP#Šôî5<¡³wY…êOe¬…82´¡‹Ö&ٖгʷ©åKŠ lZG×”–]­äzÅtGäT>Þ&aÉ€ÀZ­Èœù÷L‚û´a¥m_{AÚŸiÊö™Ž?nò%I·¥/6ï¤5šÙT}µ¦ÑðÁ½¢77ð³=Ãàº`˘ÉÇñÂé#誜öv…Žd®S~¿q: <ÀþRRh\ßvh_ì5 x†ûüyÍÜ’Ùýd šûaõ3ù xÛô­ùw8¹J<2¯iùýúqõWl:9¡§mέû¤yø´ßã;åºÒD;û{̲Ù+JƃçÉ&NÎŒIèóêús sqG¸Žõ¢§Cûž2F²3¥¥kÒa`8« dë¾JÛঃ¾y>+6*B"<Ӭ̓ñì;íþùÃe'X-`£çáj_qéÚæH‰3> stream xÚuSy<”kVt:æLB*‘¼öfaŒ½ìÉ’Â(“Øy—1Ã,ŽÉIrÊG¶R–‘-EŽí³;¨¬•p²d—HTŽï¥:Íéü¾ßóÏs_×ýÜ×½= 2¶vª††hÆ ³U1jh€íúêjh„‚‚1$±!Ý„Äu …Ñ@aÑh4@bÖ|±µ†ÝÖ‰ºîÆjXŒ†–:‹(™ ¸TˆŽ@­‡· »3ŒægœÂñùÊùL,ì‡E•X’ Ӹt‡ßqh4’7ì_Oë_,É¢qÿ?ïBT6°ß¤@ïïY 6‰‘ éT ?CË ò)¶›ì¸“h,ð3î@§€LDm,h½#€*þŽ³÷€È^tÅp_("ù¾T¨˜ÛZ}Ÿ è ·îk<‹ ÒÙ€ÖW›~°_Nù¾ ps7š€27¶qt09¸1 ʔNfP :Àâ4“Iâ"жp@€àÔýÐ.¥Fg°á'€‡}pg0ëÒÄ(7&‰ Ân0±á´¾`4ÐíÆþ×DóàÞåÛM‚íó7®÷!1Aú:nQ©ßBið0¼š<8OJZ<8¯‚úœ¹¾ÿ”ÀñR¼x^‚WD›—àQÁ`àB˜$x­I47蛄œ‹ãí½ñ…(ˇFâ~#Ñ<$,ó9ÎÕ–ÑÙö\Ÿ¿×s}Ð6æ³mÇf2¼@GˆÂöàu±&±™?­†Fc`>_oÎÿ^##†€*PÅjÃÄh¢µ<{öŽd®›½ñqáõûj»CðÏW$#z»dÝ ž÷/T.Jþ¢:ÚúÃÀPUØåS¡8¶âóÔ«­7Ì»|\)ùâ®ÙöÜ éìþUòÀìr›ÿéÚ¤ó/ãÜdó™6½QMkYY¨¥‚û( j;þZØ°ÈÜ•ãRÛ]‰D¥«ŽÝŸ.âëÔ_Œe¬V±^;iOÞÀí%a™xIv¤aJÕÔæðS2òC쇧N G‰2‰ { ‡öÕÍðr«b“±í§C4 ²îD„›elTòè=üÄyNSS¢Å*ºÿ€º}F•.WWÍÇ¥ÚôØŸñÖˆÕaív’+XVµôÏB2œFË}ËWƒÒã‚ÇgâñBÅRw«V´ÿqydÀùôÇèöå;®S×vÄîÉ] 8Pæ´µGÔ¢?”mŽ~¶åî¢ß¦‚ù¡k™ÜÅ£ö‹î„µúé΂ղ Ù…’µ)c‡ÛçFü–‚˜‰EÖ þÔ–¶û›²ÁùÅéšRŒ!z€8°÷ÁaÞUx-nfÉÒº×v1ê5›yÿù„ýöµÛ53m­ò{ÌF”çYð¿>(=)1\—cÖ˜«ß[_•öZÀLó½]MvV"yÄ3úÎô®÷§¥üÿˆûóùÍFŠX¿}úÛ¤¤²MfÇÓþBÔ…í…®WšËùKT^àN~ªnsmÕýHÞ,vº¡Ú¬| Ç.0íé”a£¯æÆ`Ó›ìœócüیޮЉjÔ l)£x,Šó`²ÍÑMD‰øyC'–âù{[‹„ŠÔüqó%¡ü&#¯½²) çäd>ÃîÐÅ7/®ÐÃj½8*óá1íO¶ì™½¼ ¾”¯¢7K%„ÿlìu½Q×$¾ÍŠ;‹{eW¹pk§ÔM‹Û®Á4vV†›«<Ÿ\è”REZ¼îr(p";)­ÅgG±Nsïjíõ-oË"¨^Çr]cªº\ôdh½1c(Ö»¨d÷ÊÒ½>é1s¼ žU’»²3·}Jᦋ'‹ë®ë6D·ÊÿØ°¬øYSy²+ÐßbTÖ¤\šœz®ª@ ì,­›¾JÙU¥¤èˆµélh8ÛóÐegÜÿ&±ÚÀä©…¾]™¿'óI ½ººìà"gKV¾n…vË$®}ГG;ÍnV2WßTÛ•ûÞÛŽÓo´ §9(²ry{bÆænŸÕjX÷úBý7Ây1Úï4Þ¾û°DÌw°“ɥ̿êÉê§Ç^‹ýÐ%Š‹Kêê¯Bš"Ϙ‹ê®XÅ;'”/Éfn}Û`ª—5VÈ<™Ï=WÒ¬Ê$¿ŒÙ7C$ŽR rNIpMˆÛDJ©PToI/ûµô~i‡ƒB†“RÇ1dô‰Äb¯‰“¡½vÝ—+ÚåiÏG_ïõJrÄW*’Æåý6Â4wETŠ|ø4_Ú¸·Qi™Åð´ ^îX‹˜˜åQ33A÷"’q8™Ûtäš^ž§;¨u _6þtÃoîÐÐ2ë®}þfõ ¸ó„û’  ½•I¹71 †›ouõ!{¶E¾'Xœo»i•Ð»d8lF—s /ª“hÍÍ2¶É³Èš7ÙóZîEáÒ40~áCÓ_‹q¹sÕàa`÷£½(Ê>AáÄšðW÷[ÉÇ qÓú¹Ê¡ÕØTA®4^ãñhÙ-¡­鎆êr•Á¶—¥zN=ËÜ¢Ú¯œ~¶1±@) ›>×ý|çÉ›g0(|gR(ª¸l]‘ŒˆO'¸xtë ôLëJUXÝ[mþ8þ‹l5ÒòG]›P—¨Ô”àZ 鿬iyû"8¶IÆJÝ>9OþlLª=ÛÕŠ4?p—k>™ó€hò¤"ËÖ×ÃÒgLmª´Ú•Wqè]ȯ•{»¿ÑÈò·ÄΡêÅ'eÛ† JÊ轚&°0¸&5(Æ ÷¥fb)U½i2˜í±rÉì«ÆlÔãÞñÂ" Àd1³Sçeä> stream xÚu“yÏ÷(í 8išS™GÀ½LÑÄjaŒ$èha„””,Y ˜ð2Xœ6VW‡Á`2‚FÆè‚`¦´ç°É´å NÀéâŒô…p€ QàHƒ`!íåã­ao&€ÕÿÆ©¿U-d±ÑKUôR5½’Ê„éAôF÷qèt<™ªËeýG%3 zÐÿ×]@ˆæƒªö â0~V­2¢˜Ã4:`¾!ˆ½ ©¡øÞd:üÆa*È¢C0H`²¡eGM,ó“Fô(¾0Èfzß%<(hm¿wˆ6¨ì#Øý\ È@­[=̦€0®ÆTðÀ} LýÙÔÜ´÷¹š»Øª¯ŒpE²‚)L*Óœž>@f±ÈAB‹Fz@€ÐÒ0mV[ f"èÀƒ¼™,¡å áôm²Ÿ‹úsÈôeq…ë ˜€,ˆíË #>?8úí#,2¤ƒÞÈ?°ÎwÌZÈ*G}ÒF3Ëí­2]#@‚½!B‚~@4Áö€^ÃàÐÈþ-é߆ÈŒƒü~ÌuÙ¡•û-vBXL_Т¢µÿ#ÅžŒ° @F ƒÁ¢ýVW‡ÿ븅30D‡4tuÑßgêëÿW"…Ãb¡Í­¼xtn«±7„>9 )BmLŠÉ©£E§Jç ÍeŽi5®ïé+;ëV©‡(·§‡øØñ'ïûÓß‹š¿µá ›s¦µ-Tf×Ä\S {åõˆîËûÓ5VµSšõóÓ•4PŽÊÜFk6¸rn`ÓdœãvQ/ÒZ’J¼KÛÂiƒ*á‚·™‹eìáCF£ÈšÛqY‰6O-å½à¶C±yææ.#Î"]å­›yOì\´–Ihv?£ojtoÿíõ5#2YS‚ÈÉ<±›Ã¹¢éYæøDŸD“µ +¹»«òµc¸ÅuJV½ßÛV ÁJôßh³¬yê¹`¶}ÃéH‡_pú·•½æõ›dþ:ÈÃV³“’àé¸l}y&wɹ†Ý«âYé2Xÿ)½¨]pKo¯Y®Uý¼éùrµRõ–9SÒVî›wÇ €Ç ¥-]eqN‘'2âˆÃõŸ‰¯i¡IŸOædß]/žÐlNpŽñõ”hLTS)wË«IÛ^•~ÇlS©Õ`ßG½ÄÒŽtì+üÍ€¥HeÒ]ÑÚSs‘û|e¶‰a€_¥ ñ•ÎÅÂñ‹a½ŸJƒOßËáæÌ+òÆ´½®Òw"4k1ýNq~œ+0TÞ—½€§D4rnúG%íp$<°QÆ› ¹5)wqý¥+õ7jˆ0}E4¾®˜ÑØä:œÒàªß/‹NÎÅ:G½¼[&Ú“"‘è&Ñžzr jÝÆ&ù®*{Iäêþ¡\Û¯ãÏvJ„NôE}rò°2ß&اL÷¨ýøÇŒzêú–…ƒƒW— %G¶-ˆK¼jä¼²éS|×Ý8b°¿Nü±åÎQ^çœë™}¶j±“»•ºax2ìœGÅç±Ñ*sÉ„b Å{å|ù¹Ó¢.Äpþx›íÙT\ìXØûœ¯1eýù˜:î…KÚEªKÜ)µèµ‘Sìó9ýÊ)BóŽ¿?kTš¡é ^ÌvåŸ>[f¾¸VB7ü¢”$ßü>âäsÓ¦îŒ G#ƒk£\˜öÔ $öö1¿x9ÇÂÖç6ÏÚ½x¡\ WÖ^zí9ðNÌÚ ßu¢ÞÊ|²©ëÎÞÑ’Ûž³}sç\i]ò=Aµ”»òžŸ–KÙÛS 4«d6ëûË­RISl•Z1±Õ*éò[ÖîÈÃgËVS.>õŒ¹ùðsx;ƒ–éÉOÓœšyEkÅ[êܺXŸ\©¦¾$Ñ¿ì?z2ô@Ìæò܉3—3½¾i ¤N&®ûâþV„5]ÿxpqÜC.W@`#·ÑæMŽ0ÊÞŽ­£ñH%÷$÷ºÅ¹Ö×f[ŒíÊžÖ{´1³ù†(ÞXº ~4k8ã þÞu`ò$?\!À¦Ü]v¦S±\*ôŠCô‚D·Öõ7xs%EŸÝ¿-\Ýx}Kv÷'x½e¿#þ•¡'¬xž¯8û¾§ÙzMÅy:×þ•‚ž|ççÔ;o‡‡¥CR¿’Çk ù§XöÓA˜¯vc*r%&"dS;[Û¹à¢K%Ø*Q Å“ÜSl½Ò×Q¼k’ ÛüûÞºÜ/%GþU]¢](·Ö½WúÔÓ.M$|.,BØ…ŸFñÀĹ>PÊ3žê{w{®Ûö=¶1j¢r„P«p/Sc)V$¸:ý,¦/²Ë¿jËQžÐ[ãÀ«0¡A¯°èãF­o׃ÇÅH*ûx86ô+ïj¿ ¢ª zÑGȳŽ!ZÞñú ÍKZW%{MsI ^¿(oUJnÑèt¼ ao§”,%’ö1,»ør)©©wIàn^q(ugdM™(·b¼D€ïö9ÜæõSÊý46 ôú7zÍΧLEI´~‘íÚ=êmqÑEÀ‰Q^ºÉ‚r%‰íaS‘tJ£‚ëȯ@èPPüN”w{ò2gN VüÉ ¬_Ôš²Wí4ãØxtÅ·½Kyh…QæiW/p¾¶í庖õ)û\09Åý¹\í“.ÝrÃEªRÜ´éºDÙ#ׄ‡9U7/™õÔ*fE78ô§½}µ¹:ÃÁt¡â‘púZñö¨ŒÉyϾÊ—Pb`K,áù)¼W ñEž5++§Ox¨ÄÝ!;a¸Bì×úM<…P]»%͈ïÓ§ÞÄ%Wçž|õÒèo…óƒÎ…¥Û©Œ+C goK&_=3‘™-›jP;_~öFŽçátË`þ˜ál)ë>ÇŠrïLq©¬¿)gj endstream endobj 852 0 obj << /Length1 749 /Length2 780 /Length3 0 /Length 1311 /Filter /FlateDecode >> stream xÚ}RkPWUÑ8ÆŽ¢â“‚7ƒ”ˆ$Ù¼@##†—2@ˆÊ£Öì&,†Ý˜lB2A‚/RQ+Pk¬¨GP ÃQ)¡Å†2,.ÑÂã1ùfÕjð£$!HLN™‹ü8#5¢X% ëM¤Â}ÖÊQœKG1‚Ž¼Š#cs âNã÷uȉÿ’‘-k¸œ@0\ xBOk4°‘.…„ÀÄåÝP5-‡$u¨uäV  4ô¡% €ƒ£[t°jˆ§ÚW c8nTÿß‘aÌÁëH ±Â2þ¿%¡0©Á 2ˆ A\Š§¾ÑÓ†Ïóõ% &× ,>Ÿ²Îç{‚¥þÖO*å:†ÊhøߢòÅ ŒZ-ŠP9ÝÚJÈ—g&ü˜Y÷¶Zì˜Äê¼E»ßaÙ‘Ý°]H~õG™)>dòþUw¶øÆ!'çýR,m;–ÕÜšìèÞýö¶!æÒÁô{E«KÉq |_°Êÿ~uI‰:#Äe“WñŽG3zö¬qš'›(sû&ªõý6¯Ëü®JÛ÷m×úeÏ@"0÷Û3ŠÉñ!˳ »¢‹:ÈÑ1ö¹35²}ºð ˆ¦ ™^.AO÷6ÅdyŠn·HEùì Gs ¶_o+ØÑ¿¯ßµø[½õ½HBDXN0ƒBw×¥tÊðçiÝ klËî}WÉÇ©O΂–a·<g¼ÈåúÕ…fUß­çÚÕ˜íçFhêÞi–ýs >æzzNð惛ç±B1µ/ß'8õÅ^õ”¾@DñåO á,WQærcQP-\⟺üUÆᢺËוå'ÓŸ¾éê¼Ý35ÌíM5#]]cï±óÏÌã%­ÈÞú²jp–vFp4r.OŸ%v'OÏIqï>prÒ¬Ük5«ãêüOýëˆúj_• ”-ˆÞ¸Ot¨?/ìµ5ÃXà.'ˆyºIâ_kj#ø+eWb¢J¶™ºãȸ¥½ÙUúÄü›¥Ó’LO’ —u2ÖnƒW·ÅT\«vxØô2L`^è}Ý{´ùÚõ=ZWû’Ø /äMI¾ä}¬‹5¸·wNvJ+ vL/9èózcî)}Ö½ñÃþ4»ù¥O¥¿]|äÉ»R;q°´Ûª²kj—À¼^`§¥…ôt§I¸Ï½¾™úñJ™';¯¿ÈhO;«œ]{8ánãý[5Ì™¹U´å}åNàÒ·îƒÓYÛx.‰=»E¼Ðàçpþ<;Ù‹)\8ïâç²Zó™¬K˜’gnìÇ´þ %-ãcwß46G¶5ëUÞæµåôºLïÔÊ”3« '9-ž`—¿Ò圄sg¶O܃¹Îs¿²+":š,) k…Öv¸·8Ç9­ÞdËv°¥¬ü`;<Þþ»}¦bý|·­G]­.²%‡sn endstream endobj 854 0 obj << /Length1 1166 /Length2 4030 /Length3 0 /Length 4786 /Filter /FlateDecode >> stream xÚuSy<ÔÛÿ&k–ìÙ™”%1ÆÎ û’Œ½B‰aÆÆ ³`,¹ˆl²&;Y ÙÒd§ä"J²JȾ§øNÝï½÷w»ßßëóÇçœçyŸ÷òœçHž²¶“ׇaÜà&4^^,‘>nœm!o GTPÊ")iÄ£àÿ¢©„!Å#1h#(žÊÛ{( (‚À °ª"u Rþ3ƒ¬±HLÀŽ‡cQH4•2¸|àh¼Á×…„Ãlá8 ëÇÔÎþ]`ˆñ%b‘O<@æ’핳çÎÉý(jjj܈2#8‰@¤¨ 8 ãû£5…) ÇR›†ýˆµö€Ãøãd<ñx_°‚‚¯NÅ€8 ŽW8KmÔ 3ÄøüH€cù¡™ w§ETøU7o4&ü/؉†ý FðU¸„FúàŒþL…XþÆp<@¤Ò)à~x »§Â’öD_øORñ EÃBƒ}1¾( EzÀ©?–`ÔÀc ðÐàÿKüsÇ¢¨€!Ýñ78‚z g§Âp?ö(‹ \A EèÇ÷×Ê™z¡0 Eü;Üê(˜;Aì.œûuö¿¢ 0Ô”òŠê*y% UªS¨5U•Íø—êðµ†"ÿÛ'èï”ЀæãPuüs$8Gõ&@æ§Ïþ™ßƒGºÃ2[çHDu õ§ø?-õþë×&ê§*2È êX~(‚‚bÿõA¢ˆÿãÀ¯Wà¸ÿÿÉsE!ÝõÑÔ_2!q&È@8̉w÷üÃ.ª ûùáÖòÇKÈ+ª*þÂÙ{"ݽÑpŽz?)8öKIc´;†D#vxª+¡XØ_ÀÚ€ÅRåùyAÔ³î=Ôáð@¸;ËøÆ]ë¦WíÍæ½}áù/•»Ô.íÄUXn£hç€Ç:-¼Ë…U ¶5ÒrŒ¬û<ÏW]¥ƒÐÑF' ßVðÖ4h2qóDq3‹ú¬Ó¶³­1Ë ­ƒº„’2ò ‰ÏÍVOUî8Jž7උÆð…‰î9Íö³b«,zxÜêóØYÍ“_#V¬ÌP7lÀŒ#½­F ¯ˆ[¬æm¯ö¥ÍX¿ìòhéḕž·ˆOÚºMùÈ„yÚ‘j-îDgFe©çW ³_¶€\_=!Òá.š4UÓ/sÊR^Ûøõ}¶g[³Æ9nÐ.°ì±”}?£Id%öf\Xzz¬?\æ‘róžÄÜ.½#x¼ëáy\[RƒÄÊI—RgÑßk£sh&à/…>µÙ½'ùhQ̼&söÉw +ß¼×~xÊÆ5º”²pg$ëŽ<ôóJì¢_HcîÐZÍD…Š¤H\jå0ãÓ*&%Åå’ûAÞŸn¬í‡’(¡R±‘>÷áP†!z+o\ôlô¿¨+žϼ…<4–6Wµ¿)] >$µ˜¨¿SÁ› «Yº4ÙØVt LWÑĸm;{óvkKæJ¯tÚyÿFÆRüÒ\—'%)büÍ®˜nZ+UÍ™¦~sħ·RâY”7³þ'/ä¾…eÌ9¹$æ.%vJ¹{þåtæ¸s¬'—¦Ö¯Öûxöü’üEÞ ÁË  ÆÙpÝ×#FEO—t·Ëï¦öõ\Ú-ö‡|Ó–“6íbYì7C,ßv6,<ššM9š×b„I&ì}þ¶u‡Œ=šq#G¨MGÌó͘‹%g£g§æ¸y+Þ3Vn%¡MÎêö׌äú<•k6sEm¿œc x‘Næ á\‡ûìÚ¤H!Àúœsr טþT~jŽ„e¡„Y·À]t%CØ %㙑pÕ*/ŽÁ†¢¥äºHºìÌkñ¶ÏRRSæã¢OŠv´ÄßÝ] v$ñO“÷&ÕÏw¶ pÎëÌ6Ĉô<|X5”7×P•³¿šËqä"ŽªÖtÓodÑDõr»"+nRò½àÖé;ªׂ×ÑáEÀºµöèÕêCq½æÞK1JÊ÷ÏMI ¶zAN_\ªñ œó lv¹8¥å*gþ.«q´.µ0=óH ð²RêY–)ä4ùàzCƒÖ«£¡ã¨NJ•¶÷grê,îÙ&Þåãýí†-hþ-—ÂÛ1„èºÓn]|rî1§‘¸l¤w¤L<7\—pS0C¢þ³Év®¦Ñ1°qÁéµø‹FŒ—u›Üc´ÄÇŠœÜÑý½¯û¨¬èlxÁ˲׎×ÍÛxó²Äé9ïÛÕ Øo»\èæËj"½H)ÊõÏÙÑ-ž~µ@Z‚ð´Œã·8”øÜ9d5 °dš•r®ÊJÖFÎ5Hìü¤VÄmµ$ˆµ(’ñYÛŸÑê•ìÜMKœî¸cj¸8Ä™ébå"Õ#u†‘üí'ø˜ÄÔì×ÖmõÍJ&’ùƒ²£Óö3'“É säK¾w\¬º-w ËOœšÏgßàÙ=¥·sÜ”>†Q“q6í]1î°ô}öræ>ÇÇjcåöaA¤šëdAÅ%x>q<Ù-/E^ˆ²¥fÐSˬšÒú]fšXXJϾ<ˆŽúzíÌìèÒ VTo®žÄ€ÍĪ:Ñ:Mâ3W{$ñÆ<Kߎ[1Jè?ÖšÅ\—µ9V÷´Âá‚<bK„#œç#×"Wî…=^iI£°ÓÓªõ ø’ 9nPÞ顱aöSO†³}³…(ŠëÎîÍ©¤ËåZ†t{$‡F%µS¼Ö¯åtn0§^¬ò·³Ñ;çõ6µ2Ø}*“ºuŸiª‚×GÍVï´=å ‡•Ä¢0ùjp—þM‘úá}ŒQ”ÐÈÜéÔCQgß‘ XÏmë“ÏY ÁjtzÙ(íAÃ0;ò…»[C+ô¯âÒ²¨lЕH¶¼ûúñí ?è»ÂÚÂYð4>”t¬¼+Ø<è»Qˆz®à™Só„/ea¼jü]mÓa‚#q­=o’ã\Ò6›28<É ›íÑ{¯NÖ̤ Ì×öž4smÛ+œa á3‹v†4gPÂ$¬¤ ¿Ž|ñ/Ö¯ÚCR 4„'¬:dZDžK%®;ð–ïÅž:{…p^oI»õø ž *Vvï««ö]‡}‹Þ0L9ýªã¹ÐùN…˜Êº1¬¨+.ض5›mý©d—ÄJ{¥À'Ù+°â$£\¤†ß³7õ‹Ñ*y uÈ®ÍÍña—µÞ¡N+?/ËL^£Ðeµ­ê€³}q•À±5¶{ÁW^M<‹„fV}Ûy ñÖÃø,ãV@Å1÷`ˆ^O©1sÝö>a{£šÄ¿†0Ñ2vžç­Ýe«(—u¸Ø:ÕDOñ ²’ˆB˜?‹Õ„?ÖK1•C6ªP¼}¸×% `þpãÆõŠUü9ØîY? áÙÌ{éÅÖ¼®‹B¿ï=ê $¥ÊXÌŸ‰ß¼˜ƒ³È–(·-ADÒ칦˜ïY{½ì*+ó ÝÔðF!“Œ?œùŠxö®ò¾BC¬A{ô„ÛIà7RöººV2I'UÈaÿj«ÛÌ”~XÉ©´)•¯‡Ní°àò|­`2ÄŸ?±@ΤçÞ‰ÈÐ.Î_8ÌìºÊ?Þ6§r;‚Ý/,ww¥š=í£W^›þneâá°(C¶67c™¦Óñú¨p|K<ìõ”_lºƒl¡Gåü[»>y®${0ßj jôã^õù™,“–Îô…{/xm†Œ¹‡=47v§¿Ä0 Á¼4Áž¿™O5…YdÞ P†¬Öåóèˆû”ÚAºë¦Çw&Eú–v²92H¼MD+¿žp 5”G˜!¦ÐµASë|G 19oã–ó2Ò¼©'ÍG-ûúâÛ™3_öJ[/™îÒûÐú*bz7Ìf…ÅBb·5W6b:à£l¶cYR´/²´´O&’ŸºN[AÒãæ¹ï¥‘E óÜfŠ9ÈLþ¡Ë÷Åùž .]Ȭ+E}á9ÉÝÇ»—5eX8 Œ®©‡2œhYu~þ:[ÃûU¶¤(j'Añ½”®‡Æ±Ó¿ßÏ¢Û1Š—’‰•Ôeð×ÔóX/¨âL´"IƒÿS¤Eû endstream endobj 856 0 obj << /Length1 1626 /Length2 14669 /Length3 0 /Length 15501 /Filter /FlateDecode >> stream xÚ­·ctfm—.»bó‰mÛ¶í<±mÛIÅ®8©˜•ŠQaŶm;9y¿ow÷}öù³OÿXc¬{âš×Ô=Ö"'VR¥6s0J8Ø»Ò330ñ¬ìLÜ\Tì¸åèåfV€/9;9¹¨3ÐØÕÊÁ^ÌØÈКĀ¦37779@ÔÁÑËÙÊÂÒ@¥®¢IMKK÷_’L&^ÿ¡ùòt±²°P|½¸mí€ö®_ÿ׎ª@ ÀÕ0·²D•´¥$T’ êI =ÐÙØ äfbke ³2Ú»©æÎÛ¦öfVÿ¤æÂð…%ì0¸8M­¾Ü€ž¦@ÇTtG ³•‹Ë×;ÀÊ`állïúUW€•½©­›Ù?¾äæÿ"äèìðea÷¥ûSrpqu1u¶rt|EU“ø7OWKc×b»X}©æ_–f¦nÿ¤ô/ÝÌ—ÖÕØÊÞà ôtý'– `fåâhkìõû ÌÑÙê_4Ü\¬ì-þ‹Àhaìlf tqù‚ùÂþ§:ÿ•'àËÞØÑÑÖë_Þÿ²úOV®.@[s8f–¯˜¦®_±-¬ìáÿ™i{s3Ó¿åfnŽÿ¡s:ÿ«@TÿÌ õ c3{[/€ÐŽQÁÁõ+$€êÿ®Ë ÿsMþhñÿHƒÿGÚûÿ¯¹ÿ½GÿÛÿÿÝçÿ-áfk«`l÷5ÿ¾c_—Œ±=àëžÈþ¹hÜìþ_.ÆvV¶^ÿ_NÿÝZøo¶"¶fÿ]'íjüUa{‹¯¶010ý[hå"aå 4S²r5µ˜Û~Õë_ru{3 ³­•=ð«¯ÿ*)€ž™‰é¿éÔ,­Lmìÿiû¿U@{³ÿNÿ«Uÿ"Ï(!££-!Lû¸\ÿe¨ô5®j^Ž_ÜþW*òfÿyøFDÄÁàCÏÌÁ gecþÚ½/BÜL~ÿ‡ÿbþ¯³¼±«³•'@÷+o&æeÿ¿žÿ:éÿ7q{S³ÆFÕÕØÞìkÒþSðÚÔÍÙù«ÁÿZþ¯¬ÿãü¯™=¦pË ¦¼¡ÖÙ™®uXùƒÅtûz˜ÁÃËÕŠ kº2"·¸½Õ†14Mð|´yÍŸ8¾ïÉÐì÷`ÚRv§/~àû‘R÷¢¬S´sÒî3”}Ë<ÕŒõ¹œ“Û„Ðá`ÒØßþ«¬bPúE0ÑÎê sùHHê^ˆNöàˆèošÞ€ÑÜ‚ZWtrJñýèñr`dhðO÷5dïm^,9¯1–ê qŠ«—‘ó]£éä‹;§#ˆúˆ›£ŽÌ²(C’Yóï,yóêuš€lÆ~Új-†íùÚš‡[;ùT`ýX»‡R1e·Ú£™zI'ˆÖyÆÄ,â ÈõĞǸfWˆˆ‘ƒ¸)N&vÆ:HÖªº_TQ=hQÇ ôD岞l‰#Þü,ì©“ÅAõ&þì o&Yû1ÀλÞcP‘pîsí,參 a$¨•­J>˜k¢8ñ¦öE7ìqÊ€ˆŒ<¹îoª«a“¡ ðT(¡;Þ¤dê­øê-¬á’’¢·|úš©£kX/8¬ «•ôÌý4­;‹*úõO~¼Ïņ±gPü¹’i9Ús¢ $QàO–`Ã{¦Wq~‚J‡oÊŽc4Ü«ì].ÔïN²­PSÜͽXýë˜y ÛVECéçéÇ„s ½’ 4®¯òlŠV ¨©‡«jÁ^AnüÙø£]ñÊœ¦f6šar¢Ïl:x…~ùr‘ß5UYº¡uÁtçD©3àkMOˆc&Ñ Í¨ùXI±œÃ¬'fØN}NëKyµÔ…ð&ZÏ>H˜‰áòÆ1üÔ64Ò›qRO½‡zyè Yeà*„vMÌu9i7G€í³ïþ+°¬)ŒßžHIyKýaƉÔFiçå¬ÉÖxü‚4uÎ+¸dB¿ºETW]ò˺ˆïE¹½ Y@Ta·`ÓAì×s‚µ¥*ÃtYßu_À¾¿îöaw`n›ù%ÑÝNðék¤3¦[ºÊL”í;ÀWô@/^yBvÅ£ÞG”›,ã+íÕ7·×zJN€´ÅÊóZX´ÐbÐõZ6¶%B€B.ó\wXúÖùäÏMÇL±—§¯¢?±Pl\dUŸ´“÷PÜ>–9H¢ßü¿åÄTˆ˜ý®­“…iW€!ÔýQ†~ ׳‚‰ñ>æµÃ¯ëeÉÆ›£O¡ìTÀOÝ„8«euÝó€÷ãûu¡¾úònr@ë¦qbG‰˜a±!qüÛuíe6Š¿õuÉãQÀÔfçA)Áž€„D˜ó“Œ­ŸŸf m½JœDba4_Û_}¬±«í4½xîsr4ŸiÒèØ›±²yÚde¶ç½Å#Ð*tñ¼€ØålÞeãr ¤qƒ,Ý„Å{ü¼œÕÞ±×ÉvJÞ¬ k*ó­¯ö>­ŽÄ£—äϷᜠâ ïUJ²ïç”ɬk}øçוգjÛHºp›=hNŸïI?S¶–§¨<(£|Y<ž ßÙ ¿NJ Ù½Nü‡ !A¬æ1è®wœ;}ÜZ'/+8ÌUÔ&ÞhsN)à!ë!OO½]Ý—˜gz¸¹yÙ¡îèp$#gàœ†~F ^M6ÊÕ¦’S¦r ‰ÝÕøUvˆ2½W1o"gµšâ± jQT®à˃ÿþ1œ»F:ôëf†ž6ïÛ[~DñZûØèÖtŸªÈ­%l´/Ig¨@ìA1±Eõ­&ÕýsýLÔr@S‘û“+('Wøgö‡R,²&³íÇ£"¬‹o‘ô¹T òãó÷õ¶áîakÊvÔÏlîÜúZÙŠGu°{¹I×[k]嚉Ð?[²6m;Jª(]ýEIƒMd׳¹(ðåÙÔq{ˆIÓc5ëꯥk@`µ5'ùŠ6†ó]å‚1Cë° ¥:Y"í$Œ€'hï\„8'Ad ¸^‰Y4Ê+þ»=î4‚aÃÈ:FtÀÛQL+®«®ì¯ö‡ç°‚ÞÔ¬}&Ohú‚‚¶¥É«’|‹tâñÆ„Àº¦rruèÉóƒà«Ð..¾+•çÅ4q3µ@3¹0²4b¥Rh  çæO×-úµýœíê4“ðä@´œøæMÒ2,f Ã?ìÇÃé ‚C +Ÿ&ã¬á»&g½9Özçª[§"M<‰!æÔÝí·‹øÁšîbðf‡w¬îO÷)¼÷]—×Ï‚azç]¡ ùÏ\La †T1G͓ƺõá(#t6 Ë˜s«,ELÉ £Pjœ Š$p†ƒcŸ£ûBfcçö#ðü÷†ïÁЗ²úÒÕÍ¢î\Ç¿õPò(em‘"¢¤>˹o¿FU ù~ˆ3ùI±bnŒ¡ôz³èoP}&-¦`ÁC!²EÈwwk r¢‘Í*~½ëŽ‘æÎI†ÝŸtú·AËcÿ‡Œ ûkòtìøB/V…"õö^¯qÿUë÷’­j¶k“*¥®¿*‹ÄçxØ´ç=óðOå¾H,ükð7Ç8A@aµ mׄo³Åy›ÕkÑ ¼§÷A)£K窌‘£±¢,y‹Iù¥ý=Ó¥7Öj—Ó½D¡v—V:âϳ# ıԎ±^ü¥JC€]í,Yl csïJÍd¨·+UDKž´:íP' ¾ä¿})Øþ4=âz&’áMŽ¡ÙŸyDì×’ÐÂW‰±xÝlÑ®l¹…(ÔþÔb©0º^9¹æ”¹ùýkx”|ƒ Ó&ïúØEá[E»é ë†ÕÚáø½Ì)}÷·\ï#“(¿Ë_þ4fQ›æŸ§.rX™…äý,¸µúÓèlnÏñ[¡áÚ¿8”Œ‚[°1‡ò3yfãÖ§íW«¢’À`²@å뺑wÍd0–þNxjÛ¨Q‚Ýpj{çÄe»+î!ÿÌù{dö_’Ãàf±¦Ú ƒW¸}o2+éÏ6FªÜ:÷ÐTck‡·§dÞt„§Lôß y©`²ÄµûžŒåV°@#h CwèÙfò4AÂh~ÐÁ¹ç c¬Wªv@„¢Ÿ(W ¶.• Õ ÂhX¬N•>èM'c±fw Ò7"›Y¨Lü÷²Š¿«Ïã{ÖÎ7[ø2º"v²7¦x³3Ðn~/˜ë ‰“7Ï>ñZ“­áÕôþÕp<&ÞŒ ©¦8™ê\’üÈP>šRCÙݼI-Ýì–9Q8ÜûÙQØŠ (o ©ŸA1p.觑ã²Ý0O$ éÛEüÎO¬™f\8°á ¦jz±ÀÔØ„óWÝ÷¸ î&²ˆ,;†Èá„ïñœ §ã̺P*"ÁÔ9f3Á e^ìáñ÷*ÎDÌëÝXSOÈD÷ïRÇ͆ðGg¸Y¡E! Ìr‘Æ×m•O‚Ü]qº¥Ó—de·Æœt¢ûs0¯¢±ã6G TTm Þñ‹o;–{Äší4 à8íÞó–5÷ü{ŽjÛ?aþþ#h4FmvÈ;»ÛØ\FÔ™¡ºï‘&œ$Ðø“¢Œ'®­oRi[øG²qVw™—ðJ Z±-xY¤Îù[^ï4ÆßÏ“çÑÏ Žåj¢S ãñ°°iÆTÃIDèhuì3Û€’kSÜÙ†:ª1Ë´Ô×èõ¦±ÿP‹Âåàñ¶%RWœÃ‚{)OŸÛ›k³ÀÛ"´÷ò•_zI1˜Ð1'<×$$»>ÏÖvï¥qʱž_5|CЈI© Ô‡õé4”ƒ·zýG¢ïj­Lê¶XÄ`[ ½×ZÝ¢@M]”1üOMd}›Q¯ÁÕ6cÛ‰§N[gü³%Ù *ÛUû×ÃùñŸòBÕ f$Y³¢ ‹wéAJc%ŠÈÆ–:“-kYŽàA%y–†0ÐÄü|2¹ {´Yq@!dS¼° þ눛Ë@dVI­ˆì¾ñÊ*a¤˜Føg†ä¶%>"ÛÇNC}”¹ðüÇó¨RæŽÚ£Xs-õžlÆ«‹þ_ò¬²‘@îŽV-PظC£ÛTBx?XI´(*KöÃ'?~\ì¶DñÍ@ ˆÖ]ä›á’ü‘iŠ“Id==özäòÙI…‘L: ±è|È’¥/ ‘Ã/êa=ßV傼`.ϬóA7AêâîàWÆ9Ú^ߣ`yÍ‘åÔ'Ä:²5@ÔpÒ°º?ámñnç¡r1® 9öÇgÊ›_Ù`òÉl ±iö¹¥!©U«\–çùÇÊñZþñ±zøè)¬æ­Éi È}ÐU„S]"A©lfYo ¤²g㕘Öß0\Òk¼…Ëoe5-å.ô®¹–¶ÙE³‚ÂwÄP›¡€1#ìMxrü4Æðá[¦J¬ªB6|ÞËyÚn+Âߤñóuè8M.Ns©!§‡Š#<|v€&„éʵR8’m…~ áÓõi%ñ~ñ¬+’Í7Yé}|ûö÷îünQ†;‡³ìÀ´ÖIÔv¤°¿^N;yÌxl×½¸šÓ,ùe8ßhK}ð ’xÛµ|0™«Í1PŇÇßûþÊå¸ó)téè¹* ïŒ63ùΨya ²´CºË‘LET»Qæf<6î/l.y¢Á?2 NV çHV!+ÖϘ;ø¨ «³ E ¦éxh;rsi²ÜÖø ʬ)?œ]9ø‹þî7(MÂɈRvˆ·Ðµ†;‹øZ¨ù¯ ~`:S¥£ÒÞ­Ý@?téàóÉœjïâSré_ÿ™ÍF¹0•ŒG“­g¨Üíj;þ)GR1àJ§U¼h¸Ü„U¦ÌÜCD¦ÏÓtÿ¦oq@úÏRE/®zæŠjöX,ÄM¼òwV+\ˆéÓTSПÎyߦP”Ì1LÎæåVñ¨:Ýk5ÂÂgú.¿B8Ïù0ƒe‰ÙFs‰†gP3èSêÛs0— à±ý•¡Ýï0Ú?BÎ ¸ zBõß ÍZµåÒ(h0e<¥râkñGÃIÎÒ´Ø+ý#1|€šlu½é´gÇÜc±hYù2‹[‚P# :¸“^ÜS”ÏM?P÷cúøçûˆ&¶JSõÓ¢éøøÊäQÊÐϹßIè8 †ÎøÒk{®£ ®:XSln¹†;âSÐK2ÒsÕfâ3\Ü€rxŠxöª`~9•¤Æ_ýîÍΗ‚c-ÇI*鎆 ”9¢A «dÁÐAu!p)ûÓßÃ'a7 3çJÁ³»ÁšsËZRY99ØÁø 4˜´ŸSóÁÐц׌ùÁ5¢„…õÇ›6 èZ¤03ûDÓfò-hÇåqA´ga[éá¤ÎÈpI®úg3 ž=fòÆlŽsÍ¿[©sØ÷áöüÒ6#7t‘X‹=›÷½ oõêNrŸ9ò™!õÏdu²aèäüø™ú‡”&'¿—Z?æí泌¾× DÐG³3…Õë²Ä µ1®2ºû—XòAŒQ'FòyÝ)µ-­ÙcÌs¨‚ׄÓ.ª÷x»}-1)JI?°¸AÒ}æJ ëhË ?ú™Q¡0K@ÊßÇ—ÎU[¥M²°ŸA\»KÂKÏ÷;nóˆß[ KýáNHN‡-âÀ ƒf8à –¤!t4稈Z棿µã-ÿªAÞi™]ç‚Ê­åÔb%_Y)Ø.uÚVÓÓXù$¿­d›«¥–p¤{U76¹W6æZ”PÄ (Åa‘y‰÷A„Ì‘í£<¯n&ÑJ;ÀWè„vÃÿvoúW9ãQê|´3ïa{ûqì3ÞÖÚkî¨ûúîô§×½†WùÛ÷â¹’tÍâm–Þg=NêQ½;Ì;É#åä¡Ø°Ä^;%x»y¿Ÿ;KH®ŸŠELOmwý²r¿)|z°#÷¹g3lÄä¬"FnÔÁŸŒ{óô%!ηÁâ'i §Ê2iëa9í5v%‹Ú‰ÄíD•f<ýÄd‚¶†,µ»»†9ç!iØ¿’½úƒ¯9¹#¾),»£aЯ‘+c ¢î8EÉ…Gë0¨ÓŽ9–ûNŒcâLŠÕ@¯|Ê(’aÛcwê*ž`. GÃO$¯• kýËcž˜|²+< ^MÉyR3Oœ#é–ïåe…Cºz'¹Nú¾²N¥ŒÎ-‚´>i©y÷«?ùíØ÷¤[f˜¼ª½½JþŠ 9* ýàéÓ7lôŽ5wöæ,Òª·ìDÌsvõÊKQB“ÈáGÔéó¨HQë‚b³3¶5™ÖŒGá×lüU1àjš|¤0¨Nt™oyK=NÚ;übÚÓ+(ۨˮ`¯Öø —Ò”|‰(L•~Eþ5xÞ-n¦—ý hÅA\Ä•Ú½¢1&¬y%oÉ£3¢^0óº?•»×œ×Ðí9Ô~E WÅCë‡ó(Jg¹I©™f4sÂqpÇö·ùôñŠRÔÝÝHÂmb½ šó :`»6o·´†Þ}gâ‰ÐÂͽ¢¾›WËÕýÏ”Öú[R³Ý:áÒl2Qºa„!Jb ÙØ?†ª³q0¶L\E¦[ø×xÆÙêc³E-5dIÍZ«ñ׊óü4ž^o³Ã\Ì+P&t„nL(߃½{ àryÄ—K‚„p?>)K:ˆ Îøëm oN»zw¶—Ûª`†(Æ ?ç_\–Œkd(üñ„¨ùÿp«…¼¸,WL_Ù!òýj¦…$ÎkÞäòþ®ÃBm§,7C¾–”a§×°jÊ×­ù1énÌ+“šNÎŒ7·§¥¥½¢[UéA~öh-6A6|óÂ-eª*hKÎj¸Îý¡âð¡¦íãoÑ@ÿW¬ÍÇLI1ª‡ùÜÒïì`Y‡Ö[¬‹ t€»¹,Ä ¥“¨[6;àV—ú€CXø¨š[rÚ[Dñ.m„ã 5y nãô6:¡ÍƒŸŸmâÓAç“…¼ÊµZ¨rµ§¬S¼‚×ý·{l®õ”öuƒÆØùm‹ >´ÎþšÄ¡µê l˜o¤ay.FÓ[œîz¢ôøs$¸ ÏEåÞ!MG÷#ø:h™ÓA0™U)—B”Þ]%ÕΩýST±Æ:î·öàË=ú7-g:ï’¢Wwk8™ƒlz´¢§£µÐUÇBb5.¡ë[P€äøàbgÊò*2´ßŠbÁiMï¦PF¤ÆY³í9Ï|÷Maw0Ãn% }óÉÕ…l’ô6åz;¿ùäO]&Ø,AÉé°¯¢'Uå'Gyå4¤ ÇÀöÔ×÷SXÒ•8@Q¶?ØËF<ù³Éç7¶J}NälúÄP¶×[·*ade¦ô’Ÿ'Ovb‰T7Ü.½Éÿ¬²N@‚=cáäFÅ콃؊ú}4ž”2sOÙs ɇS¬ç]1b(Ú3§„TƤKGÈý¬<¶ÔËmGÿÇ|û*a>D:îÝŽþ‘ø©-ãC¹ë£åŽ´V6üÜ:AKÃ6´Åž‹8ç3¤TUfn‚663뮜-ü²ŸÔ§Q ãÃ2óMÐÝ~ªð<ìÉC?“³ÏÍ@©Í=-a~ÚÂØó5É‚œiõ$Æì=i8zqâÛkÆ}¸Ã Z¡ƒŸÿª·eX–ÛtÏ\[-![hv¬âóºÏ­õFÄà.º„¢ Ü­£\S룖½½Ö§õ”óÂZöPì3‹4‹ö«Ì )!¨uľ”ÃùŸ­ç}aV€jùÂ〉ûÜ¡õrr8¼ìz[žªýÈ„Öq9Ó¶·ò¹0¾ÓÄ!’î>\åL®Ïtñ2Küˆ›©m13ý:0Ÿ%/ÚA3ÍZáq¢ìÈ[-•ÊÜl!öí(Uó~%°Ïz AõøìLz¿dÅÆ[¯šƒB ¥Ù·7—u™Ûž+舃²†åðÔä(•MΚÎ÷ŠKaô§3™¤œºxãàÍá7¿&ÎõŠv‰Íe¥¨¬+o‘3ê‰-,»­0þYàñ/ÕcÉÊôÈ#YÑ̹Aú°&VµÏ¦ˆ÷ÓÆë¡7Í ¨ßÓSFr '`Ò6M•ò1õŠ,ö“8¸Ô“ûÉ|b7e3q¤Œù¯°^þn^¬ëÿ*ÍjBçf~‹˜[ÄJÖw‹7 ëYƦÁ¶uÎK¤¨“ê6–¤YÎ,6œºªÝÐùeg‰ŠƒÃÒ4TƒÉq%p?ã_Ã[qùóH1 Ð«œ—¦BRûTßá÷Ä@šmý‘ äNì(ᤎÏíe1XÍ.‘øÝ-ß=4»@$7¼HH'—æ[RvÓöMä~´ýÊÉ„Ëþ»¬‰nš«,E’Éžp²oAô-VùT¡z-Œ›åI­©zµSVS‡q^^%iŠ ©?k's£IJ)áR äáWDѬã•6È"|äT*llÑ¿Pªy%¡<³— s|ñ†ŒyŽ¨H±±qýn ˆÑÿÈ£ëÿ©’ªŽŒõWò¹ûÄY»ÿ  ò›fÇOþHlO䛣´bÁKß`ë^dûCÜ Ñ.“" rÂLW+&­‘ÛÈÎÚ¸’, ë÷ žÄ¸—8™*¸&ã0v-ý)`ó ¶PèE5.?(pf~öÀì[m.6ºM¢ö4°wG¡ùXò’0»®ÝGï |Ï`fCà•x0â¨&„É’”ÓžÅqXqÌÄ»w%o•‡ažúI’±ïÂœê_¹œºörz•[J!®Edpa&‹Ü#®¬ª¸6\¶r!¶múöÀ[]BG¹ˆƒâ\:ò{«I¤”°æ 瀥vì euBY§r vÁ8Ju\ò£‰ÍB!í„;æ‹1¢l2„³mä§S=›FzCé¨ÚïïN‰C…þæ«P-W$noª4ôN¥°¹¼q³¦”qÜf³ ž±vÔP&2*Z>çUÒüEh>dâ\y]¿Æ‘Á6 ž¨ˆŸšýíPt`$ú£àòâyGÑÞ¬oùp+-hÇP±à ?p¹˜Êo°jÊ'ywØ{†¢ 6mÀ ¡Æ€™ð~N ֘ϬëkQ„yË`™»RoàÕxÖ}Oóa.(å°ç&a`PšKdr×y\*;wZËÀ r48Gs-bæÎ$Ñ£Èiû²ï¿_ÙUðˆî&.·tªÒÃfd†ˆÇh¢2Þò.0åƒœÑ ëAy@¶ ì|¿Ñ^ŠòåÑ"kaAlS}—j^ÊÒžÒؽCèäôU°D‹½ÉòiÛ ý…–œ$N´Ëë0 Âz’¾ÉÁ#l·\h ­RÒ@½‰<ºYÕaÁ»†¼›±¢'¨Ç}iâ,¡Á o§ß'¨OyˆâÐtú®ÈÕ÷ª«ˆÔ‰º^Ü¥öŸ0uAûËG.ø.oÝï8ˆA܃Ö1*ÛÜߤȰ^â€ÀÂlì>ëÉCòÌi¥@ v§Á8ͪ¼jn‰äC /7m›õ1ÐA+KlPœ£$¤ÖPàRoT¹Á ´O¼ü‡ÁN¼¦³…ÝÓÍø­Ã6N63ØërŸy€Lð©HÅø {úS[j¹´)äЊKÒã „*À¡ÏO0ô¶9¸"¿02=7щb°´cÓotQõ;ëúŒÕ.ý H‡vºts·ÅöÝfüM†§%Ô1F6»çHô>4ÏÜ#Ä5¸c˜uÔKœ†J>9ƒï®wµlB.OPÃCU’2äº#MU*>oˆ3„1Â0ƒ  Ê94½™czל®ŒÚ¿ãî™KDYØíÞÇë}XXwÐímòjËè\vÚ| +°ÿ£?ƒˆ7Ú»&¸ÕŒ.÷zL>j[¾j-xF_"ŠCóJ*WL‘ež`Üã6¦rP4©~0ÂH¶uêŽ+Q¶Mtì<ªf¢Ü\‰ŸÅ1Ö½þw R*¿n4 ).¶zíµ«óOÏ4ŠTE.Õ:nm£YËxJæª*H\­ èayâÆÛì‚x¶~æ‹'Ò6®•C3–›FùkjÝ| Œ7âr‹pÿ~žš}(Ò¹&Æ´†Œ˜©GR:úý,jÐ=mvKlÀÇ-“G0S’ñ³ÍÀ´RP0Y›Î9ˆÿ±õ·­™½ÿ¶‹šíb¶Ô£^çõÈR©Å£ÚÐD_v5iôjÆš„xÒšÐaë(g¨•šÔœƒÙt¦fˆÿ‹•ƒl–R¢ÿ‰wí&DP›¹`‘Òƒ§wC–šd’·R|·*FÒuI^â™åãV‚ûÆq5+%…律pB(ÊÙ’Þêó:nza(H3¾ýzRÐd•ºÄ ×…‡ów3ÞÉï_ËÍ—ÅRÐ̆}ý¤ªÇ¦˜ìÞ1=ãX·F~³+ {ÆFÒ__á˜Mµ(¥[å”~¨×ÔõZäÛ1T°xH7ñÖk m MÞ¾‚½ˆeZŠ{ÇÔÕRËGÅdw\wÚàqÐèq?y(e(]éf«J¦¸ ÞüÓý”gû—@è1šêl$9,F’‚¡÷Û„ý™9ãU‹§»³¶¡ýÔñ”ÇÅËY¸`ûgž9,DùZð8JJÖKÀ¯–z0×Dßø,ª=ÇæzOJsë¸}õœ ¾y BžæûŸH7%jYjº@Å®žvO÷·M6†1A÷xqÈA(3Ä9=Ú¿»bLo)µ:eNº•^}½?ÜMâ~…{À±fo<¾ç§O]‹^ e8îÚÿÒZÜÞPEr Ùm»€÷¨Þx °:ßF Š+”„Ôœ7(¤zw¹’~4ìÉKWØC35Y¹ ñ"…XtøC̘£Ôø¨³lÂ*ùð¶†U‰)¥ |6¨þ€s“>gRÇ…Þ’ \_b5ÒÁrHE]ä1ÀÐLS[…Þ«:=ýÞZD9³ÕÙ j÷žM󡆚Rà’/Šu®âÏ1T‚& \¯œAfêËŽþvºTvOŒýÅBø¨ýë.µÐBÁòá=Áü€ÎóX´©Éx×_Ñ~µWY?ÂsR9Ú´ãïl|ÔË{ÅrÙ®ÞOÅÇAÞ²Ò)jœöðåâƒìhcƒÕx¥ttó} ÅäâSíCác¡ 9{!’<ìYÝ`påùy®¿ñ°7¤5XX+kGïŠõÍ°0&!¤w0&#Æ@Ìþÿ=¡3²ää}ìY/ë2çfjp¸èpÞá_CYžrÅ·Ýr&¸P™+îi³*|èÝîH4.k½ændG JI‚¼ºmVî¹kÅç®ìï¶Z¥m/éµµÐn@_„Hì"Y¾ X‰]”í=•˜,EÔ`ø^­œ [ë4ˆõ Ú¦KÉ46`ò„0̶¸•ôò»\A„SD{½'ß| ÇB¨ôaØJa¢¡vAYaF.w9xÜL@áßð–Aj˜¢®7ŸÂ Š¾çúNýAP/Æž«3‰Œc# ” šôÒ¢û[æ—΢ÛuNIŽ‘wRn¼E Pú ¶îsš_–ìîþc‘³(ÆÈŸq!ÉxËã° H¶k1±§Sò¸ÿûè¯óÅHoc‚ƶ<¡Î¯,=v ÖãƒÉ«qÚ+…=KYVM„ƒ**è8]l3ÀâuCTòw¯¼0÷ °åfunÝqA*’ÍÒ.òFC.Êã6µ“yèÌêf1·4w6Ô9ëç|O.¦XÅßòJ áò® ðï¾ù^8Â"tÇ•çÓ$–vŒ‹9Õ™H7e#áí¤§š*†)…Ÿ±<ØiÈŸ²4ý¶nßùPfPuä?+3ƘÉTCÊ´a±ñÛ“°”…ƒä1ÉŽ‰Ì¨ìÔèAìå!ÝØ›`†ð^y)I28ßc߃ƪ¹×QyØ`±òPM9DbzÐÌÆÀì¹ÎU§®þ¢(uöV÷‡’ØΩ²,,ÞS¾Y Ø:ftce׶ý˜‡‰.Áľ:–‡ÅVn>ê&ˆišÛ{iê×Q #ê†ÌÜi¹­‰æ˜l¯¼„›Å·6Ìu|_äº&(ÇY‰»§I`èžÏvºC«í>AbÔ"ÝoÛ®Ú’Î’å˹Kò­é>MÄü÷-|ÂÌ!>Œ{´SVE^v\ÄÁÙ,´©I\wŸ¬þÁ¢wº>iõBGwò0.Zón® êö"ëÈ«Às=ÃÈ·–‘ êËêROøqìµPŸÿ£oñ×xýÆr÷¨>¢¦p2ZòeoZõaÈIõƒÚß–gaÒ4¥ ܺ>ªøÂxKÅFw A›?ùOAk¯§ TyY€V·=N£×X×I7óÇläŽFñîâãë¤/5 HÂÓùn6¬K“F´9Š±µ êO[t"-Kªë”u‡áÊಕrS¬ÁÁA‚&·ns§$Ë –¸&o9°4®¸1ŠÉþÆFI(sÇוÑoF‹NOýžNx#_w¸ž:¹ª=&8ûšú¡Íµ®¥H/+ÓDšjÀi-<Ízñc»3ÐÝb$aç·*®AÄ|:$ÂÇÇŒ|Ñ«T6Ò~6Å Ñ-C¯2sC¸0¨Ž9â±[aRDalå°Lsˆá i,ÛÐëÆð\©˜éKÁ”à=ƒï_ReVîê]ßÃÍoý„²QÆ+ž8aÆ\¿ÿd†FÆV“}6/¸c¾–JaÑÝì᪮$Š>¨Ìf“÷褟ŸÒzULaÄ‘q9oÑ~/Œ âN¥ƒ^f_¥B‚tÞ!çðJ-¦I¿¢6µÎ:l†9Ìþx9˜¶èײ£ XÊžLÓzx¡©‘ìð:h1Ý]2,°½]6åtÄò‚ë|StÈÈ×±X¶z ±fyÿîâܼ•­}̓‰~ SsH£ÔkšláÉ0{øÚýâŠÉ`#T9]eé :¥øæÈþ­éÔØm AÄmœPœÃ·JN ÚuøÝT=ÀöøBLaÏ:ÒWµ¾u†b,P q$ö¾"õtÐôUš| éy[(õÇ ´ôª§ûF<¦Y:ŸÙŽ­ÅøFéÊŒþG†¶2“$¨h®¶¾$´úô-{Ul6Úž‚qŽÐ³Ä鉤^·Š#踆=ÌØŸ,[´‚_N˜Q{mÅÖÜ9ðÁm!íŒ)|µ«ÁýÙ2§N…¡²Š+>ÈŠ¨¢5};Þ&سŽÑSõç˜íØ Å‹ïW}à« äï¾´ÊwmÀ¬exæ›G$†ôsÂ.Ž"†˜Ú¾&ç6ú†ÎgŒìøE)¾Í-‹ ›ZüƸžŒ÷p—kÇœ" Ñ‹ué\÷¬,l9¯ Ju+¨Iö>¤É®Õ°¡j :Ú[qÍ ûà =ÐðÉ_ÒüD&z7;ˆ·?ÊÚ¶‚À/Ö­÷]…âSÂù»>OGµwøw™%—ÎiªàÄŽ­Koýa6¹>7ðÀ(/$½oE©|oš’ƒÒ—|­\°Dxg‘4é'Tè8ý>v‘½ð´å÷ã7VÒ ®-ùÕÜ_ Ó&ÈoÚüªiÌÚ ™O< ›`Cë­Ä‹Æ‹™X9WÎ!þ³³ÁÄ)0ñ•ßʈàœ{W$Žcò¬¨]ð~@õœow䲪ÄfÛRˆîGKF..lY™Ñˆ>“*†ˆjg\ˆ²îŽüh»+F™e¡Ã:(àzJ"‡›gkf$Ë|8ÅÅ™÷ ABNé@½ ã%®Ãßõh°¶ Fϛ×,¹nè(ê~dšË=É“è9ç”? ea'•ì5ýðxÑ·b—@ÂÌçèùˆ}8ƒ×ŠìMœ‹)›=—',š½t Æ´U‹Ê/í¹&ѹØês¾AlE²4ìjÙc†ƒ2Ò|Íþõkƒ=¤Cìá'–fŽˆOÖ[Q6Z)Ôð¤ÉÆìX&Â}oØ6‡yohcžïø£Ž^Åø*—0¿‡&–' ìqá\‹Vå9’Oj$˜(ʦTŒ{¶tÀÌ?d âaL7Êhqá—À +ë}OÿJWGˆ eêŸx,÷ç”=+˜î­ÖiÁG ¸Š-!ù³RÛ¸•³o1šÎvU4ÿz½¿v-[wâöóLsDÌoaUMø‘À6—ÂARäºJìEU+$À×N®£~4{ ³ŸØÎDØ£¯ßY*O¸EÒÀUpiÃC9<±Æ¼ ùý!bÄÉ"¶°Û/FTƤdâ—v¸¸nx5—VAÔW®ß…´kd Ðì˜po&È'…ÒŸ ¦Œë\ÖU ¾Íö>ØÓì£fªŸÁOÞþmêˆÔ(ƒX× úR30J0¦™­ˆ¹-÷ÈŽü&Å‚ñxˆ¹œšº”ûâ¯0ü&àÅØ‘Žà'²ü²Ü…J²ÈËvÙÝGm~Ü×ë÷šÿñÏšÜó¦ ¡þrH—“«S‹T…"ٙ ͘0+À”y‚•.EqÜpÙ©XŽ3/ ´ æ‰`Òä‚Æ\ÿaµ+ø™ë©MxêÖ­êBÎ/‹q5 þîa-AžíÖ±ƒzàÑû§Úˆ­Ù—ƒ/ãã ç@¨9+—ò$•'´+ë{R»ZžQ~è«‹M»%Üj¤,è&ƒà`öE|ÈϬim0¢sÑý i3¬õ%u®ñ¡"Ì᜶㗠©$X$j û’¾µâ”FëÃÈû)<Ä|ºVx{ÀÆÐR¿þõGŽØ š)Ae6^´aUϨ\Ó‡Z¬Õ[k«ÍÃÄí¼ðŒÊ4 Õ½àËÃåšœ"ÉIXC®«f4ªXAáÀŒ‰½¾õgÒç ì²âQQ#¡ö_’ºÆ±ÈvÕ°'çmøÔEúkÃ@­ü>r²²_™ˆ½6ø€þS.u5m2Žy ¦¹¿˜Â׌¼þ"üD\³[½:ž³Áìþë:±fG¤@T'ë=°±î-Ý»%@ß?ÿLsê×óƒTƒË·]z…¾‹ è ¢'=5ZBh4^%¤¬f›OgcS7^bvb„ìj‚ø¯†¨ =x¾?hµú®Pi‘qjBú‹™J6ßà[VVeÑŽBwÜŽõ£&Q“PV‘ŸM_Ð’ŒA›´ãP g8"êv y‡ð¬’ýµOÏß‹ã ¸+F\ø~4cw|÷m§$ïDZ`?òR4ˆÎêðIp•<º•$fÐx·4_e R@L8ªAáoFxY+Κ> ô'€Cmá~åHPâ"LªêKU¥Dgtœ ‚!èIZJ Kn®SS¡PF…T½Oµ(óÁôà <Ž¾D¢`x _ßóˆGàpèÕÕ‚´ %ãYÊΆpðêºëž\W‰4 ÁÏðV~iðwMd¸w³Â.³,MíùšþD¢Åè7«}º40Û.µrßyà§ÎÒçðc·ÙT÷ÒŠäɶ >ì"m¼È­–òPó†¢u×XH?šô¸×~á+Çÿ€kŒ¦êvn¹òÈ)¾SóÎË9 `Ä µÃðX¾+ÂŽ&ž0 é¶êtmä_PŒëˆëùÝ'Yd;Øà÷Ž 0 'hÌI³‰S=ÆziýBXÎÇ6AIïåS¸Ì¢Ë_k˜0nsvv£Õ îU<Ñ=ìNûÒg™Þ endstream endobj 858 0 obj << /Length1 1642 /Length2 5195 /Length3 0 /Length 6023 /Filter /FlateDecode >> stream xÚ­TgXS벑€ ½ ‘¢ PE¥WéE:’‘4’Ð¥ ŠR¥Ié½#½w)‚"]ªˆô~£ûî³Ï³ï=Î9?ÖzÖ7ïÌ;3ß;kx¯ë )BÑv054 /$&,z¨ GÚ¹à ÑH]´¬¶ ×ă@& àåUÆÂÀx8¥ÆÃnMaP  ŠÉÊÊxÊhŒîàˆò™šò þeùå´óø!Dâà(à ‡+ Æ a(<âß4‚Á€xGÐŽ€•õôÍ5uÕ|êº&@u †%4¡ïb‡€C€Úp …ƒñíÑX â‚FAá¿Zà ¸q@0‡Aà„0˜;†ù 10,ŽÃ¾pÐ Fá w€Gá(Âú«‚Ýý» Mð@0™>‡ÇA°p HȪ¯¢öGxG0þWnœÑöO(âò«¥ß†€âÁpˆ‡¹ã岃¡pö ä&a°ðße¸àà(‡¿*ba`,Ãá4î_·óWŸÀêŒÁ <~G£{ý£8CØ ÄÄ 9!xBn8 òk^4Qöh ˜èv¨ æŎý}A|¿f†ŸPŠF!<€P˜=@D'¤òý{* ÿ÷Dþ/Hü_ø¿"ï&îß5ú§Ÿø?ýŸÿN­æ‚@è‚‘„øcÏ ‹Œv PøkÙ¸ ¿ö ò"ÁH8Âã_ÅþÝÛöGÑJhôïØüŠ(‚:Bb’ ?ÌpœÜÕ‡ã!Ž@{0‚pq¿í&(( ‹€£`ß-!HTôo˜±#â„ú¥„ä ý{Í~—/bb®£oú@à_lÚßÎú„‰À{``ÀÿÍdªƒ†þãð‹JI íô'”"—ÊÊJe¥$¼ÿŸ¬¿yÄþ:ë€ñX¸;ÐRTXTT Hxÿùüu²þ* ‚†þ!#<%LÝ? ¿`ˆ Kû÷" 4þçù÷üÃ`î0àó"÷äQ|R¾„)µó½Šek³qg&»Ü8#ݯÝä2#›o{R$\1pû¬Æãã*ætAëÖbw3#âfÓ+Ø÷7ìÞÜü-é4_nÔI ,ˆØd_IX3}îµ9ªý•ÄBJôÁâì{C›¬“K×ê@ØË›ûü~Ü®é~ô<{*H\ÙK†ú«hKÞ®®Ýˆ^Þß»ÙÞÓÕÙÑ´EÚ²À&ò’œWÌä»z=ïa‹Ý)‡œ‘¹Jc÷wý}«°ÃI÷VXL®B«¾¼zŠÙ>À'º¢°T½%ä™õçg*5GUÕ¬¬zó!Þø‡¶â‚™ ù›3˜·«ùÈËûÑämC©…½ò¦Aš”?(ã'eˆÜº]:šËŒYØÉMJ~è,ýÖi%“*ûìÑ2jå|{äÈäµrÇb˶+N<¬î¦U#•&{ˆ6¡sŠŸ“ì<¡åwJŸžÉÁSÕÖ=êÃkÉEÃHªÞÓ.C‰?^¥ð¨Ïo|ÖÜ¡5mÇbZªö·7ª²pvlʦŽxõd~³øGÑjÇÀT ÝÍ…Š Õx$“©PY‚SmçgBPg–âkˆù]…Ì,æ­®8äLán­¾¼>ÀÃ"îsšGµ­ù+ˆòt¬‘Zw£ªU´ÝS>bïü.„~_Ù[±÷Rt zŠFþ†ÁÖ5/eVýɪŸ†œ\øƒì¡#ÎÃgÎ×ìµd•ú©DºÎæöçz©\)2ƒ‚¦ßy ¾ÔBZ¼p5d´ª.J…U *yóC—j‚dº¹u|÷[/^RœªÿN5ì»èïüuöržV%uú+ûôÕK»z§¤È#¢ŠÄ@Ú;ªû$ '$Î,­K‡š‘ m‹•×÷ˆÎÛ5urÀù]w`V]›G•œÄέÚH‡À“/yGƒl~ë¹^‰ e”oðm“Û=ªƒ–OíQù$覹¹ sí~' +l Xt¡Þ)Dótôs‘c%¨<‘ëzÂb«÷)38ªRÀ/søÏF‡By– LmH_Æ3^ŒbätUSK'· Õ8/DT(‹{&(f§=9fÌ Wœ¥õ.ùbïTËtãõ÷ä´s†!ìë+y%5¤vÞ°,vÎóí?%ËÍv -a:ËùéÖß|Û3‹þh8ÕÍæ­’›Gl.v$½$ø¼ðé£ý­ üg] @»¸æ8º…”½‘1¢«‚r‹m•Cãëðñ pŠönAãî™Ü§“m¼Þ|i-.jBM¯i5Bó1ºNóÉed\;?_iïº_6d©&nTuå:’¢_–…1³,ÇKÇ…=ˆ™¾Ë»;Jÿò“á¤t-—P™`Ë>¦Å[±ÎÌàu¬Ã’nÅȸ¡çÞ=µf‘•ÕŠ•ÙÛÆõ*³"n>¯â á<˜s.^êÓ|òî}prŽ1PE3Gq­^ Ò\ª•ýì²ìÞXÎ"êÀj¾-x1åö¹úÔ†Ãw»žÆäÞ­„=öÈøÔ'aá9bS!ºya¦ÌŸè}æ7r G›/Ýù;ÍÔï8d: …LÝcIûNÉmhÉëæUþy> qLjH÷Ï™*Ê”çIƒ5k/ãÂxâŸ÷¸1i£ü³èkÌ7»ZvtïÄè¾_¦XÔv“¸ ¦Bõy‹ä¡ê[³ì7v)†Aþ[Ä!¶[‘åzä‹d¹Ò ;r;Ëù·?Å)à“úˆy4ä§I¾Û·tC‘$raSŠl3™k-ïä-‰–mâN¦-é¼µBAÍçž+*]Áâ²µÎÆ›%Ý—äÊŒ ¶Å7-µƒ'½-funë5φÈ_©ÌS.~YŒì»å ¿Í¨à,jMð.¹%÷ømåpsf‹˜1•oQÜ’Ì¢ìÐœš¨ ÐaØVÓg/ò°ßÀ~§ó“†@a…þ×û¶‰ÙËŽòW°uÑ È©Òà\²:zøäÁÉd–›—XSljˆò+W}AŸÚw ©ì®| <ÙM_ [Dàñ≭€&Ä“Ôî®ñO¡ W»t ¦7¿´“0M‘ }z0Èf¥b‚½·Á¡§A"帣63‰LÌrLjä—bÌbrz6Vó yÁÔB*Ñq£—õ%MÜÉÌ’Üœ3ËõÁƒx€þH´í”ÇÝ×7¨¥_³Ë‹'56 kõÌÕ 1+F¾É!Y¶ùAnϺå®xÚ` v_®ØbâS}ÞÙDoß÷íEÈ îˆ$§ I°d›YlËþ~WÞ; HV:-Gç,jQ3íÞï÷¡’:çú°ÀøÖ \OüŽ±W¾VžB:1›4(~/ö4(xÚ›x¶i8Âö ;µ+RÂÓ-¥Ç6øÊ÷»€¿†]Ýááe–jâ*ɦ;µwwR iâA 6JdM ƒ1ÑÍ¿WwêU©ˆ<ÑÿPãžq+X¢€"®Üf6“TÏ¥µãGõ„_ÒÊíºÉÁ)XÂ붿ªƒ8'k¢Þnq®•¡Z'Ò^k?v™Zrè À1Žd@$åø?Ü´¹zhâÀ–tRÞ<Æ,î~¦-ç eTøº\+Þh›ùÈ .`C¬²|hXŽ·õ©pŸ R²nËRÕ þø’ë;ÞfÈõ\ÃØÀ½ëwìGÀc†„%!½Þ¢‡OŽŒˆ’îÐSŒiUõw2luែ7?ér-ÙòYqS"Nך½åé¶ý¹‘fÕ÷𠵘‰îaû³­h#Q/SžÀÓ<Kº­¦Ú#ÝÚ·T˜oE(!Èö0rJÏíE›g ûF†¶«àvÙ¦É30]Ñ´œ¥ççÙ~_×Ù„æ|j‡jó/ÝU¨ÞDäQ †¬µsÚ*¹bt…~ÒEJõ%ðü ¼(è4>`Æb×3-µï)®Q™l-“,h±&tváœ(þ2gÌv¬é…Á É‘÷—"Jó•M±ÚÃÛ^L¾Ãº­hòàIbÚ¥”Ó6%Ö™ÂÑÚ±Pï2p¬kÇ$R˜eÝDya-÷­Yièeþm ö3ÃÏHcu™~ËÞŽ€î76Ëp!8>·u¸l*†[1Ç”8°178ß{%#?€{´ÞK=ÌV ?€]é<¤¡§dcžùzqé‰]•ôí™bv‡"­#›WɃ>µ"÷šÌ,56ýžòH +Ù•5ÕD‘-XPQÄö´µ ~\˜¹¼ª¡+ÔVÁÚÒyÜò2‘tZ¡WSê7Ä/hÞÙRš\½®B“³Î…~Ñ_uõ¸¢EÕžkÑL‰¤œ_búÊlðE¹³Brªö†•Ü7Þ€Œ¡“õÅ6Eîë‚l±Øjø܃֥›K6DGgßùog|udÏ“£Ü½WÖ"8ô‰e¬DæÛî–¼ò€™»Ü+¾›Ò^ 19«²¯g“´§?Œ-] |—†’OëÎN aS;Ô»ÙT`†Š¶·Ín{7Ãø,nÚLÒíêì[1®ˆ ùÃVüfG˜™´y{¼ÌàÉÝlŸ²±N3„|ÇiCQ-(|±š)ÕÆÒT˜$éqõiàÁÌ|R $çE>Ò€»lΈˆâP#"Ÿðè´·awùœá°6K½~åu`l鑶ˆíÐ1þäÑ$D­‹¯õ×2„¥÷r`H¼=SÁ×,.ŸHQêÙ¦A5$GMùHûÑÓwB Ë¢<~ô–+éz€tÉ{³J÷‘¯«USÉ=¯+ü\tpáQpiqh,ºŠç .·»Ò£e ×ßǾ|¿é±ƒƒÝ€Žô~Ü—ûðIŽgíuË»{»K"$ƒ}Wæt³rˆW½¥Ï—±2õ®âÍ2bÚYf,I‚+²œJwáW8õ?Y´„‰œ®N³z4j=•¨5ÞV9æ iîpê©æ¾ŒDV.ôù>“R¸ÆÀölű@ÍF•”V7²§h!ÖDgVëí)nÙä|ÕÆ÷UÙñÝ{ÞgÉ™wœ·õËVèÚ‚õ˶uºƒc3{!﶑Táˆ8Iî|­Þgõç€Ìañ¦hØͶ=Þk³lÏúë‹}ÝîÛͪ®U‹9 YG¶2‹L° ±â/Ì`ûfjd7Ž¦X†@¯‹ºU!âWà>M{4!Ä´·ÜŠ ”> ªQ›p_¬/¡‡•Õ“ð3 Ì(@R…Wš^y8“aØ" ãUãªÍ…]¹œNsÑbÑ‘5Á$Æ@Rê(äÄ]0u…v7+d$“ŽéŽ®ˆóõmŠn÷ê°K™`;W¥éo§Í âõ§qK•ÜéW ´ëÑÌÈæîcZø–ìàùQË7ãé¥N±Çr™P¼õe¡ãlYsºvq!¦ë\žåþ•Õú…˜ÐßÌ·2î:æöqx™5íäÕ}Û—dVXOQ5*÷9É× òÝ*=NÛ-·Öå)ТÙw$æôq £Ï~Bk[îÖ~%’­#oß\Ž êžþÎfT6±þ@V¥îI ÒÛ¦b5ùIò Z¾úžíÕ ²€Æ=,kÊà¤ìŒ ^ED~‘üvù#;æ[BèL‹¯ú=×e§^•5X gjMÊU ê›å%†žYôX‹Á$gó˜“èjeÏÍR‡ ;AãS•¦Ro•e¢Ö”-æ¯ëìK°¸›Ø  ͶtNˆÒÐÞs|^è.UAYt¢è¼#œZIZ¹}c™: M}Û¡ŽËåSMùæØÒPäí¤ŸRÂû‹kg÷º'MĘˆX“÷Ľ¹ÉTQs˜)Ý¥¼RijzØY¦U-ìçìþ=ÔÝ×Ð]Ö¸póV 1º¤#:¸9ºÒôŽ;xYá éÿ‰Ôßc¯96uy!E¦ÈµuÆ:¥¾õêˆ ù}Ž½ýM=žÛLÙq{§Ôw"u™‹vÖBÑåªÆ Ǫj0&yøµrf•˜þt> ¤#âËŸh6qióq†Jÿå½³p™6€SÄ{¹C¶ÔØÐ(é™úˆÞoT]¯Øño;bš<Ç™Y#{,¿f5ïw:{P÷ìȤá´×Õ^lØ!‹ÂüF'±Õ¦ºœ¤Vz´i¼:žU?gŒý¦Û4êà²bÎ/Sž?áf„³1S_ ø›4Kƒ‹I£‰®©\–»Ÿë*/5ÙZÚÔ*odªümͱŒ"¢þõŽ‹¿oÀ€P(Æ@Ûÿ¶SÛˆ“hîbAM=Û--•Œé"'{J!$‡g­^¸¶¿kq‰¦gœ<²É¿(¥ÙîG‡êŠ[ Û@|—ñ}•hQʨÍ|ÿ¤¶öüùûz|±ýÊÕ"Ëž¨þ¶Œ”Í,+íþÛ…û/Ûù¶·B¾§Û'õff^ª™Ö!Õ“cÚ npSŸl1õ-‚«RPTëû õ[n{ßÛØR!=«Øv´9p@ÕÚæ M¥(‘KMbļËtÜ‹ˆ\ѵg÷‰!±d½ôü'V„9;ne{=¼ü)Y“É}¡Œ~“Ky)ú?ž˜5$ÛFÆüä‡_r.[þ~œ­ž¥Õ¥æÔ+=گʠhæR+ê1Šä‹Éò-Šø¨”A’ÒóÓêYEoº£2ËÄ=ýÔJ2 θß~ý…ô£ƒI· õcͲJ3žgùCÉbY®úö=ó¯ne˜7»~Î$3˜¥÷Í _ϲ齦aM¯`›t¨³ŠC=‹­©"ŽéyËFýjyÐo±~¾ãS[š˜1‚¶ s×ÐL•Œ‰$Rurj&ŠéöügÍ6ÿì;Æ£YV‚ø|ý!#M§{è{ã¢6§ Jî&Ü~2° O¯̨X^z¦V‚jBÝ&Rx¦Žíö+Úà,? º‹¤‹üš—J* endstream endobj 860 0 obj << /Length1 1630 /Length2 18554 /Length3 0 /Length 19394 /Filter /FlateDecode >> stream xÚ¬¶ct¦]·%Ûvî8©Ø¶mwlÛfŬ¨bÛ¬TlÛ¶QáWÏûöéÓã|ݺÏkŒk/Ì…¹ö›‚DI•^ØÌÁ(á`ïJÏÌÀÄP°²3qsQq°Spà–£WZ¸þÊÙá((DÆ®VöbÆ®@€&Ð 4°°˜¹¹¹á(¢Ž^ÎV–®juMZZºÿ”üc0ñúÍ_O+ {åßw ­ƒ£ÐÞõ/Äÿµ£*pµÌ­lQE%miIµ¤‚:@ht6¶(¹™ØZ™ä¬Lö.@€¹ƒ3Àö߀©ƒ½™Õ?¥¹0üÅv\¦VÝ€ž¦@ÇTtG ³•‹Ë߀• ÀÂÙØÞõo\Vö¦¶nfÿ$ðWnîð¯„þZØýÕýSrpqu1u¶rtüª$&ñï<]-]ÿ‰íbõW p0ÿkiæ`êöOIÿÒý…ù«u5¶²w¸=]ÿ‰e˜Y¹8Ú{ýýÌÑÙê_i¸¹XÙ[ügtg …±³™-ÐÅå/Ì_ìºóŸuþ—êm½þåíð/«ÿ™ƒ•« ÐÖœŽ™åoLS׿±-¬ìáÿ™i{s3Ó¿åfnŽÿ¡s:ÿ«AÔÿÌ Íß$ŒÍìm½f@s8F׿!Ôÿw,3ü÷‘üß@ñ Áÿ-ôþ¿‘û_9ú_.ñÿë}þ¯Ðn¶¶ Ævàß;ðwÉÛþî€àŸEckìüÿó1¶³²õú?yýWkMà¿Óý?€I»ÿm‹°½Å_j˜˜þ-´r‘°òš)Y¹šZÌmÿöì_ru{3 ³­•=ð/·ÿj+€ž™‰é¿èÔ,­Lmìÿ!ýß* ½Ù­à/]ÿÊŸQYCDF[‰ö³`ÿe¨ôw\Õ¼ÿæö?ª‘w0ûŸ‡`DD<>ôÌ\z.æ¿÷ïoBÜ,l~ÿ›ÿbþϳ¼±«³•'@÷oÝLÌÿªþ|ÿyÒÿ/0âö¦fÿŒŽª«±½ÙßiûŸ‚Ô¦nÎÎIþ×ø[õœÿ5÷@ 'ÐnuÉÁ”7Ô:#;Óµ;ohBL÷W3øP˜ci£ZQa`Co@Fäw¥Ñ{mCÓÏg»×â™ãÇ̷Ñ>,[ªÞ4àUM!ê&e''ía0£A)bæ¹f¬Ïõ‚Ü6„“ÆáŠAÉ;áT'«3Ìõ3M ™{a ù“#’¿izCfJZÝϳsÊä“ç'ªß£ÃCƒ½·ýø´¹ °¼ÆØþ©g$)®^FΦŸ¯îœ.U‹9­š5¨ÉêÞøoKnÄäñ;èÌ™—¬D_ ûLBK& ²Wáðñ,EItÄc¢Ój2õ]KÒç+k <6Ù+ÜñVðGM1K’dEì[Ÿ¨ÔBR¿²uÂͤ4˜¸P2•µ5bø§9Ùêñ)Äb„êG+Rv…Šã :]ýˆ`ø{@ªÕ2êœ+Ññ[ÓÇ1yD?8ñ1 i[‡;}‚~U)g‹ÜãÃKÛl©˜´®šZ…„ÌhŽïF­õ½Ú§`}Å@•Ú´k"払Σ¸÷½O5híÕíîÉgFœ4š>þÕ$[² ãÕ8¤-Õs R& +¯ïÏ–ªsþ´`R;ØÔD§s¢·S‰ŸÓ!’–Ýͱ—ÂK˜ xºfÃÚ£tT/HC ÷Œ¶Aš¬C“}4d´`²=•©Ñ‚mðvêÄàèš7¼D‹Û%˜¬ìȽOn@‹¾i€¥Tzlä·;ÆPR8îÃúº»g/~ÏÏ«Æ$Ûï&ZüuëköIÅÖž-ÖslÌtܘv‡ÊYOûnYdb3bZC¨™Š®6)ЭÜ,ƒ_î™çî©f³õ$—£<Ќå•õ¥í­–4#«Ô j(–û!VÍn‡b^fæYï¢òäîâˆÁüX¨Ä* ¢há~ÐBC¾¾,Ò|©&E‡Œfµ5SÁ1 ÍÝS‚3ËWXó‡h«½|JR¤Ï¯þ#vpÅfÏêbhïi…Ŷ+ìu—°n!ÐoM +p¡\dŒÂôµ»Ð_ n/P ‰:ª Ÿ{*¦wßv(vq·Wž·®³ ÐN#Š4ŸÕ—a £ÇÅ€UAo?ÀËípjÆL¾¡°;¿bqz½%ý+dS샵N¶èlò2¨ñxþ¹ï‡†ŽZÓ{|PO$I98Ž’f6ØÔ§±^¼/©q©¿Ë#ƒV ˆhñÖ{©$R í÷JäŠÇrogµºÄo8ÞS4UÖÛE9~ì“•ÉÕ,êÌœ¼…7B¯ ¯’çÙÍ„“ 4ø}.¬˜Ìëg~ rYà`£QÖt3¸x+1¬3uì´ì¶Éþ«Æœ€iÚýz³¡¢Ÿðy(\îÏS¨ÇmP{ñm~Óú.­ô¾itè4ý\?«¦àÓØ%:vÛ±<§_œRý2 *+/çŸT‡Däì`Øo#÷(ôåH ;ƒ/!Œ2öÃe]¦¡lisôŸ”Ž S¡±‰ZZPruiðmš¨Ðe¶Yô"­#¤›_ w*Ž*þœW‡®Ô.>GAM±ƒaª&<·O/؈(¿f¬½¬<é g‘MM6GÀËvJgö ø CšV4Ãû÷bU¥zL"Þº}ß}µn2gûöˆù=·`&ó]ö wþø•RÚ…Ø{$…öÙñãõ‘Õî Ξç¼9cÚAÚóKÿZ\-Ù%@ $TÊt×9$²z,·œgh@ “Ëð8=@#´|SS&¯ÜtyfÎä,‡׬/ÃC4–㽒ȆACéýÂÿH,½¼¸çA¿ÀÉ¥Œm«À—-¹‡¨Ñ÷ûî]1Aíhõæw]bÞÑ Ï®ÀÇ8—pèȤˆ—ÿFh–Xqx³Ó𾶨GC³ÿ6Í!á£Öå>£‡Æ„©ñhGÚ»ž‰É°<þQx {ìþz£Ê®¼IK»9m˜L–¿fE~gZªGl¿DÑž3:°#ˆ.{îYL)E`L˜UZÃvë7,äJÏwŽÃ—´Ã>³Ü~K[a<¦ÏÙrs탂HL²c𱓛š÷‹ŸyÀºm—JE‡pgŒH lÐP³ž@sF|ã¼;yì¢'†ýbxœQâVˆài%-&4¯Ëlz¶ ÆU*¢†bÒ`ꎥaH®ÊÉãq7NÑ…ÒnXÚ%TÚXúd+z–`Èâzïq£ê£ð›í‡œû Ì®¶Njng¿¿ùÿJlüŒ˜dŽêu?PlVÍ6§É‚¨¼¤ã²–°²#³Þ­]Dò^/Žl<˜#*é¡\¼¥‹¨×lk æ—Ä´±¡Ç3„3!̨óÓÉo–Žùm³.qE]xˆ¿]ÃõVÍÜ H ÿDغuiC“Sf¢¬o]qÀ )'°k‚~-5ú¤Å1 Â‚8¸"›êÉ/¥6C¸*Ú§X+Ñzªd~цûÒ$ÛÞƒ«Þ~·=~–5…Â3Ež²`)Šçx½Ã/Íž7/¶éSì8&™ CÏ=œ‘rz÷dÙN·é‡ú#†Ñ­oë+ù4’Üñ5n3¬Y@Î1`£óŦt‡ƒ„:]][lâEþçHÔì)y -º2£÷ ,ÍBj×Ù4¦ý¯°Ó'ûXè QDÇsг¼ :Íæöš<]ª$P–Bú$·%‹—eþV‡(šA†ÙW…¬=ù£6’Ô¼˜§|HâªÍ7¬£}´K&wÆŒ`~Õ»Få!{[ôiìD½”L3RÓì®NŽ0 DilØÈÔ^Ÿy˜á,œ¨zyeð·è*Ëk±E¹‡­£Öñ‰0ìñj÷;ýäÉ~;~MÚBÐ×”½àQE[´kÙA:²¼‹‡(T0¿Þà3èc¸0ï:š£ae2Ü×ò­äCýA¢t`¥ng¼&'Fð5M\‡m½[œªoHmºä­0}($37@ðžÌ!8Ö?\Ü…ÊP;f!«Zò‘&K5Ô õíæ{Kˆc±mm `7‰q£Æ(vȧ°Þ0ÎPºCu±42ócâO,á5¾Ö’VÍÑ>}jÔUlÛcºíï¨Ìúd‹@üâ\säÁ†ú½4e:íÜ$‰‡¢úäf¨Ã½-EÊÙôä\0 Žlo;/†”Qš.lUï9[eÑÑ·ú’ÕzI“c-2­ó>öO–/Ø;#(,Iع1ÑþÉ ¯²¨pnë¤ø[:åa¢JÓ_\j§Z3¬?„-:´ÌÚU_Iö· ZT¶;>¤Ø ¸>_»±zžN6B#Ìv.$¥òÑÃ’ Ÿ@öGÇx& ÀnþÔêPl  ï2æqSÿÁQÑÖ§Qºå:Et°Ô„€}€*0Œ ·êucu“•¦_se r-«TÉóìv™tN–Å;’¸Ãö3=—aRˆ@ÊÍÖ®îV 0†=“Ñ)½7Œ§“ÐY–Îñ<ó›Äo00xسùû~±¥³•ÎcT!‹eº^ÛRmÇtì}ʃ/Z#à2):2RE!¢Ëv¬®òŒâÇ ŸEêöû…'ÞL3$ÙòúÈQθ÷"] þ:^ÂzeWº—J+MJ‚~Ôjs|swoøZî8N´ªª•Ô¥÷Í"³®5”„= Ã;â…VQ€¡×˜ª<G Õ÷ÖGŽ+rT èUÄ2n†+mñ.³¸‚½7/BîëÕœ«‡ýræî­²Ywß)Í°åìi ­WÝÍaÚnïÇA‰u› ¹{  ¦æ•›Òò:] ;Í;’ü9Rq5¿ü>É·ºŸ ATR@Óá|.M™¢NCò `Ed’1õÁJMB@ÕVDAX”|>ÿÛþ)RHÖɱ‚ª ^oO!çíLKÍåÛçûI¹r:k¤j¥R!Ôâü8gÅZäeHIåR¦æÕîwÝÙõi™|ߌóê!O_2§¸eþHMw’ƒ~_£ìZ‹Åû«äðRŒ9Ù×h 4ƒ±¢Dš6ç>v:5¼z_pl·¶ŸÐø»û2|v™VB.¿æ6ãìÛ×–僧¤£r8¬ùT‰qJ¶`¼O÷Ú»ö+l{ÊÙ^Ž‰%V*À¬WÔ¸è&àtCTršŒŸ¿dz¢ø)—ÛDMn¯Õ@›¾_ÐÑŠÎ}§ Ö3HJ5f÷¸Ï«µˆ`ûy.prÞ¶pÃd}A%„?Ü` ï0ªä†aElpkúÈЂÆlZRbPY”<ª+®Î:b"¬Œ)É~(íÞ|o=’JˆBË!8Må2`)`9/>'±÷¦³D|úñ„4Úô”:†ÃLÔ±ÞÔmñ[¡»,›•†q´(ž1ù‹L¨ãÕq.ŸõL*Þæv 6…4ßw¿ÁE¼í6z*~Ï UàF¨Ûƒ´d„ý}¡ÆÀ³ÖÄâk9t»í <ÛÙ;/x§ÌôQ Ü·è ‚_5$=´™.‹2¶+\‡D\Ç úö<½Øf§Ö"J¡e™%VÏH]Ò †Ô=¯Š¯T)¨Bç柌ŠPu±†Þ•BÌËe23%1ƒàaxè*D;ÌP~;£Ár0*ˆüûjôæ~_ô¤Å¿îüšµ¢@t†ÕËV…䧮×Ɔb€Ò`} d$¿ÉÍLÍ¡´ö¬ò~82U”ÖNµ!|s”û0ŸÒÊX;—WQIUi&¢ ±QaÁú=Džu¿9cs‹g‚†i¦ â'ëïQÿ’‹iu÷°„“T/”N¸…lÖy¸,Â+~_ŠWä =ÔDMå iˆË-%ýeF×uóÌa ÖL=Žëm¹DÎÜŽ©"ͯoˆµ¿LGµÈ‡.·ûÝW×OÈy­^Å 6ÐnnÜÂæcØÓ²:AÛçbÂK¡aU¡²`ËVQOä@ý)ËÜNÒÜEÈ{-µžê¶Øz$Ø„LCâຆ¤Ä ZÓ>ɹgù*£f@'¸QT$Gª[1®DM¥?}Áf«›'xm4Mak(Þ:Á×Vÿ¤nrŠÆ§@f±Îýµ6z¸$œ¨µO¼gr¨­]YrÛeaœ‰³Æ[Ì.Õ£®bjú¯œÔ–?dü$ŠË˜¡)¿‚t¥Ã‚n†ì”Û´Á™ÑÏóÖ%ƒ‚äiµMɪ/•p/(…Îàí³Þîp%ÎÊEÞ\Åùq8p—¯çÑþCÌp9Ô#©¥Š ÖXˆðj‰¯_ú…¦À÷ýB2ëø>‘59[à߸¤%v;ý;#©­¯©Žôa….Ã×2ïgÜï:Þy;éá$\Ш[¢‚ÚuLQ`Ä(’Nu™¢êíµ ïdÍrG½§Å¯÷.x €GÄéôp[îªxöÈh±ïO§ÈG.ê*JÖZV[Crú›It¾Ááªje¨rZ‡Ïm°‘ q C'hÅ×uQq¹( „ ø|…OĬ/µ¶”ÎΈ¶¢avËéXÒi0b¥ŸŽ±íÎSBÑÆÐ8Çæ_ !õ·þ·²Ä½"§th Ö¤©§ ú|ßC¹úß"ÊÔýC)˯pNìnP‚ÃïTÊîçÚ Àú•§VÓ¡'e oDëÞ[øØZˆ;)yÿ\ËAîvô¯qd†õÊÆvw7Âû®J83‹ÀH~\´X,‰cM6ß 'µwB+lŸéÉÑGµ—L¡>‘Œ›à”àŒ¾œõf=qGëJÛš»)Ûÿ-lÕH¸4ˆoö¯|‚È$ –ûíÖðÛ˜öª^Ç zH‹‘½Ÿ’,kŠ£¬R„¯Õæhu¾kà+›d~eÎíæKžBâùlˆÆ OòþsH6Vd­Äµ6ø4ç wêƒÁtºØZ“[ÛР•jP¹‰UsRLº(E³j_vhÀ}ßm[;þ›Ï7ažg4ÃÄÍ¢;g.úóË–×9þX×!Æ—ä ê9QîëVŘ™ ³s®ü|ýÓcmOq¹Á  dn©¡i¶«Ùyq«–‚Ì2‰ü÷Ò9¼c}aJªl†^ˆ­‰0÷ê |%·EöJÖcÚñPW˜Ù)òi÷íâ׸Bºß…äÎ8 .`£ð¡RE—äZ½/A—ŠòºA*í‰ÞŒ ÚHò¦½8~ÑÉŸy±K€Ð¤ã"©zÜtï±[˜ª5…†bI4ŠÑµ \ƒ9y¸nÃÂÉ\öŒ€¯{ùN‘ GŸÊpsïC·Àú„‹¸À¸•{ù?þª•Ò8‡”Ôž ZÛö«/¯Ou¥È^îýdŒ`´XOǘrš RÕ™IjÚ 6J¤Ôœ·ö @‹ ÑØ#d$’½˜íió„ÙÏKâIǦ’0Ø2Ï)  <À•ì%‡ݾٕuÑÈ YM%«Ä)j OücXº»¾€ü11¼éèƒ{€pÎh`Óæ#+"å÷}»wå\Ž÷ M &ùù).^±iËÊJp_zó*Eõ1¶ àÖ³ò ­Y,üB¢\¦®ß™o­9©ØoÓ¨ºŸž[py/ó¥¸;BŽpBÊH†”=ßEž!Öm¸¾ãBN &¬õWI6Fç lþ÷Ž‚‹på\µ\«ìi“JôíÇD1Ô'¿Öl"á¯F~ð€ñ.F‚M¸ùôééÂY†ßÉK¾Àˆ3ö‹ád´ÁZâo‘Mâ`cÛ²d:ߥþHÌìë8)'ò{QÄÃùj=§B£°:áJ¶è"@wìÒ(€é<‚ˆj}ƒoŽŒò+¤×¼OÃïHÚæ‹Ho³(±w³„yŽ Ë`¼ŒõçjGeÈz©†µ§ "úÌ:úšÍ.Îü6)8Ì5Ž˜è––1f™LÆ ‰xH—r8çö#¿Õm¼AžÎu≠»?¾ÁôZL² g?Ò èÓî.;KJ}ž*ö·Ð8EE*TÁ¹ìWìAؾcÓØ¢ÊÅ›f–œ‰kögAþ@1œ„-„¡1çÌ7¨áV&_ÙÀ+J…©ç,QÐçWd“£»æÔd]IZ—™çOùupíùZZ›Q»íž $”ÒM;ê«%t $ú¥rç€lÿÇ×õ ûÔXÙ4² ¶]KÉ՘l©KmlbñƒXáã@.X÷G­>’`Æusñ˜ŽY¾­ªÎÚÜök–Ÿ N¾7Ñ…s›J| 9ämçMȦhñÞr °GVÞæ²¾ÍH|8·´ˆm€¬üÀf¾”z KLJ?ˆøDÔ TãnRú9^Ãí`2õÅK~Ç0<„’V”è–º:Û>võ‹ÅtIÝt†Ã*yPŽ}hü¨H“§¢zëƒ5KJ p$5͆{ŸÎ x±ÞÜÖƒ¯ä8A¾§é~õû[Åâ¸FÕÓŠ©–tÎL²×éOõ¨ T<"Z²™åŒ‡µ ¬Óµþ‘XÄûmljÒ$3wšÒ|nU< ßÂZˆý$£èكȺ3#a¯%‚ðˆ'äʘb©7úxHù€.¤†xË‘\š"p(QX…•Û ÿ·Lo£í%{,.¦HýÁÛÒ>ÛÊ©Xš«­*ã[«IR#IЫºPΔý˜Š‘tç·O ¶ÄÑ àkVÅÓ¸íK¾uVÅ"Åg •šŠŽ²æµ2™–¼;Œó¯ò‡NzÒÇÔí'®«OØÜ+••B«O"9‚(pÈóHP;Â¥R==¾3,wuFéÕDàý-ËÔ©»CwUèOfÕb˜‚ºFsÌ_ þR¢Øº*!ÂDlš§¶c`ºpriæÔz¯šÇÝë’‘¬ó •|4q°z~8VÈV ‘øØa­à3j)ÜK¾SÝiV"í ÿ„ú®7:k©ùŽùŒ“uˆ†%äDíšp§‚qõ¢íÏ™â˜kf˜Ã!¢íy´Çšh7^‰Û:^þ,ïÀ»>bÓ~°fÐï".!à•¬Ú(¡V#l:#$?ɪ%e 9à ’ YÄp©kú¾©6 ì[Þ¼ø® m.^B(÷N$µi^²×Z»% ªÆb8Quë*jk–Ó£"uøÓª(Ù°ß±ÇV;=©}5E|œ7Þöÿ}]ƒ^MSøe=ˆÑÔV 2ŽHÿ„5×®û2ëÔOÎa)AªbŸï*ÀÁ‚ñ¡úÌ57˜vGý Pf3ž( î·¢Jà ˓ôrQr¤2Gí;%:³ü}=X›OE¬n$m¯i*0`—mÎß¼Rø]“­m½¡9Xüî‰}“GJÐsí, Áž»8|Õ3Høžçß3ÈDñ™Èºx{àŽ¥GVš1úÀvõ{ùV¬@ÝÎ3Bo >áéTR‰ÈÇÁ¹VB´‡í ¼å°eQT9›H‰‹,¼p|‰®bÈ¿U(óÒ—ÈõpQA«¼éI¤V­aÃr¼ÌÌO©Ó{Ý°c§”. 9 +1xã·dwá {m½«Ÿ€@ÓÃź³ç18©Ü`m4¥'Píöz¢‘_‚žéÓ†|7¢ùSÔ $Ÿté©)ù¼ñ/h‚£?Md§yŸcY e5`éòò­Îxuܘ‘õ%¶Y§;_Ù àÕdýö;—ïÏ„Ÿ$¶;™íÈ7S¿ÍÂÒ°a~d/á±ÏÚÎYl©i@Ô6­é¥9,“â[ÐÝа4$³˜D󦌥˜Téß?ÖãÔmö+ÜÏÇ`9Ž¸‰ÃkjÿÄÅØ$•Œ´Î$ †'Wõ}‰šÜK‘ãÁœ“Þâ üÍA ¾Æ¾‡/.p¾iÚ4Š¯=¹äÀ/Í?eŒ'd×µh‘»»B8pYªjéæJÖ¡š;ÆŒ~F§øQÆ~™¸RÓ ¡ a„¢ßºƒ@øõhÕ÷' «]GPØBïu¡z4±×ýgqÛØ¡áÊ!áOãzzŽFÎÐê°_gÜ‘Æ3©Éû{““J‘î–”÷Û^š @cß®·[Ú¦z&‰´‚°¯H õÝa$uÁÝIÒ£¨¢® ܼðS¬¤ƒDCHXFô©ç:‘¦‹ÑýbµýCa˜¹ X/ŧ:>Ýò5¿ˆí$ÓBg»ûȽé:üo»ê%Öo !€òq 7ný·ªˆæ+6]›Cmí¾¦Wý{@G™x®Ù ‹H—ȯºÉ\wV؈{QBã7ÍÕžú.Ñ)“4æÞ`Á,Ñ>Úix(äïKбd?ùï‘©G™W€¹hÛ‰r,­¬µÑÐAlC0°–X%_¹MoqìHŽôf¤ï>¦nã;Oh<Ôg1`yâ5P vãĦã6^ä8ínG(Gž*d(«ä¦Râ±òÊèwE¹8¾úÉ a‘èÔ°']b¨°"$lÁ^^ ž©¯?3ÒcÖ¿5ý -ÆÑw§Ì<‹¹ƒ2Ÿ-bI¢‡ø\dÆÙ9þŠr ‡Þ ±”x®xºÙVè´÷×{àó4œ!™†e(3@ö‚úø ?ÄÙ<£]Y)·ìyŒÔlP†®÷dR¡há†ã§™"ë¯~s9º dF4~t´Feeºù³> î ÔKÞ Ïq @Gºå#×ñÆ#ž/·gõ“ ±ÈϬ·º{0Ižvs¨¢Z›š¯«ë\/îÁCG«Ç–(„@0°ž]m–QW ~-ÍñFœ=›…‡ót\w…†fª¬.ÅàIdÿt„~náÁ¦5¡"%¡lxJ½-Ý£×:‰=ˆïTMñ±5p·ÆodÉED´ˆ=7aJõÆ×ubè’¯ª=eÍIÏÞBõʇ¬¶8’4¨ý}ôÄCÞ\8Ùt-IJQ""ùÃE†å©kbÅ#{íõŸ"ïÓ¬îÝ)Ö¼O079åëöpz&ÑuR4öi‡÷"ï?i°g-ŠE.$Œ­Ø´:Y,E.6õñȉÐáêž–†šù@äš›*`rFŸõ¬>­eí÷«9Ä–ùÓ¬2n9œwñSžåNU™.­)÷Å6i{$á2âá‘Óýõ_õ@>ð3p̬ëG‡lþÛ4Nöz`ˤBÝ°1눳™ó9ZVIæŒXð9îÓØ0±Â÷+(ãSjéŒÊŒÛMZ³R¹ŸÄc*n7¨œïN™óÆÀÂàá¡3f•Jm•i¥(õvº),óeðàGã§.r/ä蹶ôçßõ-I3Ò:ñ×ÉâM,ÀÄ~¬§×ÄЇ Q-èœ'åä0â&ã?Á¥ ³šhETÑ•ŒV¾K+±5’G®‘å8”˜©Vq ÜücÆD5˜¦û—rrMW´x:¶Xª|¸z”V>šLä5/É£—k ñžG1¬Æ’Öœ º1ø÷ó6CQ… g|¬P½XÝ Ù?ñ às»¿Ã ckFgÓ;§Ü,,¡ÍŠ¶µtÊÎ$}²î¹¿= 9ïˆ?ÔoÃ2=ˆh}D8–ÀÝÐ_Új?ˆÛƒì¦8™ "Ÿ–÷2á˜èCl·pÙϦ=7v·™,ÖHãþÊc†£Œg³ŽöÃMŒë¡á"ÆÑÖèGפȺ½R6Š¥Z¡ÆÅzøá“å”Tȸ¿Ûq~n~°€Xt!S QIŠw£UðÃqµT´Â`óT¥ùñ¥ÝcMÔÏ©‘`ɱ?ɽÌé~p\Q€k%B‡}xlØß/=/î _Ü–yùA×Ü“ÑÕ'5!›¢Õ$ 9UÐ$ºÜsøDC_i°DR‚ö(¡<=ùkKý2%²¥*øa3sÓkõò÷ øü2ÀÈa<:4XsÔeøæ[Q&‡õŽ)'j6]§øG ²Š­­vL§•»B²aWXoK)Dþ/×Õì„©ì?ÞÆyToßú#´Øs“ž2p±£3ôž*¹õÈkém×EÒ5ç~É%‹MR¯9iÅ+Ú >Én¸®ê/®öF=! d=¿©&%Á{ˆ&ó¼#Æ‚T* Ë?ŸCÏ Ú Ü©ê°Oýú,͉ìÂ_èàÈz8ÔN8í°àv¾™ÄçŸ0¾“\œCÇ“2QAžuÏ )õŒæøÊŒ!fõY˜Xj%XP†ôCJª5r?DÄú¾‡»Zú½a¹·¸ŸY=võÛîc`rù€É:ºcyØûrR ò˜®7Ge,–ÓS·s‡^N·$9”ÙK™+lá¾ ÂÀ?5þ9&ÆI“¯Û ÷Ò Àé>OùŒÉE@I NS&åÁ„Uˆ ¡òÇ0ì<#¡À?œƒRÿÇ$À'Ä8.àC°¿‘º½™¬Uv8´Y<’x3s>Ö¦ X½„”ìd0ŠŠÚ%N;6¹l£\0óï=I<ÖHÚ…¨ð“qžQhuÙ1Ï*ì~£šŸêvõþÂ.©ý3œãûä§ü3&žáÂŒÁW´I@‚´ä>"ÁsFR ƒžg‚Í®øjž&OÔ[ËUL´ØíPšÌÒÃëU¼ƒ ANÂdi8i™–/·½œÌöï ’)Í–Ù=´O_¥NŠŽ+Ë͇0Dú¹é“Ê[Âo3Ó¼còÇðÔP¤«9 +{©5ýº­¯ð‚Î@ÀÃ`Ä<ØÆXX:0}í{«5d8ÒMò-ˆÇå ë1»e×Bƒ¹´¢á4/mM×ÊÓTMÄemt…/Åeó‰â—Ù€ ¿ýPàgPiq³€Ãi8òÄAîš;Ô/ˆ˜g¼^fƒÙ±§H±ãÐqžÅ›ŠDxŒ‘žÑ¥€^¾¶Q¡ª®åzqßÆž§¬¬’%ø_L±iç8ù;kkØœ„ƶ}ëea 2èÎÆHoÒú í Oö†"*‡Êx¸Øߎw­ 9øJ'xÀ3í¿éÖŒŠ§‘5 ¼ØN8 ˆ9ŽÐ’›â·î3 Åv9õ’Ø«)\€Ü ÉB† fGä}“W]ºgÄÔßÏÝYtcUM‚%]XVk aÌÛÃsÔe›[‚Ì<xÜyãí®ëÅs›kæR“+´‚ÿlHf˜7l–Þ‡"Ž/Ùª,mT"3ÂoD¾õÿÅ¡æ¡Ã8°’>ð¢¼b±rêî¿ÏhËÐuˆJG øóMl”œX* E ®Úuð´zê.jtã'„ƒökêTÑAúÝ(.­«|¬Ù.Nœ/Ù(Š(?~ˆ’}Ô·• Š0ž.ýÿ+ø'ÛÖÒs£TM ªD^Èè˜<³ÐoÀvoŽ°¬^¡Oéž?~çÔ-ð…ÿKiÓ–’_îOÆß,Ó@÷ÈQL0ãŒý»ôˆH’÷Z<Òcš±õ'þZ–.l:¼WU{8›T|,Q(uNõäQú÷ÒŽ¡U2_W“uí &ÀƱŸjÔ6-ùÅ^z5CÙ™GŸTœå)4úà\Ïõ18¨-_Ó÷Ä?„ù³ “uŽï n2ã Õ¹Ö¬nÕ’õ\qÒMz]ˆzjãmñv{+RF˜3ÚSµ‡ç?ÅæÜm‰èþÔsq ”mp@ÊqV´ÓÏ}Ñ.s]ПҙöCà1µes+ °áÝ'©µ¤'Íšt%Gr*!ªŠ®¼Z¨ï©ãÊ­&XÁ«Í©¢!7·bS±z06\5™÷õªîs£IŽ&ÙÖƒÕm]Ów ò3Q„è¥á}´ûKÖé’=ÂWáæyBØZrÃ>A)ý8(ª~ÈGÆ_w ‚ö7ñG«ç5¤¼{6Ba€Õòõç!Z§ PŒpÜ”®m0þ™“ ‹>Ñ'‡Õfœ6½+êSáïË«l)_9„ì¶b$Eϸ÷¢¹R¯c)Öoµ2ùq˜ìls™^ôÉ{ùv0Õt㕤› «õy“ûae6•îm[ìN:ÚkíÔ=ï3Øç Bh¹D+,Wá`Oíϼ¤ÁÏ 8ôäDMRÛÓþø–D*W;?!‡´}û(ÜnUÆk’ñyàíÓ[åGšXFƒ!i.ªõõVÐæü)c©×”Òµœúq¸ú$¯/rËRBôOZTŽÎè—Ë&¹‚i|£úÊäòáÂ]ÛÇ\FþÞ+ª‡ÈŸ´¢¥´úµòâCñ<ùÉOä*AÖi G“EÓó©2ƒ%Ë“Uª7ÎØM®Âô°ÞlÛA"^è„>6 ~NßiŠÐê0ÏÅ%ÌÔÌ.r’èÀÊR 7þÖeû®Ø‡äbìòØ}$„ò zbˆ1©,~y—öäÔ¥à$ÖpƳ­Ñ8#qör@½†Mü€ñ!ÃÒnèœ ^Y˜uÑŠ ç0¾*3xç aÂ?lZšÈ5,Þ²çäSñ>&3­õ2‚–ƒá ¼vVZ‚ €5…ÑðÚþ‘Ãù0,QÙÇÄе:|D]uªñKÇ?¢™VÑ ²ã§ß›ûÁñ~ÙëÌ‘uå!L’Ž)á÷þìŧŽd'‡‰F]M‘˜™z¢:g±Å›¾Ì}øŽ›msMþ“txg•‘Üjw6;QKjåiœ”÷ªyh-ž°Ü¹Vw£ŽÃÔJšž: y„ñråiLoF§çæÒp„vÀ—c0Ÿ$i•Éœ’]ˆ£Y.KÅ€\jY¬fÉZí¦3ô¢þ2Ãü¹Qñ®5'|Ûc‰'fh…÷݆'àÑLü¬#þT3•³¢¨C»¡œóGp`ÉÄÆD}ªïÿ³ î乃ÏÁË,}W¡vÀZÒÇ`I‹)fÚ‹'úg”¨æ±dg›°¤y$JªJîOò£© &xvÃþ3!ï†táÕ­š·-V ¸w§éoƒ”ëþrã„oVhém‚EŠw»#tjá2üU›Zæz‚±Ó®šX˜•qÛR¿—š40*ß°õÖR=¤,=ˆìG6«Â5IDthÖIžÛ†u;÷ô¾Ë¤ínˆì;]p#‘EŠðlzõˆZ½«”Š*GQ{ª)«m(ì3«/HÙÄFWþ²ÅõGã§rz r( êxy=ÄßlqúÎÏB’JýY$]Zo•¶<‘‹Î×@˜¡zІ.r’gAÈM#©¨Õbñ ÊÑdßXãÁTÓ•Ý„'bíæõ2_Ë|˜%Њ«¸zpöàï±sšÒä` Q­ïï‹Éõ¢A\)%èü½_ÊŠ|ýwV§ / ãux#¹aDâ÷¬Ö+êY¡žÝÛ´`/Ò‚NâUQ< 'NJ­gµµ_µ•–§½øNVJ‰F%Wxõ$\<ÑL݉»›Ÿ)¡x2Š yß–¯YÄÕìœ[ùÓÎ$½y=ýp24DAûnÐ}L8³Až^Eä°¾¸>­8E}ÀqÚ G™ÀõèB¼@d+¬eC~ŽmÜX7uO]j`B›AŽ¸_ƒ{%‚ùRöèÜÂúù„ Î=On]ÅÙ(pÃëÿj,I<ô Fß¡§¥.Vœ¬ ÕÈ›£x;ç Bïs2œ#ûµÙ"Ù1^9ŒêB9î,,@Ñ ®’ûªÁQqu_™~ñ Š0xÔˆÊöËî$À,EÑmq<%¹¯q~äî…1ª‘ð{áÒmWúOS¶Ôb‰õö*ióKp¹Iöqî¥Ò\ºîü:‹º+œ“ž‘ÐdÔ,¦ú\[°=ô=y”û“0Ç{BËÒ°¸¦Z¡^Á¼Û "Ýør÷ÊÄ1S­]Ó‘“‹4ó(ʤÍ@ÒÕÍOq9IcŸ5K¹E·òÑâŸ=½pµÀÁ¹ $¨}ë×IuñNá>ÆYÆ¿ E؈Å(ðÏïi#¾ãÙ>ÒÆ<ôvvÔÝî¦fW T§(2 Ë›IþѵÀ+œ7•b2 gÑáÝCvéIŽ‹c¢³ŽgGp­ñ†ùƒ™FáÕQOÔî÷zÀá ‘JƒÁ‡¼ êK†©CÑÝæFÂèg.ÁaîÕÀ\Ó˜IP&#Ã45¤©Ø³DPü PHÝsý`¯ˆówí  ÞAñÏæó¬õq·,?ΩSð¹)‘Døf£i—ìDGîSW‰™¼l‘çwë½úŽ³eG5D–ýÓÄÑòÔ¨t†':‘¥& ÿõ¼w"þ#cÂåY ‰ë¤8’xŽ4°²N€gj+éHcrum¡¡+:¼)wI÷RóÞ`öáÆ7ô¤,# *‹J…h?SòÊYz 1¸1DòB`Wˆ[Ðï—Êßúe6Æ-IN½ðÝúvÎWÙÛyûâ7Î?‚áu©(Ë8ß±,\K­ÿ`Ñ/iõÎÆèzyÑ ŠF\×;Z#`Ô)̨¤Õ,c‘ø„žå! 'ØÇðžK¾À\µD%÷„Ôt5h9xdêÿô*™¥k¢Vyê2–šj;+ÏΥسï¨\¨)Ô*Ã⟡îq}Ê}0«´EAÂ'¼|xݵPˆ“°2´Ò{*Z…µaÆ_dX‚®HS :4a>ƒ. §ë{poÇMù¦¢}ÄWi{Ò@øxÀßÜ´©Pb£¦:÷Q2 ×ç®Þ»Jè3ðõ} ^‡@eo ã±–kÏÒõ[ªþ`=‘C¶Ùä¬ æI²±ªrõ/í¡8Ö #ãÓpëŒi8^(ì™›WSÀê’N˜ý2¾@¬ØLÔ»ÊúBú~؉éõY¶UF™Ëtßl«ž¦ùa$± V¹.­5­MÔ®`]Uȱå]0 `ö ¸ò?•¿ÍqêìUü‘…žÏ¡›°i9QGø}÷µ)ó|d3Îé\ x´¹;¼2©êµƒÞ˜Ÿ„šäö[ÂõÏÀŠ¥Õ2¸?[È…LèvÊR¨ PB÷!~oíãú³„_öÛ}®Õ©½õfRÏ`±rí¨´•œiykŠ®V—ô&æcY2¬F4eŸ ?÷ž[×_ îg~êx‡¹Lt×ÓåqGr0·—¤G^Æñe“]÷9wð‹fÛöo½Ð/¦þ Uy‡ß9!r‹xmÝŠtq–ßü¸¿–¹ós¨èNÔí}?3¹Ÿ˜/OӴĬáÌâàw¾¤&†L0º¡TìB1 CÒ߸\$NNoíûhA±¦›Tż›Ùí–qܯîÒq´ˆñ«=­áSñ}FºoA¸4ªÓ¿q—Sj]¤¿"ªn»c •ÿÔ°#¨Ð°ÞÈö›] h )a¶&Š ¼@{¨8û¸˜¢ÈæW*|ðˆ‰÷¢òQ‡•IóT’ðã_K[ë|f!Ù`×?Ĭ÷ô 5Nú½@ô·ç6äx+ð$™Ìs‡‰ð\pJ9FCÛ¾ ¯¬Œ8×iÖ€…†ŒŽû<ª èô‡WRé‘©fÒe‡ö‹®c» ÿáÎÜU£Ñ7PôòA#Õu}#¸å¸›åcW ¼Ýkê4õzóé»îúe¯¤6“íÚµ5Ï£tÇ`§7^ö4ªy>ýˆ~䳘uìêè ‰lj€· ÈRÙ&çìÎþàVèúÂ\9fªÐ‡q€ÆªøŃû¨z½Ü¹&Áh®%l¯üYÅtÖ80í;/òc•èÖ°õØÃïuë°Ä²­ß<(|ª’ÄY"mJ0hµÍ7ߧúdÀ\1·Á{dE™7|’]®U›‡Ý0g± é~¢ÓYÂUô½]gþã$êÝDÖDÛ­žmûø÷¯«Ã,yÀ' L€/ 3­°a€êGwÇÙ ,ÆCj ˜ b•}IçšôlԴÑ~WßFÅäô„Ç"©—’m:K…Im¬# p:lÏz½cºt°ÿ> S±E/'9ùÎÔ"ƒ^—4z™Å3븄\Fí^G!±"ÖèÌOµµ*iº\hWÀXcå’8žÜîk· |½'imûÞÈO¼:J†ëÕé€úYD€s÷ðçÍ«pUŠÛÛØ?}V÷ãÐù¢ / G\ÝI’`}ÖÇ95ôáI9â2í¨>çWônTEšè?)²™>(ö’r úÉiÕ§9ÉÍüÇ?½G1x3;=áGxæ&ñý,Tüær²§·z#ãÚ#^ šŠ&nr4mS)úœl6ˆ’ÁÁ‹=šýŽ°îêZM—”-ùnè^~¼<›œÿš·w§·¢ª0ŠàdVl©wè2ÙzWÝk¡I."ŒH«ðìžÉN¨9¿3iÛøÐÖ¨‘‘p\ͤÎçê s¤‡ÌÀY:»çXœH¹@£)S_—±„l–Ñoùþš²Ì¾nˆ÷šV—DË=! Zeáò’M²uík¢gk‹z¨ˆx^ê£FÃEtÛ$^–ú@å¡$«úe dãôB|î„ø½¨|g’W7q5°¥ŸÉfé¦ 3a©?%OÈÞ ü©.µô‡0_uÕ—'Á†Ñ"ŸŸ'ï‡üÓvCiŸ>Ú~c >ó.Üû¡žBåoÔÖšEf£AÆA()œiü´þ„ˆ^@CB&î<ÌÝxMù/lÁá=&ÛȯÍOšþºüí:[ËhL>´70&F*û~ògD-,ðZ«ž¤¥ ›~+šÃBéšj¡òªÿ/¿ @õ0Ïô Á¹ò³_’F³Ï¿pÀŠ‚ÁœBí|<þeuî\¼Ù½&F>ºjO®¹üí!Xƒnzm.Ö5ðšî€ŽÏš*¼~#•£ÝÞÊ“ÿÞ†ðμè»ÍBWNYM)a¤å‚Ó± …Ë4mWØr!½“5Ÿäp=˜Ä¼ŒœÃäK1xF›ê%‹~»ýâ´×¢×}ó;[3‰ÝZÛØñ.p~&U~ìlàÅ’Û7 Åün•VãÛ²ïžðÓò_Å™ Ÿâ$fVlE$ípKSó$ú+7˜úÒõ³”w•¼ã´šœ¢}Æ}v]‡KÃÞ´‹æGŠDvµM¦D¡Ôäeö–ò·:ôC;}¤§tû¡k®›åD\’,qr)÷¬Z\¯Z…àÅækøîRj‹hügx"uªjáöûÇI ’'O—2NŽ¿á;ªpË‹%LäIõI‚kÉ+‘í‹RÄ|ø÷áåK‘^ÊO#‡û…ý.ù^·À¯úœeKd1B í·ˆpb6â8áZlkx ±n4ïU¸[{ü rR¶Éñ&ú‹¼§l€HÔž·Ïöz¢K-Š°©ž˜{Âb˜ÝMbSz@Ág¼ƒkàì^²ÖvDBþ“^ä5Ç56ð÷©œ¡’ 3ŽÆ+{9“&Rˆv» þ˜» aTÁ˜À` £Ñ«,ôšcˆµaJáml’ÒGÿÓw¡[:Óz%Ü~w÷W{Âð,Åè&ìlpæ)0ÿ¹öž÷G‚ºê,¨Íd¡û±ª®#ÿ½&åâÄŠ¯ëý°†¿ÉKiµ€XZÔá4ãéÒCâ ö«ÑœÈFð%^G×]&ÆöCÙ™Ïe_…ãâÎï™v{úÎlò¸½a8щ¯P@δæàNU!. .ß\%Üxº5×r6r nÄù+1ß½LµtRÜûÐS Ð €€ŸÎŒÔÆò¦úö7[’°©û‹‡ß3F³øÒú2ð8€¹•Vmè´â""£ öˆärpuî¬Ó6ª'@ƒ3–ôg•3Ì Þõ¹¦u6{ý¹õ$rXD§¬òàž0‡.]ÁáãøWœõþ%ÈzÈ7|Žpä4GïÂå AB¦O¥ëÉpÈ»s¿Â¯“õÜù. ·Bøgœp QÜb_ÒQáhÐœ´)…ž¶TòØ…”ó¶q/MÐ6²~Lˆ¼‡ûx`‡Pß~ya¬™èp.ã-5qqÇÌ‘œ]xšI†îlV¸Þ8Éç0‘VÙ»§i“:®þ>ÚõCöR+õêe.èÏïŤ3Ê Ká¤X ™Ü^%ÿ҉Šl°Î›åxOy0ŽAΣmqGR xÉÍÎ ch¡£zåb.F0Jr!ÝÌ`ÞäèxquÎûpd±NàÄ ¹ÙC9zÝs\KÙ«ô{‹3šˆ*5jÅC/éâüÿM’ >çò ÷Ž$¦ÙñOXŒm\χqõj`¾” „:^”h”Ùà º¨0ŽG‚ì b9΄ƭŒ[–„£Í—‘¨8ƒÄOrõÉf÷÷ÂZ±æ¢6„‘+S¼²×¾t!>Q|Žs­ƒG^AO¼9ÍËè$ƒK7ËÃ2«f‘é™Ã›ùF’¯"*šß}J®õ"s ]…ÿ‹ÈBoìE,*_‚RVvdôæ*M<#+™´¥1’» Žú@¾2˜|2Z¿¢ŒØñKž?Õ,xéuÃ9=²£Ék5ÏÜLç*¡K–oOÀ€•Òôì8@b?Ñ‘d¼û 4Dô“2A] ¦q¼‡ýT%d㧅c2,ÖðYë ÅÞÔO‡^€HÝÔ VÀnç|¥•§n£Û&'FiJBkùà2!îÃ’‚8P¹'ú˜ƒÞ7êç9Ê Û ÈÚ«¬ßẻHêS=±öSV~ŒÒ;Þ­Ø6á‚À¨ vÜáùæoá$hBœ¸ý +s `Ä"»ÁGQ1.˜»“>·‘œJz„´…¡ñ•ìJ,¢ºÌÃ=ó’:£2`fêB­ù‚ Pse ¬[[Lh¬Â ¢Ð9²Û‰ã:hŲñI Â.t6¦A¢¿ÁË {QËøY))ʧ)}S-#C3lÛl ÈíÕ¸W$—«O ¶ÎŒuÓå æu93”3‚hjÏDkÚËèÀpâ{¶QWkîÝ_ªK î‰`w•–Ë Á°ì=¼O­Ï0ÐÉÃѬ>ÄÁ$<Ý͸k«Í¸Ìt…ÜåETÖàKíò“îJ§”@–.gPóvkSJýÚÿ{lw¸rt©xï Ÿõ’õØ„™ 4˜){( Œ¹Žß¾û²R _î*IˆÁ’m»`ì†,kú²¢Qþ?ÅN¥_ômæs¦hf}‚gMü—"oæl|ß‚í$­×ÿ'j©ý‚P¤°Œà‰Û˯æüú9&m?©jÄ)m¼[0ÁJ]úwM+îª-Ín&˜´bàkeò}=܈–ÆbqS«0Õ¬bu68$-ÈQ(qøب¸|*õ=PUÌí÷è3­Ñy÷8/Ô‚¾£´‘ÛKŠ Ûëe÷T­hvT;·}Ô»d{CÔk  /¹†QÅã (0Ò‚¬n)²$\û]­Èjf§hÃç,°{yJV½©:õ“kÉØìòªÊÎk—i­£eþ@ °´¢^mÉÎO¦ÐÒüÕÛ{ÔY±t¾Š¯vÎvß~½îÀ –ÛÕI„ÚÕÅ´Wñé4‹Ÿ±fEÓTåNÖ²5Áþû>6éš»išÂš©ägëÈQO@_ýI€ß_‹e$g 3$½ëbÎt¢Ê¬i<¨°‹MÁ«¼ÿÂÇwÐćDÀÃñ?gÆXß.1‹A’>ßít;…ïª0–Ÿ‰¨p¨–ëmzöúv”5d ²Ï¾ñ…Ží‘›ˆM3ûlçŽP¡€Gy¼‹?íK÷ZÁÕMfµÝæ¶JÛ(”» ­kŠ–U³ùà0ÙðÂ;ï&9©îG°à*|¨º ‘¾1 )Bqe­ÚËf/î¼ÈrËò3½”Êíà‡ãoò0áȈÛø.ÌE3Û~œ7<ÝË€Ä4žž%ÇÉ54m¨è†Ñn$ñK× endstream endobj 862 0 obj << /Length1 1647 /Length2 15542 /Length3 0 /Length 16395 /Filter /FlateDecode >> stream xÚ­·cx¥]³.Ûê83¶mÛÉŒmÛ¶mvÒ±ítбmwìœ~¿o­½öµÎÙöY?漞Q÷¨»ªÆ]£®ç!'VT¡6µ7JØÛ¹Ð330ñ,m]•ímì¹åè•æ®Ò.F6€¿;9¹¨ÐÈÅÒÞNÌÈÈКÄ€&37779@ÔÞÁÓÉÒÜÂ@¥¦¬AMKK÷_–¶Œ=ÿùëélin øûà´±w°Ú¹ü¥ø¿vT.@€™¥  ú]QKZA@%© Úþ¡èjlci³4Ú9©föN›/&öv¦–ÿ”æÌð—KØ`pvšXþuz˜þè@'[Kgç¿ÏKg€¹“‘Ëß3p±Xڙظšþ“À_»™ý¿rp²ÿ»Ãö/ö—LÑÞÙÅÙÄÉÒÁð7ª¢˜Ä¿ót±0rù'¶³å_`oöw§©½‰ë?%ý ûKóu1²´s¸=\þ‰e ˜Z:;Øyþý—ÌÁÉò_i¸:[Ú™ÿWt' ¹‘“© ÐÙù/Í_îNç¿êüoÕ98ØxþËÛþ_»þW–.Î@38f–¿1M\þÆ6·´ƒcü§_¤íÌìÌLÿ¶›º:ü'ætú×QýÓ3Ô“02µ·³ñ˜Íàì]þ†Pýß©Ìð?'òÿ€Äÿ#ÿÈûÿOÜÿ®Ñÿv‰ÿÿÞçÿN-ájc£`dû·þ=g‘àï¬Èþ66FN€Ž¥ÉÿËÕÈÖÒÆóÿäüßwkÿõpþwøß!„íÌÿ*DÏÌÎÀþo³¥³„¥ÐTÑÒÅÄ`fdó÷ðþeW³3:ÙXÚÿŠü¯óýëÄÄôß0U Kk»Ô`ÿ7´3ýï5üÕí_0ªJ©jjÈÑþ¦í¿6+þí UO à?"iÈÛ›þ¯Å?T""öozfn= 'ÓßËø÷:r³°ùþ„ýó­å\œ,=:L LLÌ€¿ÿÿùû¯•Þ£·3±7ý§T\ŒìLÿ¶Þÿ2ü›¸:9ýUü_Óàoåÿ¹þ×%=€&p«Kö&¼!VÙ™.õßòG~‰é ô1ƒ„:”7ª–ü°ïõψØá®6|¯ ehšæùl÷\;§H>yz¤îýÙ€G›KÎkôÍ/õŒ8ÅÅÓÐé¾ÑäòÕÓÞÑ=óÕ4ÐŽ$Ã^ô§×Ü KUgù]ð´®HºsãªE“×õȽ…(?) ?ô=‹Ó'P¢ž4Î$ìöå`€÷™ ?u‹VF-«ÛM3à¡q% ÑG‚‘Š5ëBÚãÖ¨@ÃH“Í¢•WbJy ^åQÄU©åÄ#©0wÙå×Hálo,Q]ƒlc©8eâÉ[TÑø6–û‰T™ü¨ê +ªM9šWbþë)Éã…“ˆò3F³ŸëELeÕV«1·^ï œR~ÓÊ=è…ès¾úCÒ5 >£ŸZ™[~‰ßõ‹nucWæ­q̆Ör€]%çñѽ¥*ÝžBhÄÅÞü(Û…EYè&z‹×ÏÞËUe ”½3|;m0®nüö­qf9xN5 F® ß”F>'Á"R ÑÏp¬^ª}ŒdÞ ·d #ºÔK§“mNË{Šà%~IìC€¶¯»×ÄÏIZá<„OCgÃÔc7‘>*ïGEÉUúJÇW7£2=ÏÀ4e!v6‡~§0HÄ©¶ÎÈíbó‰D÷}?%h”jyô´í¨ØPÿ>ß{‡e üÏ?'¥hžòîˆÜ¸` TïU߇¦®gКUVÞµ4£“ÈÕEýž ¹ÓÓ¥äòË‚\Ñí2~%l (cÏ ‰hmk›êE{=ˆš1qåÉ{‹¸abž t»B¸y* `ë-ÿX€Û É£Ëjd–=µ²žµx®‡ÊurXôœ`¾k&úD7Mƒ9ë§I§ b¶B…°P‚¦ùžM_s#*@ø$µWœä89N×Û°mÔ™ûƒÑ;•®æü³Ÿ»'ÒÀb3Œ² çäì(t¨*)á$„þ¸‹AÌø´EÏÔ"áŒh‹°»àò3Ë“’”‰HYNXÅxÆ)»ˤQ"ônò²?î—&¿‡Oekx䣫^T©\½­èt ct_ß“„F„ÄÇJ¦#c»kçÜ~ò¿É<Ëø•€(âò²°Ó„ØrM‘àýRtódkoý<ØF“´Z”°µ2’ ½¢¬I,ÇÌÃЭ$iéìX¾a(ßdlMþjé>ËJ˜îßéJÉÉvzÙ²ŠßG†mØ ée^»íèüs+˜¾.•¡?óEŽåŸ½@B_Ï -¥rJ‡Ö`E’z ² °JöP1-}A¸€ÉÖ¡1ÝHÄ…l}›ïÂúS§¡r„lhùâí Úi¨xªŠ¥U×Ò=HF|¬¤ÊÁ˱'ðwV 'ÈÎ}ƒ%^h˜ ±3Lðø/¹l¾ áv¢Ý œö¹3ÈâóuE^*s-¨Ç Q¬œ›î×zŽño×}æhÌÊÿ)ÕEeWXÒ¼¿Û™@h³Ë‹SÆtØ”b`4›;¹¼OBYËw7`±Í?Ã=¢ØLâ‰Ý]dfª¦©¼´\ªtYê JâõN~ 9×s.M›ðQIíÙ1}jÓ.µ\SØÌçyqÍú•û[Ö |¶…ùî•êçH ¹£NÛx„¸Í/²6¢Z­ZÇâ3‹öË`œÖÓCÅ›4MI¶„ó• b>„ Æ¢q‹kŠÜò¼„  oÿJyΔW O¸-9ç(¤FšëûÕ±çY¢,Ÿ!¨¥‘è×$G±]¢b}q{çîú¸xòT¾7_G€<)Z"Ï÷ù}Óø˜ç4pBoxO^ §@ËPY- XÚoª>'ÙGsÐïùòÇ:tƒ¯L7ñw !Ÿ¨ßR‰w£GgX¯òÚ2¦ ç,»…ƒÿÄb ,ÛÊ™Ï~ÍŒÛèc8ˆåÿ,®Sš‚¾{M® ?òT½)‹,jÞcÐÎà¹ÂzÎb×Ìþ%~uMGÁ“§ˆ[×ý9 ­ãª¾U* :x íg-EÙxªÛsÕ¶B÷*Ú6fxŠ¶5ÑgM•xx¶gçwsbyW–«› —“.iénmZ cm?iÕY ¹eŒ^³-Z«™ñÀ–¼!èl‡@‰Ÿi?tPWמ¯t±Pº`ÑíA è‰~>Ŭ ¨6oñ}æ´ùZ%Ž¯¨¢ÝÏIå å°ÞV·Ióþ2º+ãâe&ñ)ly³K.õ]û„AÍÁÔ¿ä·©B]4D$! þ•çà#ì‰î Ùyƒ3Hҵꓶá¶x@u¾wÎ/¶öíw zòöV=øxñx's*o‘\y̳›«/z­îürMôïëáo?§ž{†&%Œâ_ꃱ²m8q—Ï¡Ü*Œ: ž[ö´b8zËüp‰ Zµ? ãŠw{„æqá:z‹ý¹l_oܧ„E ‘˜B¿ÍA6DD2ïµQÑ^:'ÕÁ¿Ù¶Õã‡:;pð+ƒ°s}C²’‰ Z•Z ¶ð¯Øâ¢ßñ ‰ $F‡ZÓο·_rO Ò]žñè‹Ÿ÷îÿ4eÝwäwôÕx†£´á?ÆÁŸÄ’_SŒËŒû-ÞV»4”‚°ì@¡×}H¤Ý$ËQó¼Úïza[¬©ƒ ÎNâcýÕM~_M‹~½L‰’#z߀&X$—.UþÃDññ»%MC`†Ây!9`^'R¶Î|D°þÜî>Ÿ,ÅØÜã’TyªÈÙ›–ÆóúCöÀ²ØêèêlGS:&<1‡\B9{Æô½ùãfÓW Yü¬âçWÔýêQ´ãeB¾ñê3¯Î]Ë!™»¼h&–¤‹)Û¾^Sf•È™ µ aúZ†^ˆ–È´g ½º² nw£'ó9$%+o¯@,„×Kª”¿îdÛé¦æÑÁߣûMk¦øç²´ˆÔBO Bï±s7¼©§‰†ù5{7ACÇŽé£R& ¬A µ½ÌbÅ®2Éóý[ì›ôCR»^"õç·•0lÎVÄh¢ÛÒc²¨ß¾À|„v^¶’CŸrµ"3Þk! !AÔüÁùBµhØÒ—î Þ1¥t8ˆn‰ ‡ t¦ ?&OM kü4ŠË|@ÉÄS¼î×~ž¹½‚Ë0ê‹/Uõ´1Kß²7»®¡pEæIYp*©ˆw÷F.ÓÝt[9âsœüÿч¸¦Zº Ñ?“HÌa$-èzŽ¬8¡F-6$&…âÿÝ ‘ÄJé4w£sò˜®Ç´¶:ž×»w¸Á»·•á9PèŸ ýèlÎKVF $ì*o1KÍS`YBx5§]«º|o†®%¦Ú7˜Âõ¤¨.™5×,!jÉÂdÒ7âü‚²ßžoÞ¦SùïÕëŒû$îi¢%¥â)`jm8Õ°j½=&¾Ž„õÝßåƒo&uJâ$Zt Õ5%’½sž­Ò5ˆòL\çðò%wÃP?XCtÏ÷”±K†4—Ôûˆ¼:U‘»-\RnþÊnR0Z u@'¶×PÖtÆ éq:MCAœú‘ÉNúÝ=Ç·¦ž0N¸ÁÜ Ò𑿶"?þI™C4ƒ^ï$‚tÖù^2f(ùêÞÂÔŠ_î%'ÑwW@‰@šá¢úƒf/ÿ»ÅhšZ|Ô´²´‡?uLAn»Wy0b§¸Ç*"¬ÃÓH3Ô¤u·£ŠãÉä½ñžãÐø ´[’Óq]SBº¿Ï„R•í1úQŸ¢, ‰ÖLf€çÌ÷ØM¹q3Rk0{^nú­£åuÆ~1‰¹ùtIdj©Iγ|j|/•CÇ>Mw†Ž‰1´Æ鬦x"cRGñÃ2+;ÙG¬‹$­ÏËU¶ƒÛäODÃ@ æqoz£Ä]Èœ[¾%Ô÷Þ…5Wò‡:më©8!Ï–DY¢Å9çØ eÈcªñ³8e)¹\¾¸,ÉGu„ v»‚#ÂÝr%¨O•AÀ^4äÀã=s3­Ó†GóKŸçÒ´Íq7Ë·!Ò”(KKÀ†Š¿¡žÛ•d¼”v¢vb¬¹~c.¤,Ñ4aœ7^¤¦v­oÁ“ècÃÎ"4f®K@ßÞ¶K%x^oVýë»ýíâ+³¶ ~¶¡ù\¸ÌoÔ âDˆâô3¦nuÐé–YÏNÿ«2ãGÝ0U´(,^«ÕX=Ã_ÃçèS#‹WÛá(H7ÙðÅú %†¶PR$Ò1Û°m?;麤t>í´ruàx -Œã8 K•÷ñ¡#¼‘½žÄƒô¹N×OÖ·¹w-Ï<üœ Ì%½­d£ïÌÒ‘DÒÈb³]d&ñ髹‡í+¹ñðÒ@RÃrù™v Zµ5d—2œ²„Ñ TŒ;~p»fÆÛt2‰‹ÒØ2(&œe5l§éÆ»w#†ù…ç4#{@vÐd×ØÍö=§”:²6*ÜË¿-4…¬.yÝlü ú½Ö(ð™QTa7c“[ÝÖ{ÌÄAø›PDCaÓIN?¨ å—êõT»ÇfoQfK »Ð|£mjÛ2îÙ‚·\_ˆƒ-5—Z*@euW\[yó†‹ÿ MuAE?ûÀg-¶³l\f!Dtü9œ-kØÎKº³þû7q®' ˜î'y¶¦®ŠšFéÎ*”—}"%\ö ,T¬ !C`;9xb«Š¾ì²ëÏ©РÌ÷P³óŠ¥R*EùÔ• W±{&¬NÿÞ2Í=ýw:óf'âä×ôÛŽ~ f¤-kHÅ’$æ¸XºEMÆÉ<á*ºß¥Œ£çªd‚üy »ûdÁsqÛs$£–XE‚ÉÖøxýWõ´e>ÖÀîv{l¦bKÚÀkË¢Ò‹á«×ˆÜ{¢o—1î!\TQÓ”XM>ˆ$räõ†Õ-¨\¬ ⤀–u.â8lÞ6=n‰ w?ê.{®-Ñ%•3 JcgŒ›«:AƵéPÖù-Wè(Øt´­ˆ5áóeóA°_ÂÕ?¿÷#®_xøyçÑà;ç¹Üê†xdªIßuüï"AxÆô—мÔO‹ ²aøSûÒà$ÒGÐÂcP…Ý­ÑëDü>ï1m+ööï<ªªH²¦îðQÅ°Ou&àPK~pµˆÕdxRešø¯ŽžÎŸà¼ßf32fˆí?pAw¯Ä2ZŸôZ”XÊ‘©›Ù*Yƒ–ñÄÂG›Êk€²Åæ– "YçÍ©bioÈe ³eR&É’õ£ÒA²~¼ý®ƒÅ8‰œƒO!Ë9ŒðrNR¸„Ÿ¹WmF_îIñ®# 2­vœK˧cA)z?–ßôY?”ˆçÝáeûK²{~¶£kÀ)5¼Ø21<'òY· òg®—”‹4eòÓ-à»'ÚÃNÜQ&Ræ@O_F„*q˜Lô•ƒGGûT™7‹òñÙoÑ™uG Øån™ª™y^çkc¾úäíT2öa±Yö)¬Åk5ÞrÁЦ8ÄÔË:m˜"Hcä9A†òP`‰~8~7)’©ŒÒ¼]ûî¯K¦-7™+ HÒæݺžÍ*DËb\ÚRF/×á°B£yA9×î°Ü˜†×VÆø© ¦ï$/ØÂTt[D5e¿oo†¹³C\3ë‹rˆ£„È¢£g"/ÏLx]kø ÆÏ «JQý¬K.5;‘EôZæ"JG…ì: ¬Í}‰Êÿ ºÆCFMÖœµ#ÎÃ…F˜{êeìÛ¢ŒsTùÛ:} «É€=*²Á  OÊXDaŔۺóêáqZìvŒ7ú„ê2é»ûǰ*Óß/¤ŽŠn.<ºÔUÖ¯"Hl‹ .‰¾ß×O:\Xû’óžâ¹d]¡ ׫¼„³_ù©àªËMóAEëÙ%XÈ•ˆºÁ‘²é›¨½ú^ß^êÿAs›¸©vå”_|ñŠO3k?ëgú;d˜®ª6½â§vxýQ¤ wñ}hÎánÜš"í&¼ ;f³å¾1 ŠŸ]„÷@jì; ©:¹%4,VÚ_nÄùG¢ã·¼K+Ž+(Š—úßê¼Í…û’Ôåœvè§ïF‡iolÝ꜊‹ß¯ÓG}™½ýtB1¸ßè¦ô0süp¿o’VÍëRf€å@çWwkN…*B髬Y«x²JÜLëÜâk¿}mµ1žy é‚~¦SÅ=0É—·’™û<½IW¸†º£ôš5¯¡Bæ"T„õßÖ¾§"Vˆÿof‹¦ÝZÈÞ§ öGƒJ£R6ðYÈ5‚ØqF*»)ÂмO¢G¤} yH{îS–ã®óFŠÐO2_IÚר_b¦‰Zˆ× <µ)š>x!ÈüRD•0¹‰WÕ=åäìÔÒûÛ?̃Ó‡KÇ>²Ùo¸ò£ dWþ…?Ç—;* å«ä²‘˜Å¤ £vdZä(œÉÏ!÷~^³ (Zðz¶”‹á^>îÁõ3ÌAú …ÖZ…×VM3Ëÿ¾ ò=õB%åã“ßN„Ì$4 °€íA»ÄÂg ^3±#{»w¿˜ûÄu¡9{ÆÅïOŸ‹ îã¹íÝ.5?xÆ ®n­‡ç] §Ç@jé;Ìh3!a­XŸ`)¥ª/sc?¹OFmð™.e´d1áØþµÖhDáÑKVÔü>Óò1ôYµÐ©(¦´à%ìŠcy7o˜Ã1ú˜õ®H—…f£ÀÄZ'¥ÈSÛ®Ñîr‡iL¨ªë@M-Po'Ž“_ñ²aÓædQ Åu¨~CZ¬wÞy®¯¸dÝ÷fMBž{ÉN|Z£Êùƒ"DqÑ<;…jk"/ ýö­üÄbKÇ«sâ4ÜIþºc}œÇè5B·¹Søù¸Çýr‡R2\³F›lîý¼„Zþ[~ãr2âÁàe¼ä'Ò¾'ÜüJuN&™JÛ;uß}cûAØÙ»+ðAlVøÔž“Χ©Q;Zw.Î×úPævÉèx“ª•áís½æù¶aévdÒ>f¸0pC1ŒOuèoæéÇdùƒ± G‹(ù[ÙBúé—©”`V1°_eL¶>>‚ʼ4K}ÚpÁf çcƒ´!T$W‰ªëMkd¿ÙÒÏXÁÖ’8'±€û_³9_þ¥>þW-a ÄC½9*¶ZdÙy€™ú‹%æt°ÙŠ,e2ÌþU_ÍŒæV„e|hdV{:¥Ãûžß–5/–öÍàU)“aY6d@éˆwF:Ó¾9ö/óÈ[aj……€+–¬ãå8Ù&@Û•‘Or[˜¯DÇõ hŽtÎÛ —l(ÛÒ¤^Àc^ä»ÆàuE 6yÏ™¥<'bÅóA¹{–vÝ·ŸþeB¯Í’ûí6ÆK|OéaDzótGØÔ6‹-*jײ¹D` úS£ô UЈ~(› œëÀroðû":@K‘°ÈŒú‹} ܧÛΨXÜŽN;±ìÌqmý€p>“€æJ ÉÇ7Ö†÷é„&KÑà-?À+¸ ¹´7,\§›øùõ1=<|l 7.¿ˆ7—v±Ã»v„®Âí/ÓZŒÛä ;tŠlËG¦R-ÊñÙw—ó%u(.¸Oˆos9£ e§»²!ׯnWón˜`µ*>(.žyå62²ô…´ =²´GÏ—»::õ¿-nr|4KÅ>^YÃ’õ­,‡Ã¢(b —5¶0YÚ¥u…Ò³ôjrfë¯>Td`Äê˜ýüÔÊ`•iF¬ÂƬõ¼Vb“ûƈ\R¯}]8^d|˜–Ú³à“¾­4åæ1~pç^’Ñ7¨ò¹4ƒÿÐE´Å]kª¼C ¼suÞÿžúÅ6¤õM^†=å½'&%Ð_Õ$«s¬9®›s´åHÌpðtøZIrOMÄyå¥Üïûo}ùâh7*ì·F*ñ†1ɵž¼ltVÓ¦²9DnݤJœíîºÆ¶ÖqZKŠ÷ƒ§ñýùÉpÕP©EÊÍÞn^ß²aì€FËÉú1žlžÚåˆn¥ù„^þö,Zø£'é8+†ëØ:«j)f>è0 @Fr5§‹ZÂ@ÁïÆÁöú]Ø¿í‡î¾7NDÈ®IË!8ñ(þÙy¼Ã…ÁÙ=¬|¾ÖSt+ôÌé"ƒ%Ù³}LÇÌ”¶èhÇèÒ¼Ok~*ÂP°”Ä¿Îm[njcq‹\;µýÎ¥i®:Gâa›œJäd\ÁÀ*ž›iEåX0I©ÛcâÏ w…+:³×ZÏÝ®ëÕÈhHï¡C(»öªûR‚½Óó=? æÓtiw$$ÍEÿL-[à­ptÝj:ZM½UéÏ$â×^3wÛ r š}ŠTlóTÚï-,þbV,Èw¸ã¿£V¼©‘ôæVóT÷¤€z žMaRl×o'& EFjëÍß-œiÙ¯ZßRöƒ¦–ÄÃWïà’—˜µ¨É.œtê€ó §ç¸þQ&+ 7Up˜ž‡·­• F!ÒÛ6FÝ 5øì’‚Ä<+½ï/5Ž†ÉRñîȨ?Û+6yÑW–º± BÁ_¨üD0ÔñÆ_ºß×?Ê1ù²ox,"­œÈ£9‡k¸Â·÷ˆ¦’&\õÝ.8@P ý#àÙÒƒ|+º”\çŽY{ýÄMd¯q2‚Ó(Ãbg@YGÕñèjßI7r¨µHLäÛ¼ú‡Ö/M÷ßlp[b¹Á‹vnNGÆG×DInlê¦ 0²lPcûvY¯¥„Ù±óø„ÉwiÞQ3¢9ÕŸgRâ\½wkÎØ7̾mŸå¼EµQêY‰¥T6Õ‡0©äÓSŠzƒË÷^-_-ï¦dÈA™oû·jñ¿‡k‹ ®ðdWqãÚ»@—"@-Ü{¸,pÌ>"¯EX™›Têj†Üó! ßÓ´ £ÉôEì÷}79ŸC3Ël-NŽP'¿Íà@_LÔ[ŠÅ£ H¬Åé4¬à3_ˆ­ª>ädz(ôØç¢VõQ# ­™?pMýT1JЉCþýÐmñ+ÿ®›ÏÅÃÐEÓ.#úO_=Æ^‹Ž.SRÊwžõiè9•áqJCS;sÀò ‚þ1r:˜°±KOù7N!u숙$iæ‚sfÅbQ“.öYå5|þ–c{«•q2^µ9Ÿ ‰_°H®ôS®ôkx¾Ä¡tž|ŸË;j1iå;w¸R?7í)ç‰_(Qûb¹’\—A N7 Ê|̸!“¤L‡˜4`[÷ Æ†¤H‚`h¥è7½:æQX ?h1‹fˆ™õwüÑ‹ËXßSû»¢0 6Áy ¼ð‰„J µ´BŸÍR§çt÷rs$í%Ài¤;–—k¤.èk?WˆÂ»¦­ŦŸT±MþU? `¦KRÔ×<²*$Š-ïö½'B‘÷­àøÂLÒ>žWÏ/¦³6SË T–/o@ð×P¾H*‘ìõÛtŸX%‚p)aÿ=4hhúád›ïÝþCªÒ1Zm°Òˆ¨Á,¨©ô÷äm‘`ds´o 3j §¹KÕÔzí;2!-œŒž&í‹Zo:·ѶµûueÅxÂH|““úÅîæ‘r;wNg¡ ÞÉ›cTÇ°HŒ;³fðGcô"ÚSí•_jéÇz3j~5ª>øðñ—Î üÕš¢VÒÃ\†žŠU·¾Äß'‚ZÓ¯k kczúº¯^RíªôôºÃM“çJr 3Í 'Ke×ʪt^Úq  °HGLón1,¹$Œ{¥=>­¤¢©¦I¹-çS…I";ò¾…ãC˜ SÇÆœÜ$§SÅŽ*ëßÎÎê±übˆ¬?€¸]ä"þѬ¬UÏ]çÛÌ¢­ˆ½æ?2ÀÈÚ»64ò‚8ÛË]œy‰ˆé³†¾‡LÖ'³×æ‘žTPÓ¸¹<[ZŠ±/ Ö ã.¦‡6×™ÌÞ`Õ☈¦Œ9ýž|»t€ˆ˜~ÚY° µãÂßÁí¤º/A¹¦ŽŠ`ÑQ/šºX&‚›¡%!òm©Êñq8,fh§Ã 5mMŽná»ÕD”ñS;«°GŸš¡A‘+ÌÀp^›ý½¦¥C?Ë{ïµ\u3vB¥¡}Z)t¢ŒVG¥ÿok`ÓtEá ¢ü.à’´p%Qw4Ä·0¹%\¶ýPëPI“Ïð9ö&Øi‡:Ó´v¸’í=à)>}`ѸëÎ.j3Ïó3`ÑA†‹4ŠÿYù­ƒ,l½%AIâ-ôL·§ä—ÖD >…y{8µy­\ºêÚw½Ï3䉸¤!-iŽÇn‚<|ÕBH N¾dî ò|RåÞd÷¾Ñ©¦0Â%õ œñ’ôÒÓ—ÓÑžf œŒi/š2¢;PˆŠÕ±x¿½…Qd­?eñêžSm7ÉcH£ÒPßy‘4K˜îZ"(s´ëÏžt ùYÒ÷0@.ºDqªUÆ-qª{ð“x8LàOë±òRFT¬÷‡WLÐÛEY0’–íCm½Ö(®mzëXÂá‚Ô@®¦É[S–ûV²n„jù‹¼AY„¹ŽZEb7ìÉ„SŸ”þÞ¶±•x/±¥øy-ñâ 1w7Q ™&M¶ûä›’[W^oTx÷ø¶š{îøÎ$ÎÔ@“‘/M-bÅ: ^Ò´K˜&^Í$¡اå|¥¥å÷^s=´ÕªR úêœ¤tÃ/ÉogSwO$׻ĠfX"‡¾QÏ‚ûi®7Ð?ÞZýàø¾ß#â/+™Âå ˜ÊZDèƺ7ÙÕõ…^àµìfŸþdw<ó*È¢|™QP2D¤^ '&(hÉ,J!n¹·F‘Ì;ï;%jߌƆNÊJsXš«¦ `€FzûHßtâ16UP þò0mŒ0á5ÑŽùlæ62Mó‹ÿÓÏf c(M¨†fŽÐk›m¥+j]ž'üœ,ÌSf M¤t 3¹‚¥µµ[ùü0­~Âébl›c 壸D¿´nñ)õ½Ëœ—;lNKÒäƒXî‡Õí•ePÔæ˜Pþ±"X¨Êëî0DIÄGbŒJ#lº4>2†Ïç#8«=ÃÀ4 FÓJ—Uéi¦G`N ÂÎ<ãSp¶a£3«/\¥ß,[§ ?‡f®O ¶ø3W…Ûh™]Çh Ýt‹ü0¤²÷P>ü³â·_LåHQ6_ÌšÕ|P¾¿ Œí;2¦ CÏFeáy›ÈŠZKÆ–sÿ‰ž¦È×ð3âLVt-ÎM0zZÇ®¥Äâö^ëºRé‹™`׊}Öî%sµwþ‹-Ùóšæ0$d*îêDDFÑžî·m97^íš°­óK B8^:tGú¾Iz{´ôqTù`ÿ»h»¸×‡¶Sûyv­ßU f°Gåcž¦+¬Lòx#¬LõÜx+ZùIM¸ê—2ú§Ä›QoFZ{oý?ü³Í»ûMßÉÌØG%2Rš—9GË\Z$¥ž€èy̼÷ Ö9­š |µm ââ£&åƦÉÁfË_‚ISh}34ÖËåt™z&£¯M‡²â(”~¸rUNÃØ@@d“j–©áB¶ ±Áª7'‡YuÜiþYýÔ_˜ë 7ÁgîÝÅò ^°yÒ=ûF#ØÝC–W–Ñ_î·ƒÚ…gRB/ôµìduî×-:•˜$ ZÚËŸyDÅUÜÍï¼ÔîLÎ|C&=JX”/ý‚Ô¢g#(hnmU2/f˜($MÃþóëT`ñ‡Û¬ÆÅ錥I¨aÎò\#8Çy—(mkOÒ5¿CÓÈsõx´á˜ðöÊËÓ"§±A¯»»ˆÜÆt®À[ªI·„8—Ù’tT>f-Ÿ1r"ìÚï±%ËX–É™1 Ó|ÐVÅ]öŒÈÔÈ¿ÚšЪýNg™÷'Ó€¿d°‹Ä–xHŒ|§Weƒ ˜ÅªÛ¿9ê¼ÑùS#jjøgH6—î{»--ÈY „ ûV7…Æ_®Ç‹¾pƽºNÛ5I].Æ°/Oa¬Z¯ïÒ»ËVd²ýíM cŽ|þþÞekÎÉLì»VdN’÷^Ö eK—¹p‚7RQ´ôéIÝã˜ÜuŒ ¾‰øt 2“íCž–“]gÃQº3ã Ú³Mo¶`Jü© VŠÒ‹‘µ;»ýV˜¤¡Nz>y3^güËAHäÍœ›÷˜bˆw;Ò“ÝŽ‰ÜqËž¦n]Óš7.5ôñuâ»yðޠć¦ðH”ÈûÌðïÜ$Ì6#Ä°C½BqìEýhò̸µpƒ@f.ÖSÝðzó, þºØ}¤L ÿÙÜÂ#kÒ µ0 º TûŠ¬.Ö9:³c±VF*¬&Äô²¨õãF´Na.Oób‰|ÆËC³„¬¶ÏóJ¨È³+U<Ù*ðî¶Qq÷‰L]ïÀÅ—ƒš ¼Ò _8æ´¥9”A¶¦ÌyÔ†CAcÛH=W3Ïš]ôdU,Qî„*}•Ò{' ¦Ü9M‹fˆ7Íßp,Pui_å8qª;zKVm‰/Ò$¡[{F#ÇP4›ÃQY Z8éó¢Ù#H*Y¨ù(=·M™¡b8£Ã„ڥ̩;¡wƒ¬ª¸uoJ6o¡l½Ùìš¾§9Ñ,©¹4³ƒÕ-îÏÙ›„ƒ"¹¾m½+3Ö> ¹Ñ®'‘ jsó£õj䫦H¼ËÀQu®ò ûdr9Py°l!ëx»Oxx±ÞÌqsHL«·º"&ˆ°g÷’W5.8ákªƒ—Öˆ7Ÿw kW>¡EÌ^ºNƒºfžþ·:ßÑ1½bH@¶\Ä%«Ô•½&» Z¦°R&…{(lõ½4Ù¡ÆÕö‘ÍtIHÝ£˜ @µœR„[Tõ¤ŒªIÌ UàÕeoDÃíáˆ}FC÷ãnŒßvûÄ£ŽG^`Z£)m´ ‹Ao#?tlÎiiøhæä}ûOéÝ6{’HæþÝ¿°f±¦ 7Ïf@z7| pÛNòu€×éÿÐΞǗqùUðñŒkp§[ò>/r(ÓfŸèµ·½úÖ¦th€s2‚(¶RÃ…¥SJçáU9Ù$žàŸÚGKì*/ti㘠_‘H©¤0 ïs¾PèJ7¯iá%ÿa&g›É¼'gý>n*á‹R†dÓb°ÄÖðsd*·jƒ"£—¡]f>…`GŽ¸N^‹7~»ÞP(þO¢ÚéÖ„`ÒñËlßÌ ½<òմ怪œW$úd}ÄÒ1óVã;(6\‡zU‹ìÉn»%{•ýivËÖÏß&¹ !ô¿nÊ[*Lî¦Íö\7cîÑ™ŒýéÁ¦’÷ÓYÎyj\¨Z¦¾FÔ%c‚—C³[¢ÒjÔÈe |ɳóâ˺6†[V.…ú%ƒJú0Epºpþ š’·Q ·û*p¨ÿ&—¾‹;XêVî‚[ÿ¸¶„õ‘d³D fÔ½;°¬ñÐÉ,%ß×Ëÿf Þ¬¼œcGÚ{âÚ šAù—×VL óÇ'¤ÆÓ²ÀctÃÄ×u#ñ°m¤xåêš°i+[œs•Íõ‚uÖqgó´Á¡³û–’däÔC*Ý?=ÊYd>x ³vq­x;U>€ý™&Æ\`ÃãYwf›n‹]çã¨íºoõ”g(Pa”‘”¯€™‰q/±ƒ LO}ØŽ!žÈ˜€L‹SWС¹ _ZÝ¡}Jdìý€gi; Ä9øòpŽqÞé1×# í±ô„Ÿv¢õ±”$ôì’… ½ZPÒ7åëâ‹oýêï?iíµÞ-r:œ- NO²&Ei\*6{ÔŽÈ,ʲ&NO\9‘UóR»ê›`ýõØjœõk|©úäX¾Æ„Ñl%ôÛöXxA×N͈;E‘Ÿ¯äÕäœ"¾)QyÚILÂÝ“U°•\š$‰ÿ¼hìK È=ºžX=eÁV(mjÐþ B¯>fÚðr÷´p±æÂ"þ3³ð±2žÊãŒ*Âí0#°ð•0}Œ9ì=×5â’ÄÓlnäŸÊã‰2}_ôÅ'ø~rkp%·½Fš¶›Õ Ñøñ25>ÄÒvÖKïÀ»DpÕ¨Á1o-!*O€¬ g]žTÞ"ÿx©ücD#Pæó(KrŒ, <Õ¸Â,ó’ ÛGù8|WÚAQÕÓ_C­G½œŸÃ°zØG ˆ.<©Ÿào”éò»’Ö¨Í)™¬üÀdÎ.øÇäpßé†8ùü6-¦z7kY=}\«¸ú×Ú)ÜN×Dì÷Ÿêytiu“ 6[ ¯Âþ.yXÌkÚ²ÈùB"ç– )é…e°Š<¨àR)úu¡cûØ|`Ó$Ô–™üžSâIëk/ÔZ毥3œ^UÚÓöª%„·äûçUº.Šñtd…&ùk09ݸÐÏo»¾‹û­£š¾ ðSרLß|ãË@Rß"Ю{¡9ß«’|o±‰TZB¥;²†Z˜£±%K} aYeÖs{/M@¤ðò\LE€„_K=×"Œ‘W™næ£Z¿ é`Ìr_ƒ6ÖeùÁÛf›Âè@„úXȘfªý\pžW S„ü ÍL%WuŠh׬±ùkÝÜlHÞþ\”©­%E;«Öìí-J—ÇÍÛc—;¶º“\ƒ+½•ˆïmÊSÜ H£wê£<´’ÕßõƽH#HE:‡!´ËC‰;1¾Ëñ¿XŽL;); £íÃ_ABþÒñ}5Gë]Ð3CÉÏbj®á…ÃZ{v¿vµY¬¥I×%…¡î`úiÓ«á&©ìhí‚å±[ÎfÓ°”E¬)ß$ª± ÒdïS 8 i„­ßD‰–nvi«fvð²Òê¨Ù¯Ÿ5–EËØ=÷-kva¿˜a½ ·ƒÄ¿\å7eƒTZ,K-vÙÇÚ MBz¾‡R7Ëù¶®§ fÔ ÊËŒ5W1RLúïjÞNŒ uXì=ctód/­—~”º³®à«ÿˆ”iFÒtFüˆÀ­5T$‘þæ Möl8jV£68AË!¤8áYi7ßx%Ð,g GÀoÃglÉKwö8ÿ•¤»;;3~Y.#¾YtÌ û å¤büêÙ€ú“Ô.ýkÿSXŒÅ‚4VÅňп²o"ßÁzÅ&‘N«EF_,òDê«ÈÐltŠ’œÉ€šQ²¢†µÐ7¿)qâDL¤MxÈÉ ºÑÊs8Ud\u½EGuœÙ§Š~稨œ1N3&¿¶Q{!¡® ^³ïåqW}Ëø2˜U·Ï˜#6µS<;Ñmz¢Wš3<¦m»¶~o¨}, ]Kfx‚u`þ¼×XÚFAu#œ6˜ˆpÇy7RE4[•Äýh¦ªTYpæ=<’ê—Þ}§‹?d£Ü K±¼_m~ÖFí8É-iÐÛJÜhÇò­˜Œ–EvP^~e®ÛÌ*ëààgiéZýH~3¢­º6­×§øp•þ¥ÎˆÞ8‘õ]|AºMs‹+·G÷+BM¦Íù*cJ$h§Õ«{=Qóàç8¶*|7© ¸„1³»‡|{.R¹Šn‰R0ŸÏ.Ù§òËMi8ÚƒŒ ÈÕ¾°iÌ„£ßÌúU[{ô‰eeÏÊÙòÖgZ.¥ Ýxy=0g>]ÀÌYÙ/E|µ– «„¹hÙ¼Q$sº·æ4áW Ê.GhŒáÿ˜’þ endstream endobj 786 0 obj << /Type /ObjStm /N 100 /First 905 /Length 4240 /Filter /FlateDecode >> stream xÚí[YsÛ¸–~ׯàc§º";0Õu«»¤¯“N;›“TjŠ–h›‰¶–¨\§ý|\Ê’C¹33/] L€888;@ÈZ—¤‰u"áÒ&Ü 8Ž¶L„ÐxªDØ4á–'’kr4ÒOv–Êä#?ää B !< ¦äRQ{”üòKÂ^%ìÉüõí…55,ÿ>¬mäÜ°ˆ¡‚àïæ^I‹ÆbȲÚº¡@Pè = tÞ=™ÏÊ ¶„8ÒouRÇǺá‚n«y»nz8ŘG G±†× Aƒl—Rø¨à01{¹œ^åÐiÂ^Ÿ$ìu~S&-M•™¼Ì®ò;}ù¬\!XÚÈVóõr”ã•sÕ«çù¸ÈÏo’`>r–\ˆ‡]¼Ì–@oT<–’ïg)uxŒÙºKf±˜î;ºÅ> ëþì-óàÇY™'?ÿ—HáÅ «G*ë|¼·ïuQNÐs–ùš—Å(°_g£ùèªI¥J¹åÃÇóÉx7ÞìÙÑ ã­¤·Ÿ¢¹šÖv<{WÌÍVŦû¸¸¼Ì¡IRýÇDi6-fëUâ ûs=/óI~YRÀfã9t³Zìj™}ÍY6Z—9ËÑzz9ÉoXYLÆ9›f£å|Æ.–9`0$`$l\`ŠU±bCÄèq~É–˜›``“IÖ¾¼^Ï®²åz:ÉÖ%›_Ígù6Êßj‘òN ò²˜a¹÷CMëîPÙ½ÎØñ¸}žIwþûÅgZDhس)²JXþ†;Rò·åŽ” ý­hDÉ__QuaïX‘zÅwMÉâ.1v‚}Òî+5žÞ–šû»RóHÍïšzÿXɇI™Õé;„^ 2:Ñ-,Òµ>°Æ§Coûá5Ú)³î+ý×6µCÊñûÀj«‡jG‚²êSØ©õ‚ÈúRÞ –‚ˆ=aáD´/ë«aœýXSÜ{jX:7äýˆ•°a!z¢EöL{¶^°)eŠý G8Àú~Æ€ÍåÐðž°H™­ì']‘Ñ®§,dæm?cï2?.c¾wÜÔ·â&—‡ÄMÇívÜäªÜŒa› e±ýâ¦,¨á²'lê‡$®>Kþý–§ÚЮ‹¨‚«):5PØh~4ÒÐ_êÆ&M¡Ó“¸ŽE~JÒê=°8´CÁrÞŽÑtjµ ®CEJIç'á©mp†zZÕ•Rí³}¯Ã™K(Ô¦'ÑDï›qæ¢"ê>ÂhJ‰C‰XŠ‘U¡S°¨h#BQtlŽŽ|(J ¢ˆ*š¥¿U´û‹ÜŒo ¯Ž§Â¡)VbÏgTZ5¸ñ|gR¯*hº]¸ô‡vÖ-®E€²tt—j:êêEHé ŸÎ4BÙ•$å X) X ý§üŸ’:‡À4™‘©ìÎ;”‡›K:›¤“=8b¶ÆŽ\ôc…¡OgèЩUÀ¨ÉÉ`1ˆ›dcÊq:è ç…ÊæBñÈGã"É=‚›T¦;ã]Eèª ÁÕxR¢”bz]8Lá8»ùW¹¸ŠÞ¤·`lÔ"ì ŒÜ•v ªùÿ±«f›ÿ ˜ÓN«ŠX›?H¼«?ÿ(àÿUñ´"Ó:V]Z@hõmV{*i»îW™D³Êw¸û¯é§€F#P5$,U®FÞ”ÐiÓv@ˆ6¶tjñØtSðÞ#’6iM›úÔ©­S’&M¢:õy¥Úª…¢J“"<‘i'O¢pçIunÓ<·ó¤JJUÝÕõ¦¿C¹RÔˆÚä`¤ËFÆ551ö. ®©M‹ñ»º¤pµj”؈=|9Šð7\Ñ W•„£)A•˜·ÉHÛŒ·n7Zô5$øsµ=w2^Î[É6íGïki7)=U-å}oÌ[“ñ6ïc íÌxk_5ZV—O­x¤‰†rWmCY0ƦA«°eK×j«Õî]Úš‘ÐÀ¦­Ot2€ZM$i×(»y'ë¤9¼#ØXpB¶FÞó ¯Õ›ð5JÞ*ÛÛ’x{—m ѵë}³ºË0k%U›°M.dC.„œN4yÉÎâh›âm(wÁí*y•¶]çr!ÏóÝ/î¯rÄôV¿T6ì;4ÑpªP‚þ‘CR¡Hp;º<@ ÀŸ <úÄ!ºZ¢‡>TCë&H›ËT¡Žõ(šêp}À†oÂ. ‰i@”úð¤ÿ”Њxô1œ®ˆ@Ò…k ²ÑêVQØïP¹µÃô ÞA?·¢<]-¨Ý- m”'k©áÚquÝ×îÙ¼o\<¸9mÓ)ÄÖ´´›Mm#êp@Ï6ÈG¥ñŒÛµ0׳o-Ìá¼™úú.ÌÜmÖY•ºïsšá¾Á¼= k¬ Ñé¿udÑà8˜ËúÈbûø¢ Þ[ÇTÚõ»/ËAcuLŠƒø­ã‹˜öøøb×bóËrBWÕÝ®`NmðîFóÚ"\ø¼ù×.:J¨¶Ã96Wá³Õ²Í‡sRÑv8ï¶Â9I}/œ«HÛåp¾¯]{ûjglÄ•O[_RóÕhY,Êù²:È«¾£ž~8{ztòó²¼¹ÀëIvµJTÕÿ8œf>Ô6y(è&‡súp¿cEß(“tÀŽ²ÅÓ¼¸º Â_¿Vf“bôhv5É©ùªÌ§oéfË€×ðÊ qt-élð'vyù šô¤À'ÕîÁ»YøíôùÓgÇ…i±› 5<¤KO¡·àÙìaĪ˜óq‘—Cércáæåó'¯Þ½&nö2#ÈC²˜1¢3v3NG̨ŽR²Éâ:«XͧӌóIŻʨ}E÷ ò%£ïÜlºf³5›Oó«Œ-òe1³ÅuÁVÅ W“luÍÊk{St¥ã‘Γ“ÓïÎH:ËÝÂÁ¢ Ù¸p+Ã8ysíF§±`è:Ó6‹ÿ[LÖ«Š¿[*=„­óÓó÷<ú¹ä7eùðÕ$›•ÿq7‹ÈÀ"]$â”"Gú·©‹¹4ÚG6ðŬ>ô-¯±EûŽ±cök¸öœ½d¯Økö–Á(؈YÎ.Ù»fûÂ&lÊflÎlAŸ:èDU[V¶d+V²5ûÊþÃnØ·®¨Ä!¢zqöôí¯¿“¨ÊrOįŽÄ“ÒíÁH<ÆšŽx¸»K<éNétãÖ#ö¸–Ð {¤ôov I½`¿CZ°³ ±7Ù;v¹e«QQT7?.ØE¶dËlô%/ƒ¸êz%°|n2Ÿ5žÇÆóÉð9 t4&È.‹¯ÐÂ|½„"w¼f×ß×ù¬ÖKåPά˜åÐÐ÷h ~®Ÿ4þK¦ýgu¯e|1 J\å_yUÜ4> ¯^æ9+ÿ3ߨ—ýÅþÊ—ó®–ŠéOžýúôìM¥åïùÃÒ÷w8Dá]u¿'…O!ËU6Wª¯îÿÄf­ ²bõ…ìúÏ[KÈ+•Üe ›ËD=l%¿M²é÷Mæó£™­§`£¸šÝÏ~ò]qÛRS ã›Æ&PÜßÆZiŸ½x÷æøçò&¿ÙkXÂ:h¢ÝœŽ,K›}+­’n_þƒ vïZ4E¶6pQnêÓb¼i”‘ìWW›ÖE¹©U[oFo†„æE¬¿fPh`Ô2£lrQ\±Õ¦EJ«Å$û¶yQeW -èožœ?z÷o(aõmÏj'à™´} k¬Î˜È½i‰ÞhÁŠX\ï[ص؟ñ,ËùMå/c‚ËëÝ®‰S²öbvYÌŠò[u‰Ð¤,“î§JŽÎOÿxq„2Ú“Rž#eøJ× ¢DG‹EùÉN'¥íÊdÄÑeæ û·óã篞ýü¢˜^¬W¯²Ùéóüj½-Jm}FA0÷:bËrÑÑ657ù-—=Ò8-dÌÙ1âu•Êl™*YTùJ—éƒÖ¯“ß>¼?yT3}6Ÿ¾˜ûÓ‡tGdß9Ķ9lssâEÌw1ë;y½Þ»ÛJXĪ%l³€Ubé®*ÍrB+Hg‰IÿDƒˆhmèÆýÞ_ñß¼þòÝÛZ!î ?)9v\Þ‡#Ç8üt“mç:Š[Ù×­jDºwIª!1ÒB{Óåô °úÇÛÇ¿½¹Åé~Gö'\c"fÓê®ýéŽß¹Ã·O·Íï·;ÓæGµ7¢7ÕŽ’£6+šŽ)·ÈgÕc“Õ¶=ÙaÞŸëN}kû@Kß›ó´Ç«ZœM‹Šú{¸†9hÉyýôõù»Ó³ß5¸¡CúÙ^Ê·Æt}ø»NV"ß°nïBÔ5›®ÑÜ6˜¿š ó·t_)ý®Hö¹ÝÓ ã{1³c=ÝóƒHž$RÖ&›H¹Dd?»(…HkYÔ|V”'Ê¢DmEwJOHÅ_•ü‚HìM41²æ‰Ñ‡{yâyìЉ—ŸîüÂbÇÄçÎ!ÄŸÛ'ì/¤J& ô º;œ’¶“mp¡‚~•Z{>ÝpÜ„$º¨ßÚÚ§>?aÙüpeorÔqÔð«ÂpÍô¤X®Jò‰„\ê4«œÃ¥Þãòz~ñ`7?‚¼ïk6”u†·È‘·Èá¼CΆ‘>÷ö‰îöìöÖì"žÝnfçîðÙ·N`·'÷w²NËy˺>|ò»Çm›¸%†ŽDL‰ÿ&ñ½ôi›¾[’2ºC`øClö®¬g‹8Ãïü;ò'ù#hûÞ»MŸ¼“>y˜L}Ý“õ-j¿Ûèc7‡Ï½ÿø{›Ž[R¡mNdR1!íUûÿPÏ8 endstream endobj 864 0 obj << /Type /ObjStm /N 100 /First 871 /Length 2997 /Filter /FlateDecode >> stream xÚ¥Zmܶþ¾¿B} ìã;‡@À¹sí ¶“ÚõKQÅúN¶µo½½Âé¯ï3$%‘Z®.¾N$ É™gžÎPÚSÎu¢SZvR¹NKÓ)';¥|§ÉuVÙÎJ×Óã¶ó|µ¡#\ô]à«7Ú®,”H ‰%Hnœ¦Nj¾ ¶“7ÞaÈ↰Z²ZrâqéƒÃ †ˆ`Åay0j% 0„w¸q 7ªSRAæ( ¸š•"Öʘ†L6:à“­d˜lü5ì«ó+i0ÙLÖ˜Ls4&1'˜xf´0ì“` P‚.¥!‘ 7유Nk„“±+ ¸Úb<ÒLuøa¦·¢Iš”ì0¤ÉPÓé }ÏŒ0àŒ@±÷¦3pu·Œ‚xe´‘ˆ®!tðɘ€IøgA&<2ÞàŸñ±Ä<9Ü1ä5(éLÐbh6rŠP F®n,&\‚ïLg5¬“ ¸á³iËã¶!`@jÉó P G+B*8À,uÆzܸÎ9HÞÃ#ÅË é@<>K8B`ÀƒéŽY÷Æ HuÞÂ7"Ýyo劰ÀÁs`OXžƒÿ”à90£ÁY²’rî ‹(°”“ 1BòB $&Ý!µ †ƒSX0Ò<„lÞxBa1â$gÅ€… F`ò„…cP;_ƒ@n ïÙœæ¬paõý÷«Ó¿ÿñ¥ïNÿº½Ú¯N_Þ¼ÛÇGÊÕéëëžGºÓç/ž¼~ôËw{ùuy,;ï¯/v›/ûíùb±[_@¾Ù]ïÏ>®wH´ÕéÓu~À^¾Ù\î?^ƒçþðÃ7XüÓ£'/^%ë÷_~Z_íÿ%±Eqø9uÝÆÙó7¯Î¿Ûí¿6l‡*ãØ'“qýíÆ_=~ûðÍÏ0~ýÇ¡q+çÆEi[èɶ–ßnûìíÓ¿=?¶/Æõܸ)£ÏƒqÿºþÐ_¯N϶70…ýºÞõ¸ãêÕþ¼¹¼îþ))ZA‘Téêò³éJ6]C|þí.6´ŒÜ p*_}ºêüŒ¯ ÃÝlX—t8JW¯ó5?‡ŒáÿðÃȸǸ8§kÆot²alwæî6|Öã‚ž®!Å<]¥¸³ ›ã€rŸ¯)Ö6ísýäM\þY®²AÉ’ 'Ò³“ù9s‰æqgNg·sÉ/çó3ågº»Ü¿£Žtz™âà³moÒ¸O¥âN6¼Kœx®)>¤àCJºÞš»ò˜ ’·æ¤.úr1ãæÉÎóé &¥ÌÉéöö`*{ÌŸ!#Óy7¹œñ>í¦x~¼a9ºë2e®…è—›ý§Í¯‹µµK­3–ÖN¦ *i\¹Ùê1š;ÂCHS«›Üžwýâ²,Éiµ÷ÓjS­–õj¯Z«õ´Z´W?ï¿îã™ùpµ ãê!5†ÕÊ”¶)úÎ%³“Ugv;Q‹FD ØQ ðDµ¥%]ÖUºZÄÛ‰xku©J—héšÂ`Å’.SŶ‚b¦ ˜Å Séj¥‡™¸7‹Ü« —iqo&îõœû*QuË+=±­ÍâêŠmÓb[OlkÑÖ•V7‘¨‰_u„ßl»Nz]%½©JÂ}=72¯ÔbM¨’~(4ÙäàM29,Ì&çeHMñ‘Ke(T[£Y“ä,YkXœÀËVÎÉ):R,,U¨d+Tb •pKªª¸5Q‰) B-©ªâ!ZAŒDjjÊk̆‘Ø`BT—nU'Å€0%…¬‹¯ÙcêPTæ¨á%Ô“;¾Ò—‡ï4ÒN5ë¾d»|ꪳ¦ZÙ¤ÉSG5ÔzYg2™b*ÎŽt·©ÿú6ÁQ}«ùN½wÖzëT¨Êá` fÕ…3ùL336†aÖ‚‡Þ–Ò®†©ãÎnµ²j·­n;5ÛY¯­šk ÀÔZgU—LÀNdÆ™¥J=uÛY³­Õ—¦Õi§F;ë³µš²¸´šìÔcg-¶R£K¶Mƒí©ÙÎzm•M­Þ:µÖYg­³²LnWQ¯Kê­Z*NS#žõáÒR«?MÍuÞ[Ë…%תÁuÑ/é¨YRÝ>¿O'¬ãZJÚUóH3u§––|ôl4”q_/D«LÞ!†F-EKÑšúgÛRª£%ÂçëÏ=Þ»îo/î¿Ü¯wûÄ+½¸Ý{ò~»Ý_m÷ýyÒñ'æ™TAš?R}‚òq0לàœæR ©Î¯‡§O7Ÿ7û–jîosÐ?íûÏÛ`/ –1…@³ÀÃW, |!p, 9°ÁÞ0§ ÈóÚô¢}oýåKu¹ùúàá ÿd’„¸óðœ/¥Ÿ>n…+ùSU–^]öÐ.hHÒwÛ›óþêæu¿ûÌš¼RMÀÀÓ\p€>N:[ï/žmw–™!byäfÑÍŸ7²ôÓæ}à˜çCIúh·ù7EïÂL8<ùÃH’"Ë÷øW§þDÕŠ¶<ø¬¿ê…a•®0ôxýnÇ8]Òøx»½ôA•$Ò<ˆE !ç± 9‘oÈa7ÈWô<*¼dÉËÐÂ!k6Æ^èƇQøÆ€á,’ˇnÆCºT~ÁŸo¾fÉŸxH–F7Œ1îFÐ|Ä݈š¸m 7o.¼ÃÜR®¥CülͲCÄ: ÜU¾1?%Äz+Z86{rkФŠàñA¨¥e'餽®éiݘZ¶Ló(GWÛc£bÚ£5{yl”5{Ûòê_sõèÚð§ å#eË”~ˆ{™®féƒWóÝ:[È•ƒÒOñ/5Ó•ò5#8|å¨_ƒŽ÷Y>øQ>7P>jR>SQ>«¹[NèåAeбRÆšÿV†òéŒòÉ„ŽfãaãX™Ë&†?»É︔;$÷§ ÎmñÍ/²¹ˆ†ü—]áð˜TÆuÑÂyAŒ?Zü¯|s( endstream endobj 906 0 obj << /Author()/Title()/Subject()/Creator(LaTeX with hyperref package)/Producer(pdfTeX-1.40.14)/Keywords() /CreationDate (D:20150422102823+01'00') /ModDate (D:20150422102823+01'00') /Trapped /False /PTEX.Fullbanner (This is pdfTeX, Version 3.1415926-2.5-1.40.14 (TeX Live 2013/Debian) kpathsea version 6.1.1) >> endobj 905 0 obj << /Type /ObjStm /N 1 /First 6 /Length 100 /Filter /FlateDecode >> stream xÚ³40Q0ಱáÒ©,HUÐwN,IÌÉOçÒHLO-V°03W0PâÒ÷/-ÉÉÌ‹X@Düs\Kc¤Þ7?%U?´8¦¨­ 5Ï1¹$3?OÁЬÒÎŽ là ä endstream endobj 907 0 obj << /Type /XRef /Index [0 908] /Size 908 /W [1 3 1] /Root 904 0 R /Info 906 0 R /ID [<30CDADFC92696608941989F1C2F8B507> <30CDADFC92696608941989F1C2F8B507>] /Length 2170 /Filter /FlateDecode >> stream xÚ%˜IlIÇû{ݯŸ—ØN:v¼o‰“x‰8Žã8ñóšÅ[œØ‰—ıÓNfŒ¸€8ÌœØFá€4%4Q a„@ )¸Œ@!Ä!FŒŠ X$.Àûý}ù¹¾U÷ëªúµ8Š¢è…(*D¥®ü‰Âh!ŠóY4´‹h3„1HÐÆЦ ‹ E»„V&,*´i´)ÂjPƒvía-8‚ví*a¨GÛG›$lGÑ£]!<²Ê§—c´ Âã -A»LØN Õ 6ƒ´Z´K„­  íÚa;è@;†v‘°t¡eh£„Ý íÚÂ^p­í<á)Їւ6BxœAkE&< úÑÐ΀A´´!BáÚ´AB½`í6šÓ_@[Cë'Ô‡_D{Š¦ŸT§+£a>B;C¨»ŒV‡¦ÏÕ`_AëFë#ÔD]E;ƒ¦®j’§ÐÆÑNÊ Óh—Ñ4L2×,ÚZá˜G“‡4Ä à:ÚM´.Âà&Ú=4MÏ-°X¨ÌÉZáXF»Š¦©]«h7ÐÚoƒÊ0Üq4Ù⸋¦\h!\•0+ ÉR÷À}´"Ú ÂM°…–¢ÉŽÛ`MÞh$|‚]ðì}ðäà<OIe-¹”«dhP†’—AyI6e#9”ƒd^Pæ‘oAùF–e¹2@F¾/ðáQ g`œ˜™À”æ2´Z5è«è~ ƒYÒUr5£ È”âFI&,Z´ú=‹U ˜+຀> 8àì€åiH’€ùi°|À§Õ"àâÐfQÝôú üä°Hì€KPÀ¶E+`ÌpÙ¢öŸë±2Ijd³fú>Ìæ ×Æ 7Á2ÀkaÌZÔ7¦÷áÓ° p]ÀRaÉ¢ª%W3õ _…û7…-€‡ÂÀ9çL4ºNé-x(`© Yèاä>kÙ@ È4O,šùœ*älïf#ww¨ûç©­  ¤m÷è±jš<¦tÔ[ô‘ª8 ´žj¡m´èµwUÑ:VÖn õT ^»EoœV»^ EõÐRzh= ´ljMÔGYô쥞Zð.XôÖŒ4mwêù%‹¾ù½Èlö½È ¥AÕŽ–¾ò ˜c½ý¾j'‹[¹ ®ZôõeU\¬få0m…ß|Ls€%­|°h•oE°–Á xÖÁu+üaA/Xwûe6`…¿ömlXá/o*ÜÀ=pl‚-° vÀ ‰ÊÁ«¨Ì°oqãC½Š$É«øµ‹¿±…æP°¸ÿg cFÒc _mqù’jk@-¨E‹GjUÁ6›Lžg;ö O~Eµìؾ` °Û{,åÙã}“ų¿WcÜäÛæòÉcßjñÊ+j‚¯<Îñ8Ç÷Yüê*p‰goôØÇ÷Z¼ù\øÊk38ÇkÏë·øàËj¢SP5¥€YýÀÃùJμ¶¯ÆØÂOZüÉj«ÄW~Êâ7¾+ #y\âµña¯í—øY‹?ýkµc·Ê˜d_ü¢ÅŸ?ü <äq‰¿ Ö~ñwnòøÀ¯Xü…ê‰  -Kƒƒ7<ÞðxÃcB¯ÂM~lYü|^ϲvæ:ii’ñ—‡XZüc‹¿:­vOØa @ PÕ ²ø‰ˆR J Ê’³¯j¹¹@xÔXüý©]-¨ à(82p4‚ Îâ7ë±&ÐZAh tƒf‹ò®žè½à$8úÀipœý`œ•tùE•^0ÈqÐÌ8¥e“h#à<Ãÿê…kgÒ¯K`\à ˜-þ]“ÓK¯ñ¬Þ7 fÀ,˜e‹?x]çÀM°`ñËgÒ®™–€J+@Ç£Û`Í’Â×Ô˜£PÏ-‹ÿùG…è³G@ñä`¬[üß¿©ÝØf44—à¸6-iyKí,™ k NáKŽŽŒô<»àØû–t¼¯Çj-­×Oº*€Ã?éÌ’Ñ—ª¨ØÑaG‡vtøÏ•,™R;æp˜Ãa‡9|åê,¹ù5áá× š,ÙSÖsØÌa3‡Ã\‹%[‰jqÃk®Ó’½ÏJëbpê(á5×cÉÓ»ªÀ×9\çpc°®s¸Î ‚!€ÿÜ0Àf›9RÈá!‡¥Ü)K^ù©ÞŒ_ Y†O3•:Gûìœ%ŸiU;¬ç®XñüjdÅ_v×¹«àšö4¬çÆ-ùĤB|êÊù;`Ü3–|êßjw`V‡;Žu c:½­8Ò¥Æ:žãS‡O>uøÔáS‡ÿܺß[Uc â´Ti.qÃu×¹M€1Ý6À±îÀa‡¹€õ‡ÔL+ ”Ë-Mꫳ-Ëa6l¥Û"KO½T-ϬÊÒÖ…:5‚ŠO{Û¥ÉC\82m‹\µ²zÐŽ‚c–ž¯Qc›,ü¢Bï¬ÕÒùaim pÛÉÆ-]{]ºté¥Z-¦ÇÚìક̀Y0æÁ¸´÷pýÊnE°déúá§-[º}N¥K÷>PiÕÒvªtÛÒØš¥ÏºTºcé;_R鮥¿ý¶Jë–þ©_¥ Kÿú\¥{–þãðÍ÷­”n©´i¥…G*mYiwW¥m+}ëm•v¬ônÎøåºÂëή‹}Ëé[Nßrú–먫%}Ëé[¾–Á XºYë*­ÿeèŸë€û`®{([j¾ ¸ûåZõ¸ñåŒs®»Ð.`ÙÌåNÝ…ôÿ êùÐé ×=‹¬*}³‚â\¥èÿˆ+Ÿ endstream endobj startxref 526168 %%EOF gnm/inst/doc/gnmOverview.R0000644000176000001440000006073712515664670015271 0ustar ripleyusers### R code from vignette source 'gnmOverview.Rnw' ################################################### ### code chunk number 1: gnmOverview.Rnw:65-66 ################################################### getOption("SweaveHooks")[["eval"]]() options(SweaveHooks = list(eval = function() options(show.signif.stars = FALSE))) ################################################### ### code chunk number 2: Load_gnm ################################################### getOption("SweaveHooks")[["eval"]]() library(gnm) ################################################### ### code chunk number 3: migrationData ################################################### getOption("SweaveHooks")[["eval"]]() count <- c(11607, 100, 366, 124, 87, 13677, 515, 302, 172, 225, 17819, 270, 63, 176, 286, 10192 ) region <- c("NE", "MW", "S", "W") row <- gl(4, 4, labels = region) col <- gl(4, 1, length = 16, labels = region) ################################################### ### code chunk number 4: squareTableModels ################################################### getOption("SweaveHooks")[["eval"]]() independence <- glm(count ~ row + col, family = poisson) quasi.indep <- glm(count ~ row + col + Diag(row, col), family = poisson) symmetry <- glm(count ~ Symm(row, col), family = poisson) quasi.symm <- glm(count ~ row + col + Symm(row, col), family = poisson) comparison1 <- anova(independence, quasi.indep, quasi.symm) print(comparison1, digits = 7) comparison2 <- anova(symmetry, quasi.symm) print(comparison2) ################################################### ### code chunk number 5: EriksonData ################################################### getOption("SweaveHooks")[["eval"]]() ### Collapse to 7 by 7 table as in Erikson et al. (1982) erikson <- as.data.frame(erikson) lvl <- levels(erikson$origin) levels(erikson$origin) <- levels(erikson$destination) <- c(rep(paste(lvl[1:2], collapse = " + "), 2), lvl[3], rep(paste(lvl[4:5], collapse = " + "), 2), lvl[6:9]) erikson <- xtabs(Freq ~ origin + destination + country, data = erikson) ################################################### ### code chunk number 6: wedderburn ################################################### getOption("SweaveHooks")[["eval"]]() ## data from Wedderburn (1974), see ?barley logitModel <- glm(y ~ site + variety, family = wedderburn, data = barley) fit <- fitted(logitModel) print(sum((barley$y - fit)^2 / (fit * (1-fit))^2)) ################################################### ### code chunk number 7: termPredictors ################################################### getOption("SweaveHooks")[["eval"]]() print(temp <- termPredictors(quasi.symm)) rowSums(temp) - quasi.symm$linear.predictors ################################################### ### code chunk number 8: RC_homogeneous_model_1 ################################################### getOption("SweaveHooks")[["eval"]]() set.seed(1) RChomog1 <- gnm(Freq ~ origin + destination + Diag(origin, destination) + MultHomog(origin, destination), family = poisson, data = occupationalStatus, verbose = FALSE) ################################################### ### code chunk number 9: RC_homogeneous_model_2 ################################################### getOption("SweaveHooks")[["eval"]]() set.seed(2) RChomog2 <- update(RChomog1) ################################################### ### code chunk number 10: Compare_coefficients ################################################### getOption("SweaveHooks")[["eval"]]() compareCoef <- cbind(coef(RChomog1), coef(RChomog2)) colnames(compareCoef) <- c("RChomog1", "RChomog2") round(compareCoef, 4) ################################################### ### code chunk number 11: Summarize_model ################################################### getOption("SweaveHooks")[["eval"]]() summary(RChomog2) ################################################### ### code chunk number 12: RC_homogeneous_constrained_model1 ################################################### getOption("SweaveHooks")[["eval"]]() set.seed(1) RChomogConstrained1 <- update(RChomog1, constrain = length(coef(RChomog1))) ################################################### ### code chunk number 13: RC_homogeneous_constrained_model2 ################################################### getOption("SweaveHooks")[["eval"]]() set.seed(2) RChomogConstrained2 <- update(RChomogConstrained1) identical(coef(RChomogConstrained1), coef(RChomogConstrained2)) ################################################### ### code chunk number 14: Eliminate_Eg ################################################### getOption("SweaveHooks")[["eval"]]() set.seed(1) n <- 1000 x <- rep(rnorm(n), rep(3, n)) counts <- as.vector(rmultinom(n, 10, c(0.7, 0.1, 0.2))) rowID <- gl(n, 3, 3 * n) resp <- gl(3, 1, 3 * n) ################################################### ### code chunk number 15: Double_UNIDIFF_model ################################################### getOption("SweaveHooks")[["eval"]]() doubleUnidiff <- gnm(Freq ~ election:vote + election:class:religion + Mult(Exp(election), religion:vote) + Mult(Exp(election), class:vote), family = poisson, data = cautres) ################################################### ### code chunk number 16: Contrast_matrix ################################################### getOption("SweaveHooks")[["eval"]]() coefs <- names(coef(doubleUnidiff)) contrCoefs <- coefs[grep(", religion:vote", coefs)] nContr <- length(contrCoefs) contrMatrix <- matrix(0, length(coefs), nContr, dimnames = list(coefs, contrCoefs)) contr <- contr.sum(contrCoefs) # switch round to contrast with first level contr <- rbind(contr[nContr, ], contr[-nContr, ]) contrMatrix[contrCoefs, 2:nContr] <- contr contrMatrix[contrCoefs, 2:nContr] ################################################### ### code chunk number 17: Check_estimability_1 ################################################### getOption("SweaveHooks")[["eval"]]() checkEstimable(doubleUnidiff, contrMatrix) ################################################### ### code chunk number 18: Check_estimability_2 ################################################### getOption("SweaveHooks")[["eval"]]() coefs <- names(coef(doubleUnidiff)) contrCoefs <- coefs[grep("[.]religion", coefs)] nContr <- length(contrCoefs) contrMatrix <- matrix(0, length(coefs), length(contrCoefs), dimnames = list(coefs, contrCoefs)) contr <- contr.sum(contrCoefs) contrMatrix[contrCoefs, 2:nContr] <- rbind(contr[nContr, ], contr[-nContr, ]) checkEstimable(doubleUnidiff, contrMatrix) ################################################### ### code chunk number 19: Get_contrasts_1 ################################################### getOption("SweaveHooks")[["eval"]]() myContrasts <- getContrasts(doubleUnidiff, pickCoef(doubleUnidiff, ", religion:vote")) myContrasts ################################################### ### code chunk number 20: qvplot ################################################### getOption("SweaveHooks")[["eval"]]() plot(myContrasts, main = "Relative strength of religion-vote association, log scale", xlab = "Election", levelNames = 1:4) ################################################### ### code chunk number 21: RCmodel ################################################### getOption("SweaveHooks")[["eval"]]() mentalHealth$MHS <- C(mentalHealth$MHS, treatment) mentalHealth$SES <- C(mentalHealth$SES, treatment) RC1model <- gnm(count ~ SES + MHS + Mult(SES, MHS), family = poisson, data = mentalHealth) ################################################### ### code chunk number 22: RCmodel_constrained ################################################### getOption("SweaveHooks")[["eval"]]() RC1model2 <- gnm(count ~ SES + MHS + Mult(1, SES, MHS), constrain = "[.]SES[AF]", constrainTo = c(0, 1), ofInterest = "[.]SES", family = poisson, data = mentalHealth) summary(RC1model2) ################################################### ### code chunk number 23: getContrasts_simple ################################################### getOption("SweaveHooks")[["eval"]]() getContrasts(RC1model, pickCoef(RC1model, "[.]SES"), ref = "first", scaleRef = "first", scaleWeights = c(rep(0, 5), 1)) ################################################### ### code chunk number 24: two-way ################################################### getOption("SweaveHooks")[["eval"]]() xtabs(y ~ site + variety, barley) ################################################### ### code chunk number 25: residSVD ################################################### getOption("SweaveHooks")[["eval"]]() emptyModel <- gnm(y ~ -1, family = wedderburn, data = barley) biplotStart <- residSVD(emptyModel, barley$site, barley$variety, d = 2) biplotModel <- gnm(y ~ -1 + instances(Mult(site, variety), 2), family = wedderburn, data = barley, start = biplotStart) ################################################### ### code chunk number 26: residSVDplot ################################################### getOption("SweaveHooks")[["eval"]]() plot(coef(biplotModel), biplotStart, main = "Comparison of residSVD and MLE for a 2-dimensional biplot model", ylim = c(-2, 2), xlim = c(-4, 4)) abline(a = 0, b = 1, lty = 2) ################################################### ### code chunk number 27: Set_contrasts_attribute ################################################### getOption("SweaveHooks")[["eval"]]() set.seed(1) mentalHealth$MHS <- C(mentalHealth$MHS, treatment) mentalHealth$SES <- C(mentalHealth$SES, treatment) ################################################### ### code chunk number 28: RC1_model ################################################### getOption("SweaveHooks")[["eval"]]() RC1model <- gnm(count ~ SES + MHS + Mult(SES, MHS), family = poisson, data = mentalHealth) RC1model ################################################### ### code chunk number 29: Normalize_scores ################################################### getOption("SweaveHooks")[["eval"]]() rowProbs <- with(mentalHealth, tapply(count, SES, sum) / sum(count)) colProbs <- with(mentalHealth, tapply(count, MHS, sum) / sum(count)) rowScores <- coef(RC1model)[10:15] colScores <- coef(RC1model)[16:19] rowScores <- rowScores - sum(rowScores * rowProbs) colScores <- colScores - sum(colScores * colProbs) beta1 <- sqrt(sum(rowScores^2 * rowProbs)) beta2 <- sqrt(sum(colScores^2 * colProbs)) assoc <- list(beta = beta1 * beta2, mu = rowScores / beta1, nu = colScores / beta2) assoc ################################################### ### code chunk number 30: Elliptical_contrasts ################################################### getOption("SweaveHooks")[["eval"]]() mu <- getContrasts(RC1model, pickCoef(RC1model, "[.]SES"), ref = rowProbs, scaleWeights = rowProbs) nu <- getContrasts(RC1model, pickCoef(RC1model, "[.]MHS"), ref = colProbs, scaleWeights = colProbs) mu nu ################################################### ### code chunk number 31: RC2_model ################################################### getOption("SweaveHooks")[["eval"]]() RC2model <- gnm(count ~ SES + MHS + instances(Mult(SES, MHS), 2), family = poisson, data = mentalHealth) RC2model ################################################### ### code chunk number 32: Homogeneous_effects ################################################### getOption("SweaveHooks")[["eval"]]() RChomog <- gnm(Freq ~ origin + destination + Diag(origin, destination) + MultHomog(origin, destination), family = poisson, data = occupationalStatus) RChomog ################################################### ### code chunk number 33: Heterogeneous_effects ################################################### getOption("SweaveHooks")[["eval"]]() RCheterog <- gnm(Freq ~ origin + destination + Diag(origin, destination) + Mult(origin, destination), family = poisson, data = occupationalStatus) anova(RChomog, RCheterog) ################################################### ### code chunk number 34: Transform_to_counts ################################################### getOption("SweaveHooks")[["eval"]]() set.seed(1) count <- with(voting, percentage/100 * total) yvar <- cbind(count, voting$total - count) ################################################### ### code chunk number 35: Class_mobility ################################################### getOption("SweaveHooks")[["eval"]]() classMobility <- gnm(yvar ~ Dref(origin, destination), family = binomial, data = voting) classMobility ################################################### ### code chunk number 36: Class_mobility_weights ################################################### getOption("SweaveHooks")[["eval"]]() DrefWeights(classMobility) ################################################### ### code chunk number 37: Salariat_factors ################################################### getOption("SweaveHooks")[["eval"]]() upward <- with(voting, origin != 1 & destination == 1) downward <- with(voting, origin == 1 & destination != 1) ################################################### ### code chunk number 38: Social_mobility ################################################### getOption("SweaveHooks")[["eval"]]() socialMobility <- gnm(yvar ~ Dref(origin, destination, delta = ~ 1 + downward + upward), family = binomial, data = voting) socialMobility ################################################### ### code chunk number 39: social_mobility_weights ################################################### getOption("SweaveHooks")[["eval"]]() DrefWeights(socialMobility) ################################################### ### code chunk number 40: Downward_mobility ################################################### getOption("SweaveHooks")[["eval"]]() downwardMobility <- gnm(yvar ~ Dref(origin, destination, delta = ~ 1 + downward), family = binomial, data = voting) downwardMobility DrefWeights(downwardMobility) ################################################### ### code chunk number 41: UNIDIFF_model ################################################### getOption("SweaveHooks")[["eval"]]() set.seed(1) unidiff <- gnm(Freq ~ educ*orig + educ*dest + Mult(Exp(educ), orig:dest), ofInterest = "[.]educ", family = poisson, data = yaish, subset = (dest != 7)) coef(unidiff) ################################################### ### code chunk number 42: Unidiff_contrasts ################################################### getOption("SweaveHooks")[["eval"]]() getContrasts(unidiff, ofInterest(unidiff)) ################################################### ### code chunk number 43: double_UNIDIFF_model ################################################### getOption("SweaveHooks")[["eval"]]() set.seed(1) doubleUnidiff <- gnm(Freq ~ election*vote + election*class*religion + Mult(Exp(election), religion:vote) + Mult(Exp(election), class:vote), family = poisson, data = cautres) getContrasts(doubleUnidiff, rev(pickCoef(doubleUnidiff, ", class:vote"))) getContrasts(doubleUnidiff, rev(pickCoef(doubleUnidiff, ", religion:vote"))) ################################################### ### code chunk number 44: Scale_yields ################################################### getOption("SweaveHooks")[["eval"]]() set.seed(1) yield.scaled <- wheat$yield * sqrt(3/1000) treatment <- interaction(wheat$tillage, wheat$summerCrop, wheat$manure, wheat$N, sep = "") ################################################### ### code chunk number 45: AMMI_model ################################################### getOption("SweaveHooks")[["eval"]]() mainEffects <- gnm(yield.scaled ~ year + treatment, family = gaussian, data = wheat) svdStart <- residSVD(mainEffects, year, treatment, 3) bilinear1 <- update(mainEffects, . ~ . + Mult(year, treatment), start = c(coef(mainEffects), svdStart[,1])) ################################################### ### code chunk number 46: AOD ################################################### getOption("SweaveHooks")[["eval"]]() anova(mainEffects, bilinear1, test = "F") ################################################### ### code chunk number 47: AMMI_model2 ################################################### getOption("SweaveHooks")[["eval"]]() set.seed(1) barleyModel <- gnm(height ~ year + genotype + Mult(year, genotype), data = barleyHeights) ################################################### ### code chunk number 48: Spherical_contrasts ################################################### getOption("SweaveHooks")[["eval"]]() gamma <- getContrasts(barleyModel, pickCoef(barleyModel, "[.]y"), ref = "mean", scaleWeights = "unit") delta <- getContrasts(barleyModel, pickCoef(barleyModel, "[.]g"), ref = "mean", scaleWeights = "unit") gamma delta ################################################### ### code chunk number 49: CI ################################################### getOption("SweaveHooks")[["eval"]]() gamma[[2]][,1] + (gamma[[2]][,2]) %o% c(-1.96, 1.96) delta[[2]][,1] + (delta[[2]][,2]) %o% c(-1.96, 1.96) ################################################### ### code chunk number 50: SVD ################################################### getOption("SweaveHooks")[["eval"]]() svd(termPredictors(barleyModel)[, "Mult(year, genotype)"])$d ################################################### ### code chunk number 51: Biplot_model ################################################### getOption("SweaveHooks")[["eval"]]() set.seed(83) biplotModel <- gnm(y ~ -1 + instances(Mult(site, variety), 2), family = wedderburn, data = barley) ################################################### ### code chunk number 52: Row_and_column_scores ################################################### getOption("SweaveHooks")[["eval"]]() barleyMatrix <- xtabs(biplotModel$predictors ~ site + variety, data = barley) barleySVD <- svd(barleyMatrix) A <- sweep(barleySVD$u, 2, sqrt(barleySVD$d), "*")[, 1:2] B <- sweep(barleySVD$v, 2, sqrt(barleySVD$d), "*")[, 1:2] rownames(A) <- levels(barley$site) rownames(B) <- levels(barley$variety) colnames(A) <- colnames(B) <- paste("Component", 1:2) A B ################################################### ### code chunk number 53: Biplot1 ################################################### getOption("SweaveHooks")[["eval"]]() barleyCol <- c("red", "blue") plot(rbind(A, B), pch = c(levels(barley$site), levels(barley$variety)), col = rep(barleyCol, c(nlevels(barley$site), nlevels(barley$variety))), xlim = c(-4, 4), ylim = c(-4, 4), main = "Biplot for barley data", xlab = "Component 1", ylab = "Component 2") text(c(-3.5, -3.5), c(3.9, 3.6), c("sites: A-I","varieties: 1-9, X"), col = barleyCol, adj = 0) ################################################### ### code chunk number 54: Biplot2 ################################################### getOption("SweaveHooks")[["eval"]]() plot(rbind(A, B), pch = c(levels(barley$site), levels(barley$variety)), col = rep(barleyCol, c(nlevels(barley$site), nlevels(barley$variety))), xlim = c(-4, 4), ylim = c(-4, 4), main = "Biplot for barley data", xlab = "Component 1", ylab = "Component 2") text(c(-3.5, -3.5), c(3.9, 3.6), c("sites: A-I","varieties: 1-9, X"), col = barleyCol, adj = 0) abline(a = 0, b = tan(pi/3)) abline(a = 0, b = -tan(pi/6)) abline(a = 2.6, b = tan(pi/3), lty = 2) abline(a = 4.5, b = tan(pi/3), lty = 2) abline(a = 1.3, b = -tan(pi/6), lty = 2) text(2.8, 3.9, "v-axis", font = 3) text(3.8, -2.7, "h-axis", font = 3) ################################################### ### code chunk number 55: Double_additive ################################################### getOption("SweaveHooks")[["eval"]]() variety.binary <- factor(match(barley$variety, c(2,3,6), nomatch = 0) > 0, labels = c("rest", "2,3,6")) doubleAdditive <- gnm(y ~ variety + Mult(site, variety.binary), family = wedderburn, data = barley) ################################################### ### code chunk number 56: Compare_chi-squared ################################################### getOption("SweaveHooks")[["eval"]]() biplotModChiSq <- sum(residuals(biplotModel, type = "pearson")^2) doubleAddChiSq <- sum(residuals(doubleAdditive, type = "pearson")^2) c(doubleAddChiSq - biplotModChiSq, doubleAdditive$df.residual - biplotModel$df.residual) ################################################### ### code chunk number 57: Re-express_data ################################################### getOption("SweaveHooks")[["eval"]]() set.seed(1) subset(backPain, x1 == 1 & x2 == 1 & x3 == 1) backPainLong <- expandCategorical(backPain, "pain") head(backPainLong) ################################################### ### code chunk number 58: Stereotype_model ################################################### getOption("SweaveHooks")[["eval"]]() oneDimensional <- gnm(count ~ pain + Mult(pain, x1 + x2 + x3), eliminate = id, family = "poisson", data = backPainLong) oneDimensional ################################################### ### code chunk number 59: Qualitative_model ################################################### getOption("SweaveHooks")[["eval"]]() threeDimensional <- gnm(count ~ pain + pain:(x1 + x2 + x3), eliminate = id, family = "poisson", data = backPainLong) threeDimensional ################################################### ### code chunk number 60: Calculate_log-likelihood ################################################### getOption("SweaveHooks")[["eval"]]() logLikMultinom <- function(model, size){ object <- get(model) l <- sum(object$y * log(object$fitted/size)) c(nParameters = object$rank - nlevels(object$eliminate), logLikelihood = l) } size <- tapply(backPainLong$count, backPainLong$id, sum)[backPainLong$id] t(sapply(c("oneDimensional", "threeDimensional"), logLikMultinom, size)) ################################################### ### code chunk number 61: Constrain_slopes ################################################### getOption("SweaveHooks")[["eval"]]() ## before constraint summary(oneDimensional) oneDimensional <- gnm(count ~ pain + Mult(pain, offset(x1) + x2 + x3), eliminate = id, family = "poisson", data = backPainLong) ## after constraint summary(oneDimensional) ################################################### ### code chunk number 62: Get_slopes ################################################### getOption("SweaveHooks")[["eval"]]() getContrasts(oneDimensional, pickCoef(oneDimensional, "[.]pain")) ################################################### ### code chunk number 63: singleExp ################################################### getOption("SweaveHooks")[["eval"]]() x <- 1:100 y <- exp(- x / 10) set.seed(1) saved.fits <- list() for (i in 1:100) saved.fits[[i]] <- gnm(y ~ Exp(1 + x), verbose = FALSE) table(zapsmall(sapply(saved.fits, deviance))) ################################################### ### code chunk number 64: singleExp2 ################################################### getOption("SweaveHooks")[["eval"]]() saved.fits[[2]] ################################################### ### code chunk number 65: doubleExp ################################################### getOption("SweaveHooks")[["eval"]]() x <- 1:100 y <- exp(- x / 10) + 2 * exp(- x / 50) set.seed(1) saved.fits <- list() for (i in 1:100) { saved.fits[[i]] <- suppressWarnings(gnm(y ~ Exp(1 + x, inst = 1) + Exp(1 + x, inst = 2), verbose = FALSE)) } table(round(unlist(sapply(saved.fits, deviance)), 4)) ################################################### ### code chunk number 66: doubleExp2 ################################################### getOption("SweaveHooks")[["eval"]]() singleExp <- gnm(y ~ Exp(1 + x), start = c(NA, NA, -0.1), verbose = FALSE) singleExp meanOnly <- gnm(y ~ 1, verbose = FALSE) meanOnly plot(x, y, main = "Two sub-optimal fits to a sum-of-exponentials curve") lines(x, fitted(singleExp)) lines(x, fitted(meanOnly), lty = "dashed") ################################################### ### code chunk number 67: doubleExp3 ################################################### getOption("SweaveHooks")[["eval"]]() gnm(y ~ instances(Exp(1 + x), 2), start = c(NA, NA, -0.1, NA, -0.1), verbose = FALSE) gnm/tests/0000755000176000001440000000000012515147523012232 5ustar ripleyusersgnm/tests/nls.Rout.save0000644000176000001440000000350512515147523014641 0ustar ripleyusers R Under development (unstable) (2013-09-11 r63910) -- "Unsuffered Consequences" Copyright (C) 2013 The R Foundation for Statistical Computing Platform: x86_64-unknown-linux-gnu (64-bit) R is free software and comes with ABSOLUTELY NO WARRANTY. You are welcome to redistribute it under certain conditions. Type 'license()' or 'licence()' for distribution details. R is a collaborative project with many contributors. Type 'contributors()' for more information and 'citation()' on how to cite R or R packages in publications. Type 'demo()' for some demos, 'help()' for on-line help, or 'help.start()' for an HTML browser interface to help. Type 'q()' to quit R. > library(gnm) > set.seed(1) > DNase1 <- subset(DNase, Run == 1) > > fm3DNase1.2 <- gnm(density ~ -1 + + Mult(1, Inv(Const(1) + Exp(1 + Mult(offset(-log(conc)), + Inv(1))))), + start = c(NA, 0, 1), data = DNase1, trace = TRUE) Initialising Initial Deviance = 15.301005 Running start-up iterations Start-up iteration 1. Deviance = 0.7139315 Start-up iteration 2. Deviance = 0.7139315 Running main iterations Iteration 1. Deviance = 0.350113 Iteration 2. Deviance = 0.07852421 Iteration 3. Deviance = 0.005771315 Iteration 4. Deviance = 0.00479172 Iteration 5. Deviance = 0.004789569 Done > coef(fm3DNase1.2) Coefficients: Mult(., Inv(Exp(1 + Mult(offset(-log(conc)), Inv(1))) + Const(1))). 2.345176 Mult(1, Inv(Exp(. + Mult(offset(-log(conc)), Inv(1))) + Const(1))). 1.424054 Mult(1, Inv(Exp(1 + Mult(offset(-log(conc)), Inv(.))) + Const(1))). 1.041453 > > proc.time() user system elapsed 1.848 0.048 1.889 gnm/tests/logistic.Rout.save0000644000176000001440000000534712515147523015670 0ustar ripleyusers R Under development (unstable) (2013-09-11 r63910) -- "Unsuffered Consequences" Copyright (C) 2013 The R Foundation for Statistical Computing Platform: x86_64-unknown-linux-gnu (64-bit) R is free software and comes with ABSOLUTELY NO WARRANTY. You are welcome to redistribute it under certain conditions. Type 'license()' or 'licence()' for distribution details. R is a collaborative project with many contributors. Type 'contributors()' for more information and 'citation()' on how to cite R or R packages in publications. Type 'demo()' for some demos, 'help()' for on-line help, or 'help.start()' for an HTML browser interface to help. Type 'q()' to quit R. > library(gnm) > set.seed(1) > DNase1 <- subset(DNase, Run == 1) > > test <- gnm(density ~ -1 + + Mult(1, Inv(Const(1) + Exp(Mult(1 + offset(-log(conc)), + Inv(1))))), + start = c(NA, 0, 1), data = DNase1, trace = TRUE) Initialising Initial Deviance = 15.301005 Running start-up iterations Start-up iteration 1. Deviance = 0.7139315 Start-up iteration 2. Deviance = 0.7139315 Running main iterations Iteration 1. Deviance = 0.4943494 Iteration 2. Deviance = 0.03849465 Iteration 3. Deviance = 0.004895279 Iteration 4. Deviance = 0.004789635 Iteration 5. Deviance = 0.004789569 Done > coef(test) Coefficients: Mult(., Inv(Exp(Mult(1 + offset(-log(conc)), Inv(1))) + Const(1))). 2.345183 Mult(1, Inv(Exp(Mult(. + offset(-log(conc)), Inv(1))) + Const(1))). 1.483093 Mult(1, Inv(Exp(Mult(1 + offset(-log(conc)), Inv(.))) + Const(1))). 1.041456 > > Logistic <- function(x, inst = NULL){ + list(predictors = list(Asym = 1, xmid = 1, scal = 1), + variables = list(substitute(x)), + term = function(predLabels, varLabels) { + paste(predLabels[1], "/(1 + exp((", predLabels[2], "-", + varLabels[1], ")/", predLabels[3], "))") + }, + start = function(theta){ + theta[3] <- 1 + theta + } + ) + } > class(Logistic) <- "nonlin" > > test <- gnm(density ~ -1 + Logistic(log(conc)), + data = DNase1, trace = TRUE) Initialising Initial Deviance = 14.865539 Running start-up iterations Start-up iteration 1. Deviance = 0.2026326 Start-up iteration 2. Deviance = 0.06637329 Running main iterations Iteration 1. Deviance = 0.01595068 Iteration 2. Deviance = 0.00479133 Iteration 3. Deviance = 0.004789569 Done > coef(test) Coefficients: Asym xmid scal 2.345168 1.483077 1.041450 > > proc.time() user system elapsed 5.116 0.056 5.180 gnm/tests/unidiff.Rout.save0000644000176000001440000000313712515147523015472 0ustar ripleyusers R Under development (unstable) (2013-09-11 r63910) -- "Unsuffered Consequences" Copyright (C) 2013 The R Foundation for Statistical Computing Platform: x86_64-unknown-linux-gnu (64-bit) R is free software and comes with ABSOLUTELY NO WARRANTY. You are welcome to redistribute it under certain conditions. Type 'license()' or 'licence()' for distribution details. R is a collaborative project with many contributors. Type 'contributors()' for more information and 'citation()' on how to cite R or R packages in publications. Type 'demo()' for some demos, 'help()' for on-line help, or 'help.start()' for an HTML browser interface to help. Type 'q()' to quit R. > library(gnm) > set.seed(1) > > unidiff <- gnm(Freq ~ educ*orig + educ*dest + + Mult(Exp(educ), orig:dest), family = poisson, + data = yaish, subset = (dest != 7)) Initialising Running start-up iterations.. Running main iterations......................................... Done > > print(unidiff$deviance, digits = 10) [1] 200.3259229 > print(unidiff$df) [1] 116 > > getContrasts(unidiff, grep("[.]educ", names(coef(unidiff)))) estimate SE quasiSE quasiVar Mult(Exp(.), orig:dest).educ1 0.0000000 0.0000000 0.09757438 0.00952076 Mult(Exp(.), orig:dest).educ2 -0.2253718 0.1611874 0.12885847 0.01660450 Mult(Exp(.), orig:dest).educ3 -0.7434235 0.2335083 0.21182122 0.04486823 Mult(Exp(.), orig:dest).educ4 -1.0389384 0.3434256 0.32609376 0.10633714 Mult(Exp(.), orig:dest).educ5 -2.2495023 0.9453761 0.93560617 0.87535890 > > proc.time() user system elapsed 4.476 0.104 4.581 gnm/tests/gammi.R0000644000176000001440000000062112515147523013446 0ustar ripleyuserslibrary(gnm) set.seed(1) yield.scaled <- wheat$yield * sqrt(3/1000) treatment <- factor(paste(wheat$tillage, wheat$summerCrop, wheat$manure, wheat$N, sep = "")) bilinear3 <- gnm(yield.scaled ~ year + treatment + instances(Mult(year, treatment), 3), family = gaussian, data = wheat) print(bilinear3$deviance, digits=10) print(bilinear3$df) gnm/tests/gammi.Rout.save0000644000176000001440000000243512515147523015140 0ustar ripleyusers R Under development (unstable) (2013-09-11 r63910) -- "Unsuffered Consequences" Copyright (C) 2013 The R Foundation for Statistical Computing Platform: x86_64-unknown-linux-gnu (64-bit) R is free software and comes with ABSOLUTELY NO WARRANTY. You are welcome to redistribute it under certain conditions. Type 'license()' or 'licence()' for distribution details. R is a collaborative project with many contributors. Type 'contributors()' for more information and 'citation()' on how to cite R or R packages in publications. Type 'demo()' for some demos, 'help()' for on-line help, or 'help.start()' for an HTML browser interface to help. Type 'q()' to quit R. > library(gnm) > set.seed(1) > > yield.scaled <- wheat$yield * sqrt(3/1000) > treatment <- factor(paste(wheat$tillage, wheat$summerCrop, wheat$manure, + wheat$N, sep = "")) > bilinear3 <- gnm(yield.scaled ~ year + treatment + + instances(Mult(year, treatment), 3), + family = gaussian, data = wheat) Initialising Running start-up iterations.. Running main iterations........................................................ Done > > print(bilinear3$deviance, digits=10) [1] 52494.66316 > print(bilinear3$df) [1] 120 > > proc.time() user system elapsed 5.832 0.140 6.001 gnm/tests/bwt.Rout.save0000644000176000001440000001437512515147523014650 0ustar ripleyusers R Under development (unstable) (2013-09-11 r63910) -- "Unsuffered Consequences" Copyright (C) 2013 The R Foundation for Statistical Computing Platform: x86_64-unknown-linux-gnu (64-bit) R is free software and comes with ABSOLUTELY NO WARRANTY. You are welcome to redistribute it under certain conditions. Type 'license()' or 'licence()' for distribution details. R is a collaborative project with many contributors. Type 'contributors()' for more information and 'citation()' on how to cite R or R packages in publications. Type 'demo()' for some demos, 'help()' for on-line help, or 'help.start()' for an HTML browser interface to help. Type 'q()' to quit R. > library(MASS) > example(birthwt) brthwt> bwt <- with(birthwt, { brthwt+ race <- factor(race, labels = c("white", "black", "other")) brthwt+ ptd <- factor(ptl > 0) brthwt+ ftv <- factor(ftv) brthwt+ levels(ftv)[-(1:2)] <- "2+" brthwt+ data.frame(low = factor(low), age, lwt, race, smoke = (smoke > 0), brthwt+ ptd, ht = (ht > 0), ui = (ui > 0), ftv) brthwt+ }) brthwt> options(contrasts = c("contr.treatment", "contr.poly")) brthwt> glm(low ~ ., binomial, bwt) Call: glm(formula = low ~ ., family = binomial, data = bwt) Coefficients: (Intercept) age lwt raceblack raceother smokeTRUE 0.82302 -0.03723 -0.01565 1.19241 0.74068 0.75553 ptdTRUE htTRUE uiTRUE ftv1 ftv2+ 1.34376 1.91317 0.68020 -0.43638 0.17901 Degrees of Freedom: 188 Total (i.e. Null); 178 Residual Null Deviance: 234.7 Residual Deviance: 195.5 AIC: 217.5 > library(nnet) > bwt.mu <- multinom(low ~ ., data = bwt) # weights: 12 (11 variable) initial value 131.004817 iter 10 value 98.029803 final value 97.737759 converged > > ## Equivalent using gnm - include unestimable main effects in model so > ## that interactions with low0 automatically set to zero, else could use > ## 'constrain' argument. > library(gnm) > bwtLong <- expandCategorical(bwt, "low", group = FALSE) > bwt.po <- gnm(count ~ low*(. - id), eliminate = id, data = bwtLong, family = + "poisson") > > coef(bwt.po) Coefficients of interest: low1 age lwt raceblack raceother 0.82301890 NA NA NA NA smokeTRUE ptdTRUE htTRUE uiTRUE ftv1 NA NA NA NA NA ftv2+ low1:age low1:lwt low1:raceblack low1:raceother NA -0.03723429 -0.01565301 1.19241318 0.74068485 low1:smokeTRUE low1:ptdTRUE low1:htTRUE low1:uiTRUE low1:ftv1 0.75552835 1.34376336 1.91316582 0.68019547 -0.43637967 low1:ftv2+ 0.17900851 > > summary(bwt.po) Call: gnm(formula = count ~ low * (. - id), eliminate = id, family = "poisson", data = bwtLong) Deviance Residuals: Min 1Q Median 3Q Max -1.35204 -0.74802 -0.08152 0.31117 1.75472 Coefficients of interest: Estimate Std. Error z value Pr(>|z|) low1 0.823019 1.244761 0.661 0.50849 age 0.000000 NA NA NA lwt 0.000000 NA NA NA raceblack 0.000000 NA NA NA raceother 0.000000 NA NA NA smokeTRUE 0.000000 NA NA NA ptdTRUE 0.000000 NA NA NA htTRUE 0.000000 NA NA NA uiTRUE 0.000000 NA NA NA ftv1 0.000000 NA NA NA ftv2+ 0.000000 NA NA NA low1:age -0.037234 0.038704 -0.962 0.33604 low1:lwt -0.015653 0.007081 -2.211 0.02706 * low1:raceblack 1.192413 0.535981 2.225 0.02610 * low1:raceother 0.740685 0.461765 1.604 0.10871 low1:smokeTRUE 0.755528 0.425035 1.778 0.07548 . low1:ptdTRUE 1.343763 0.480634 2.796 0.00518 ** low1:htTRUE 1.913166 0.720758 2.654 0.00795 ** low1:uiTRUE 0.680195 0.464350 1.465 0.14297 low1:ftv1 -0.436380 0.479411 -0.910 0.36269 low1:ftv2+ 0.179009 0.456390 0.392 0.69489 --- Signif. codes: 0 '***' 0.001 '**' 0.01 '*' 0.05 '.' 0.1 ' ' 1 (Dispersion parameter for poisson family taken to be 1) Std. Error is NA where coefficient has been constrained or is unidentified Residual deviance: 195.48 on 178 degrees of freedom AIC: 973.48 Number of iterations: 4 > > anova(bwt.po) Analysis of Deviance Table Model: poisson, link: log Response: count Terms added sequentially (first to last) Df Deviance Resid. Df Resid. Dev NULL 189 262.01 low 1 27.3376 188 234.67 low:age 1 2.7600 187 231.91 low:lwt 1 4.7886 186 227.12 low:race 2 4.4628 184 222.66 low:smoke 1 8.0834 183 214.58 low:ptd 1 8.9699 182 205.61 low:ht 1 6.4559 181 199.15 low:ui 1 2.3177 180 196.83 low:ftv 2 1.3582 178 195.48 > > drop1(bwt.po) Single term deletions Model: count ~ low + age + lwt + race + smoke + ptd + ht + ui + ftv + low:age + low:lwt + low:race + low:smoke + low:ptd + low:ht + low:ui + low:ftv - 1 Df Deviance AIC 195.48 973.48 low:age 2 196.43 970.43 low:lwt 2 201.86 975.86 low:race 3 203.32 975.32 low:smoke 2 200.46 974.46 low:ptd 2 203.77 977.77 low:ht 2 203.01 977.01 low:ui 2 198.44 972.44 low:ftv 3 197.10 969.10 > > bwt.po <- gnm(count ~ low*age - id, eliminate = id, data = bwtLong, family = + "poisson") > > add1(bwt.po, formula(terms(count~low*( . -id), data = bwtLong))) Single term additions Model: count ~ low + age + low:age - 1 Df Deviance AIC 231.91 991.91 lwt 0 231.91 991.91 race 0 231.91 991.91 smoke 0 231.91 991.91 ptd 0 231.91 991.91 ht 0 231.91 991.91 ui 0 231.91 991.91 ftv 0 231.91 991.91 > > bwt.po <- gnm(count ~ . - id, eliminate = id, data = bwtLong, family = + "poisson") > > > > proc.time() user system elapsed 3.004 0.036 3.056 gnm/tests/stereotype.Rout.save0000644000176000001440000000276712515147523016261 0ustar ripleyusers R Under development (unstable) (2013-09-11 r63910) -- "Unsuffered Consequences" Copyright (C) 2013 The R Foundation for Statistical Computing Platform: x86_64-unknown-linux-gnu (64-bit) R is free software and comes with ABSOLUTELY NO WARRANTY. You are welcome to redistribute it under certain conditions. Type 'license()' or 'licence()' for distribution details. R is a collaborative project with many contributors. Type 'contributors()' for more information and 'citation()' on how to cite R or R packages in publications. Type 'demo()' for some demos, 'help()' for on-line help, or 'help.start()' for an HTML browser interface to help. Type 'q()' to quit R. > library(gnm) > set.seed(1) > > library(nnet) > .incidence <- class.ind(backPain$pain) > .counts <- as.vector(t(.incidence)) > .rowID <- factor(t(row(.incidence))) > backPain <- backPain[.rowID, ] > backPain$pain <- C(factor(rep(levels(backPain$pain), nrow(.incidence)), + levels = levels(backPain$pain), ordered = TRUE), + treatment) > > noRelationship <- gnm(.counts ~ pain, eliminate = .rowID, + family = "poisson", data = backPain) > > oneDimensional <- update(noRelationship, + ~ . + Mult(pain, x1 + x2 + x3)) Initialising Running start-up iterations.. Running main iterations.............. Done > > print(oneDimensional$deviance, digits=10) [1] 303.1002779 > print(oneDimensional$df) [1] 493 > > proc.time() user system elapsed 2.452 0.048 2.502 gnm/tests/RChomog.R0000644000176000001440000000116512515147523013716 0ustar ripleyuserslibrary(gnm) set.seed(1) RChomog <- gnm(Freq ~ origin + destination + Diag(origin, destination) + MultHomog(origin, destination), family = poisson, data = occupationalStatus) print(RChomog$deviance, digits=10) print(RChomog$df) ### Fit an association model with homogeneous row-column effects set.seed(4) ### Set diagonal elements to NA (rather than fitting exactly) dat <- as.data.frame(friend) id <- with(dat, r == c) dat[id,] <- NA rc2 <- gnm(Freq ~ r + c + instances(MultHomog(r, c), 2), family = poisson, data = dat, iterStart = 0) print(rc2$deviance, digits=10) print(rc2$df) gnm/tests/logexcess.Rout.save0000644000176000001440000000303712515147523016041 0ustar ripleyusers R Under development (unstable) (2013-09-11 r63910) -- "Unsuffered Consequences" Copyright (C) 2013 The R Foundation for Statistical Computing Platform: x86_64-unknown-linux-gnu (64-bit) R is free software and comes with ABSOLUTELY NO WARRANTY. You are welcome to redistribute it under certain conditions. Type 'license()' or 'licence()' for distribution details. R is a collaborative project with many contributors. Type 'contributors()' for more information and 'citation()' on how to cite R or R packages in publications. Type 'demo()' for some demos, 'help()' for on-line help, or 'help.start()' for an HTML browser interface to help. Type 'q()' to quit R. > library(gnm) > set.seed(1) > > time <- c(21, 18, 33, 17, 35, 23, 43) > age <- unlist(sapply(time, seq, from = min(time))) > lowerMax <- min(age) - 1 #16 > upperMin <- max(age) + 1 #44 > leftSlope <- c(0.1, 0.2) > leftAdjust <- log(lowerMax - 14) > f <- as.factor(rep(1:2, each = 39)) > > family <- binomial(link = "cloglog") > > y <- leftSlope[f] * log(age + exp(leftAdjust) - lowerMax) > y <- family$linkinv(y) > > #don't test as Log not exported (N.B. fails if use gnm:::Log...) > #set.seed(1) > #test <- gnm(y ~ -1 + Mult(f, Log(offset(age - lowerMax) + Exp(1))), > # family = binomial(link = "cloglog")) > > #set.seed(1) > #test <- gnm(y ~ 0 + Nonlin(LogExcess(age, side = "left", > # slopeFormula = ~ 0 + f)), > # family = binomial(link = "cloglog")) > #coef(test) > > proc.time() user system elapsed 1.636 0.020 1.662 gnm/tests/diagonalRef.R0000644000176000001440000000122112515147523014564 0ustar ripleyuserslibrary(gnm) set.seed(1) count <- with(voting, percentage/100 * total) yvar <- cbind(count, voting$total - count) classMobility <- gnm(yvar ~ Dref(origin, destination), family = binomial, data = voting) print(classMobility$deviance, digits = 10) print(classMobility$df) upward <- with(voting, origin != 1 & destination == 1) downward <- with(voting, origin == 1 & destination != 1) socialMobility <- gnm(yvar ~ Dref(origin, destination, delta = ~ 1 + downward + upward), family = binomial, data = voting) print(socialMobility$deviance, digits = 10) print(socialMobility$df) gnm/tests/RC.R0000644000176000001440000000165412515147523012667 0ustar ripleyuserslibrary(gnm) set.seed(1) mentalHealth$MHS <- C(mentalHealth$MHS, treatment) mentalHealth$SES <- C(mentalHealth$SES, treatment) RC1model <- gnm(count ~ SES + MHS + Mult(-1 + SES, -1 + MHS), family = poisson, data = mentalHealth) print(RC1model$deviance, digits = 10) print(RC1model$df) rowProbs <- with(mentalHealth, tapply(count, SES, sum) / sum(count)) colProbs <- with(mentalHealth, tapply(count, MHS, sum) / sum(count)) mu <- getContrasts(RC1model, pickCoef(RC1model, "[.]SES"), ref = rowProbs, scaleRef = rowProbs, scaleWeights = rowProbs) nu <- getContrasts(RC1model, pickCoef(RC1model, "[.]MHS"), ref = colProbs, scaleRef = colProbs, scaleWeights = colProbs) all.equal(sum(mu$qv[,1] * rowProbs), 0) all.equal(sum(nu$qv[,1] * colProbs), 0) all.equal(sum(mu$qv[,1]^2 * rowProbs), 1) all.equal(sum(nu$qv[,1]^2 * colProbs), 1) gnm/tests/logexcess.R0000644000176000001440000000141712515147523014354 0ustar ripleyuserslibrary(gnm) set.seed(1) time <- c(21, 18, 33, 17, 35, 23, 43) age <- unlist(sapply(time, seq, from = min(time))) lowerMax <- min(age) - 1 #16 upperMin <- max(age) + 1 #44 leftSlope <- c(0.1, 0.2) leftAdjust <- log(lowerMax - 14) f <- as.factor(rep(1:2, each = 39)) family <- binomial(link = "cloglog") y <- leftSlope[f] * log(age + exp(leftAdjust) - lowerMax) y <- family$linkinv(y) #don't test as Log not exported (N.B. fails if use gnm:::Log...) #set.seed(1) #test <- gnm(y ~ -1 + Mult(f, Log(offset(age - lowerMax) + Exp(1))), # family = binomial(link = "cloglog")) #set.seed(1) #test <- gnm(y ~ 0 + Nonlin(LogExcess(age, side = "left", # slopeFormula = ~ 0 + f)), # family = binomial(link = "cloglog")) #coef(test) gnm/tests/RC.Rout.save0000644000176000001440000000350212515147523014346 0ustar ripleyusers R Under development (unstable) (2013-09-11 r63910) -- "Unsuffered Consequences" Copyright (C) 2013 The R Foundation for Statistical Computing Platform: x86_64-unknown-linux-gnu (64-bit) R is free software and comes with ABSOLUTELY NO WARRANTY. You are welcome to redistribute it under certain conditions. Type 'license()' or 'licence()' for distribution details. R is a collaborative project with many contributors. Type 'contributors()' for more information and 'citation()' on how to cite R or R packages in publications. Type 'demo()' for some demos, 'help()' for on-line help, or 'help.start()' for an HTML browser interface to help. Type 'q()' to quit R. > library(gnm) > set.seed(1) > > mentalHealth$MHS <- C(mentalHealth$MHS, treatment) > mentalHealth$SES <- C(mentalHealth$SES, treatment) > RC1model <- gnm(count ~ SES + MHS + + Mult(-1 + SES, -1 + MHS), + family = poisson, data = mentalHealth) Initialising Running start-up iterations.. Running main iterations........ Done > > print(RC1model$deviance, digits = 10) [1] 3.570562451 > print(RC1model$df) [1] 8 > > rowProbs <- with(mentalHealth, tapply(count, SES, sum) / sum(count)) > colProbs <- with(mentalHealth, tapply(count, MHS, sum) / sum(count)) > mu <- getContrasts(RC1model, pickCoef(RC1model, "[.]SES"), + ref = rowProbs, scaleRef = rowProbs, + scaleWeights = rowProbs) > nu <- getContrasts(RC1model, pickCoef(RC1model, "[.]MHS"), + ref = colProbs, scaleRef = colProbs, + scaleWeights = colProbs) > all.equal(sum(mu$qv[,1] * rowProbs), 0) [1] TRUE > all.equal(sum(nu$qv[,1] * colProbs), 0) [1] TRUE > all.equal(sum(mu$qv[,1]^2 * rowProbs), 1) [1] TRUE > all.equal(sum(nu$qv[,1]^2 * colProbs), 1) [1] TRUE > > proc.time() user system elapsed 3.420 0.080 3.509 gnm/tests/bwt.R0000644000176000001440000000136612515147523013157 0ustar ripleyuserslibrary(MASS) example(birthwt) library(nnet) bwt.mu <- multinom(low ~ ., data = bwt) ## Equivalent using gnm - include unestimable main effects in model so ## that interactions with low0 automatically set to zero, else could use ## 'constrain' argument. library(gnm) bwtLong <- expandCategorical(bwt, "low", group = FALSE) bwt.po <- gnm(count ~ low*(. - id), eliminate = id, data = bwtLong, family = "poisson") coef(bwt.po) summary(bwt.po) anova(bwt.po) drop1(bwt.po) bwt.po <- gnm(count ~ low*age - id, eliminate = id, data = bwtLong, family = "poisson") add1(bwt.po, formula(terms(count~low*( . -id), data = bwtLong))) bwt.po <- gnm(count ~ . - id, eliminate = id, data = bwtLong, family = "poisson") gnm/tests/doubleUnidiff.R0000644000176000001440000000051112515147523015131 0ustar ripleyuserslibrary(gnm) set.seed(1) doubleUnidiff <- gnm(Freq ~ election*vote + election*class*religion + Mult(Exp(election), religion:vote) + Mult(Exp(election), class:vote), family = poisson, data = cautres) print(doubleUnidiff$deviance, digits=10) print(doubleUnidiff$df) gnm/tests/unidiff.R0000644000176000001440000000045612515147523014006 0ustar ripleyuserslibrary(gnm) set.seed(1) unidiff <- gnm(Freq ~ educ*orig + educ*dest + Mult(Exp(educ), orig:dest), family = poisson, data = yaish, subset = (dest != 7)) print(unidiff$deviance, digits = 10) print(unidiff$df) getContrasts(unidiff, grep("[.]educ", names(coef(unidiff)))) gnm/tests/diagonalRef.Rout.save0000644000176000001440000000305412515147523016257 0ustar ripleyusers R Under development (unstable) (2013-09-11 r63910) -- "Unsuffered Consequences" Copyright (C) 2013 The R Foundation for Statistical Computing Platform: x86_64-unknown-linux-gnu (64-bit) R is free software and comes with ABSOLUTELY NO WARRANTY. You are welcome to redistribute it under certain conditions. Type 'license()' or 'licence()' for distribution details. R is a collaborative project with many contributors. Type 'contributors()' for more information and 'citation()' on how to cite R or R packages in publications. Type 'demo()' for some demos, 'help()' for on-line help, or 'help.start()' for an HTML browser interface to help. Type 'q()' to quit R. > library(gnm) > set.seed(1) > > count <- with(voting, percentage/100 * total) > yvar <- cbind(count, voting$total - count) > > classMobility <- gnm(yvar ~ Dref(origin, destination), + family = binomial, data = voting) Initialising Running main iterations........ Done > > print(classMobility$deviance, digits = 10) [1] 21.22093265 > print(classMobility$df) [1] 19 > > upward <- with(voting, origin != 1 & destination == 1) > downward <- with(voting, origin == 1 & destination != 1) > > socialMobility <- gnm(yvar ~ Dref(origin, destination, + delta = ~ 1 + downward + upward), + family = binomial, data = voting) Initialising Running main iterations........... Done > > print(socialMobility$deviance, digits = 10) [1] 18.97407303 > print(socialMobility$df) [1] 17 > > proc.time() user system elapsed 4.296 0.064 4.363 gnm/tests/logistic.R0000644000176000001440000000152312515147523014173 0ustar ripleyuserslibrary(gnm) set.seed(1) DNase1 <- subset(DNase, Run == 1) test <- gnm(density ~ -1 + Mult(1, Inv(Const(1) + Exp(Mult(1 + offset(-log(conc)), Inv(1))))), start = c(NA, 0, 1), data = DNase1, trace = TRUE) coef(test) Logistic <- function(x, inst = NULL){ list(predictors = list(Asym = 1, xmid = 1, scal = 1), variables = list(substitute(x)), term = function(predLabels, varLabels) { paste(predLabels[1], "/(1 + exp((", predLabels[2], "-", varLabels[1], ")/", predLabels[3], "))") }, start = function(theta){ theta[3] <- 1 theta } ) } class(Logistic) <- "nonlin" test <- gnm(density ~ -1 + Logistic(log(conc)), data = DNase1, trace = TRUE) coef(test) gnm/tests/doubleUnidiff.Rout.save0000644000176000001440000000224312515147523016622 0ustar ripleyusers R Under development (unstable) (2013-09-11 r63910) -- "Unsuffered Consequences" Copyright (C) 2013 The R Foundation for Statistical Computing Platform: x86_64-unknown-linux-gnu (64-bit) R is free software and comes with ABSOLUTELY NO WARRANTY. You are welcome to redistribute it under certain conditions. Type 'license()' or 'licence()' for distribution details. R is a collaborative project with many contributors. Type 'contributors()' for more information and 'citation()' on how to cite R or R packages in publications. Type 'demo()' for some demos, 'help()' for on-line help, or 'help.start()' for an HTML browser interface to help. Type 'q()' to quit R. > library(gnm) > set.seed(1) > > doubleUnidiff <- gnm(Freq ~ election*vote + election*class*religion + + Mult(Exp(election), religion:vote) + + Mult(Exp(election), class:vote), + family = poisson, data = cautres) Initialising Running start-up iterations.. Running main iterations........... Done > > print(doubleUnidiff$deviance, digits=10) [1] 133.0429799 > print(doubleUnidiff$df) [1] 78 > > proc.time() user system elapsed 2.768 0.112 2.894 gnm/tests/biplot.Rout.save0000644000176000001440000000224412515147523015335 0ustar ripleyusers R Under development (unstable) (2013-09-11 r63910) -- "Unsuffered Consequences" Copyright (C) 2013 The R Foundation for Statistical Computing Platform: x86_64-unknown-linux-gnu (64-bit) R is free software and comes with ABSOLUTELY NO WARRANTY. You are welcome to redistribute it under certain conditions. Type 'license()' or 'licence()' for distribution details. R is a collaborative project with many contributors. Type 'contributors()' for more information and 'citation()' on how to cite R or R packages in publications. Type 'demo()' for some demos, 'help()' for on-line help, or 'help.start()' for an HTML browser interface to help. Type 'q()' to quit R. > library(gnm) > set.seed(1) > > biplotModel <- gnm(y ~ -1 + instances(Mult(site, variety), 2), + family = wedderburn, data = barley) Initialising Running start-up iterations.. Running main iterations......................................................... ............................................................................ Done > > print(biplotModel$deviance, digits = 10) [1] 41.02423508 > print(biplotModel$df) [1] 56 > > > proc.time() user system elapsed 4.220 0.104 4.343 gnm/tests/nls.R0000644000176000001440000000050312515147523013147 0ustar ripleyuserslibrary(gnm) set.seed(1) DNase1 <- subset(DNase, Run == 1) fm3DNase1.2 <- gnm(density ~ -1 + Mult(1, Inv(Const(1) + Exp(1 + Mult(offset(-log(conc)), Inv(1))))), start = c(NA, 0, 1), data = DNase1, trace = TRUE) coef(fm3DNase1.2) gnm/tests/stereotype.R0000644000176000001440000000120512515147523014556 0ustar ripleyuserslibrary(gnm) set.seed(1) library(nnet) .incidence <- class.ind(backPain$pain) .counts <- as.vector(t(.incidence)) .rowID <- factor(t(row(.incidence))) backPain <- backPain[.rowID, ] backPain$pain <- C(factor(rep(levels(backPain$pain), nrow(.incidence)), levels = levels(backPain$pain), ordered = TRUE), treatment) noRelationship <- gnm(.counts ~ pain, eliminate = .rowID, family = "poisson", data = backPain) oneDimensional <- update(noRelationship, ~ . + Mult(pain, x1 + x2 + x3)) print(oneDimensional$deviance, digits=10) print(oneDimensional$df) gnm/tests/biplot.R0000644000176000001440000000032112515147523013642 0ustar ripleyuserslibrary(gnm) set.seed(1) biplotModel <- gnm(y ~ -1 + instances(Mult(site, variety), 2), family = wedderburn, data = barley) print(biplotModel$deviance, digits = 10) print(biplotModel$df) gnm/tests/RChomog.Rout.save0000644000176000001440000000307012515147523015400 0ustar ripleyusers R Under development (unstable) (2013-09-11 r63910) -- "Unsuffered Consequences" Copyright (C) 2013 The R Foundation for Statistical Computing Platform: x86_64-unknown-linux-gnu (64-bit) R is free software and comes with ABSOLUTELY NO WARRANTY. You are welcome to redistribute it under certain conditions. Type 'license()' or 'licence()' for distribution details. R is a collaborative project with many contributors. Type 'contributors()' for more information and 'citation()' on how to cite R or R packages in publications. Type 'demo()' for some demos, 'help()' for on-line help, or 'help.start()' for an HTML browser interface to help. Type 'q()' to quit R. > library(gnm) > set.seed(1) > > RChomog <- gnm(Freq ~ origin + destination + Diag(origin, destination) + + MultHomog(origin, destination), family = poisson, + data = occupationalStatus) Initialising Running start-up iterations.. Running main iterations........ Done > > print(RChomog$deviance, digits=10) [1] 32.56097626 > print(RChomog$df) [1] 34 > > > ### Fit an association model with homogeneous row-column effects > set.seed(4) > ### Set diagonal elements to NA (rather than fitting exactly) > dat <- as.data.frame(friend) > id <- with(dat, r == c) > dat[id,] <- NA > rc2 <- gnm(Freq ~ r + c + instances(MultHomog(r, c), 2), + family = poisson, data = dat, iterStart = 0) Initialising Running main iterations.................. Done > > print(rc2$deviance, digits=10) [1] 1006.906781 > print(rc2$df) [1] 810 > > proc.time() user system elapsed 5.768 0.112 5.894 gnm/src/0000755000176000001440000000000012515664670011666 5ustar ripleyusersgnm/src/Makevars0000644000176000001440000000004112515664670013355 0ustar ripleyusersPKG_LIBS = $(BLAS_LIBS) $(FLIBS) gnm/src/gnm.c0000644000176000001440000001140012515664670012607 0ustar ripleyusers/* Copyright (C) 2005, 2006, 2008-2010 Heather Turner */ /* */ /* This program is free software; you can redistribute it and/or modify */ /* it under the terms of the GNU General Public License as published by */ /* the Free Software Foundation; either version 2 or 3 of the License */ /* (at your option). */ /* */ /* This program is distributed in the hope that it will be useful, */ /* but WITHOUT ANY WARRANTY; without even the implied warranty of */ /* MERCHANTABILITY or FITNESS FOR A PARTICULAR PURPOSE. See the */ /* GNU General Public License for more details. */ /* */ /* A copy of the GNU General Public License is available at */ /* http://www.r-project.org/Licenses/ */ /* vector * matrix */ # include /* for length */ # include /* for dgemm */ /* copied from src/main/array.c */ static void matprod(double *x, int nrx, int ncx, double *y, int nry, int ncy, double *z) { char *transa = "N", *transb = "N"; int i, j, k; double one = 1.0, zero = 0.0, sum; Rboolean have_na = FALSE; if (nrx > 0 && ncx > 0 && nry > 0 && ncy > 0) { /* Don't trust the BLAS to handle NA/NaNs correctly: PR#4582 * The test is only O(n) here */ for (i = 0; i < nrx*ncx; i++) if (ISNAN(x[i])) {have_na = TRUE; break;} if (!have_na) for (i = 0; i < nry*ncy; i++) if (ISNAN(y[i])) {have_na = TRUE; break;} if (have_na) { for (i = 0; i < nrx; i++) for (k = 0; k < ncy; k++) { sum = 0.0; for (j = 0; j < ncx; j++) sum += x[i + j * nrx] * y[j + k * nry]; z[i + k * nrx] = sum; } } else F77_CALL(dgemm)(transa, transb, &nrx, &ncy, &ncx, &one, x, &nrx, y, &nry, &zero, z, &nrx); } else /* zero-extent operations should return zeroes */ for(i = 0; i < nrx*ncy; i++) z[i] = 0; } /* computes matrix product between submatrix of M and vector v */ SEXP submatprod(SEXP M, SEXP v, SEXP am, SEXP nr, SEXP nc) { R_len_t a = INTEGER(am)[0], nrm = INTEGER(nr)[0], ncm = INTEGER(nc)[0]; SEXP ans; PROTECT(ans = allocVector(REALSXP, nrm)); matprod(REAL(M) + a, nrm, ncm, REAL(v), ncm, 1, REAL(ans)); UNPROTECT(1); return(ans); } /* computes elementwise product between submatrix of M and vector v then puts result in submatrix of X */ SEXP subprod(SEXP X, SEXP M, SEXP v, SEXP a, SEXP z, SEXP nv) { R_len_t i = INTEGER(a)[0], j = 0, last = INTEGER(z)[0], len_v = INTEGER(nv)[0]; double *dX, *dM, *dv; dX = REAL(X); dM = REAL(M); dv = REAL(v); for ( ; i <= last; j = (++j == len_v) ? 0 : j) { dX[i] = dM[i] * dv[j]; i++; } return(X); } /* put results of nonlin localDesignFunction in X */ SEXP nonlin(SEXP X, SEXP a, SEXP z, SEXP expr, SEXP rho) { R_len_t i = INTEGER(a)[0], i1 = 0, last = INTEGER(z)[0]; SEXP ans; double *dX, *dans; dX = REAL(X); PROTECT(ans = coerceVector(eval(expr, rho), REALSXP)); dans = REAL(ans); for ( ; i <= last;) { dX[i++] = dans[i1++]; } UNPROTECT(1); return(X); } /* Computes elementwise products between submatrices of base matrix M and columns of gradient matrix V, summing 'common' results an putting result in submatrix of X. This version has start point in M, V and X for each "term" */ SEXP newsubprod(SEXP M, SEXP V, SEXP X, SEXP a, SEXP b, SEXP c, SEXP nt, SEXP lt, SEXP ls, SEXP nr, SEXP nc, SEXP max) { /* currently set up for single term so nt = 1 and all integers here */ int i, j, k, l, *start, *end, *common, nrow = INTEGER(nr)[0], n = INTEGER(max)[0], final = INTEGER(nt)[0], *jump, *ia, *ib; double *p[n], *q[n], *dM, *dV, *dX; dM = REAL(M); dV = REAL(V); dX = REAL(X); start = INTEGER(c); end = INTEGER(ls); common = INTEGER(nc); jump = INTEGER(lt); ia = INTEGER(a); ib = INTEGER(b); for (i = 0; i < final; i++){ p[0] = &dM[ia[i]]; q[0] = &dV[ib[i]]; for (l = 1; l < common[i]; l++) { p[l] = p[l - 1] + jump[i]; q[l] = q[l - 1] + nrow; } k = 0; for (j = start[i]; j < end[i]; j++, k = (++k == nrow) ? 0 : k) { dX[j] = *(p[0])++ * q[0][k]; for (l = 1; l < common[i]; l++){ dX[j] += *(p[l])++ * q[l][k]; } } } return(X); } /* computes single column of design matrix */ SEXP onecol(SEXP M, SEXP V, SEXP a, SEXP lt, SEXP nr, SEXP nc) { int j, k, l, nrow = INTEGER(nr)[0], common = INTEGER(nc)[0], jump; double *p[common], *q[common], *dcol; SEXP col; jump = INTEGER(lt)[0]; p[0] = &REAL(M)[INTEGER(a)[0]]; q[0] = &REAL(V)[0]; for (l = 1; l < common; l++) { p[l] = p[l - 1] + jump; q[l] = q[l - 1] + nrow; } k = 0; PROTECT(col = allocVector(REALSXP, nrow)); dcol = REAL(col); for (j = 0; j < nrow; j++, k = (++k == nrow) ? 0 : k) { dcol[j] = *(p[0])++ * q[0][k]; for (l = 1; l < common; l++){ dcol[j] += *(p[l])++ * q[l][k]; } } UNPROTECT(1); return(col); } gnm/NAMESPACE0000755000176000001440000000342312515147523012314 0ustar ripleyusersuseDynLib(gnm) export(asGnm, checkEstimable, Const, Diag, Dref, DrefWeights, exitInfo, Exp, expandCategorical, getContrasts, getModelFrame, gnm, instances, Inv, #Log, Logit, meanResiduals, MPinv, Mult, MultHomog, ofInterest, "ofInterest<-", parameters, pickCoef, qrSolve, #Raise, residSVD, se, Symm, termPredictors, Topo, wedderburn) import(MASS, stats, Matrix) # classes importFrom(graphics, plot) # in plot.gnm; plot.profile.gnm importFrom(nnet, class.ind) # in gnmTools, expandCategorical importFrom(qvcalc, qvcalc) # in getContrasts importFrom(relimp, pickFrom) # in getContrasts S3method(add1, gnm) S3method(addterm, gnm) S3method(alias, gnm) S3method(anova, gnm) S3method(asGnm, glm) S3method(asGnm, lm) S3method(asGnm, default) S3method(boxcox, gnm) S3method(coef, gnm) S3method(confint, gnm) S3method(confint, profile.gnm) S3method(cooks.distance, gnm) S3method(dfbeta, gnm) S3method(dfbetas, gnm) S3method(drop1, gnm) S3method(dropterm, gnm) S3method(dummy.coef, gnm) S3method(effects, gnm) S3method(fitted, gnm) S3method(hatvalues, gnm) S3method(influence, gnm) S3method(kappa, gnm) S3method(labels, gnm) S3method(logtrans, gnm) S3method(model.frame, gnm) S3method(model.matrix, gnm) S3method(plot, gnm) S3method(plot, profile.gnm) S3method(predict, gnm) S3method(print, gnm) S3method(print, coef.gnm) S3method(print, profile.gnm) S3method(print, summary.gnm) S3method(print, vcov.gnm) S3method(print, meanResiduals) S3method(profile, gnm) S3method(proj, gnm) S3method(residuals, gnm) S3method(rstandard, gnm) S3method(rstudent, gnm) S3method(summary, gnm) S3method(summary, meanResiduals) S3method(termPredictors, default) S3method(termPredictors, gnm) S3method(update, gnm) S3method(variable.names, gnm) S3method(vcov, gnm) S3method(weights, gnm) gnm/demo/0000755000176000001440000000000012515147523012014 5ustar ripleyusersgnm/demo/gnm.R0000644000176000001440000000421712515147523012724 0ustar ripleyusersmessage("1. Set seed as gnm returns random parameterization") set.seed(1) { if (interactive()) { cat("\n3. Type to fit (linear) uniform association model, ", "\n using Diag() to fit diagonal effects: ") readline() } else message("2. Fit (linear) uniform association model, using Diag() to fit", " diagonal effects") } Rscore <- scale(as.numeric(row(occupationalStatus)), scale = FALSE) Cscore <- scale(as.numeric(col(occupationalStatus)), scale = FALSE) Uniform <- gnm(Freq ~ origin + destination + Diag(origin, destination) + Rscore:Cscore, family = poisson, data = occupationalStatus) summary(Uniform) { if (interactive()) { cat("\n3. Type to fit an association model using Mult() to fit", "\n separate row and column effects:") readline() } else message("3. Fit an association model using Mult() to fit separate row and ", "column effects") } RC <- gnm(Freq ~ origin + destination + Diag(origin, destination) + Mult(origin, destination), family = poisson, data = occupationalStatus) summary(RC) { if (interactive()) { cat("\n4. Type to fit an association model using MultHomog()", "\n to fit homogeneous row-column effects:") readline() } else message("4. Fit an association model using MultHomog()\n", "to fit homogeneous row-column effects") } RChomog <- gnm(Freq ~ origin + destination + Diag(origin, destination) + MultHomog(origin, destination), family = poisson, data = occupationalStatus) summary(RChomog) { if (interactive()) { cat("\n5. Type to compare models using anova:") readline() } else message("5. Compare models using anova") } anova(Uniform, RChomog, RC) message("6. Produce diagnostic plots for RChomog") plot(RChomog) message("7. Get simple constrasts of homogeneous row-column effects") getContrasts(RChomog, grep("MultHomog", names(coef(RChomog)))) message("End of demo. \n", "See vignette(\"gnmOverview\", package = \"gnm\") for full manual.") gnm/demo/00Index0000644000176000001440000000006312515147523013145 0ustar ripleyusersgnm Fitting generalized nonlinear models with gnm gnm/data/0000755000176000001440000000000012515664670012010 5ustar ripleyusersgnm/data/yaish.rda0000644000176000001440000000101012515664670013605 0ustar ripleyusers‹­VMKÃ@MÅZPöÐC¥”""bb?TD#µ~ Ac[´Ð*ØzðÖŸæOñ'xñZÜ63 ûÜ ’7›y;û6ÉÅþµŸºNÑ49Óêî(3¡ š¢Í)œy Û½G"gAu¾‚,[d\c\a¼ûˆ0Íý<Ùû'ŒË€>ãcx!‘‰.ããQ„»ŸÀOòyŒg1¼ Œ%ÆË‘OšŽŸ±çÓýÓ/XÞY—ç9ÆúÀ\ßðtÖ•„zå€×*ÄKC=+°?K‡ìùóæA/ÈëÆÔ㈤>Ç Ù¿CX_xJÞMÆ2¬÷Oâm3VctàÚû’—P¿ k™G1º‹kZOdç¥ý|ØÏœáyKÇäuíó÷oüu¼"ÔuçC¿~ð¾ X'¬—SÆÙ¯úÌš|4¿Ëw3ÞcxÜH>ÐYm`êÑ3ë‹õÒ:*‚n×`ëçÚ÷Gû ¿Ôßçy–bÞ‡².â¡Ž®ómÄð+ÚÏѵ¿ž³Iý&Íû׸?›ñwší®‚ù‘9ú¦«k6º ·¤r{ »­žúXd_wìOybøb¬‹Q£Ì“f'›ÄFEŒê¿…1gx}ãØ?L´š¯ ±Ÿ_Úb7[½¾Â!itžÑûá}§yÓð¸˜'kY gnm/data/barleyHeights.rda0000644000176000001440000000126712515664670015300 0ustar ripleyusers‹íV?oÓ@?ÛiÚ„–"ubcbBU“ÚlJìó]bgª)N‹”D lL°°u*S'&6¾¦~&¾@'–†sî÷\;†zCœôòžŸßßß½ó…?ô«e¿Ì3™eª_K‰%0ƒXIñµ§áhMÑóãɘ1kC)c«›ŠÞÙÂfñ²ùF-^¶w®yçŽÖ{gšw<­ïžà} œì/4—o`ÿq¿kîÃÞG>~í»ýxÙüÝÏš;ðãÈÏá÷äìßÙ×ÜÃqZ°x/ÏEÜvqP—‡þ9úÄ¡wPGpÃ3áHuIà$©Äi~Ôy[°sQ_¸%8,àÏ¡ zìõOñ9ôvMàÕ@¿¨o¼ ü Â×=ÉâÅ÷³ûÙD¿{ðÛC_ pY?Uëƒí½ÕÏbõ«Zßìðð>égÂÉG=öš;èiÞäOWŸæ‰æP¾×zq[Û ì\ü:° îiÞ~ðè‘þù( y‹Ú®‹~©N‰<ûå#¿D}ìzè/À¾À%y¾Èž+â]â´/ì÷óOçÐTáp¶0§çÙyõhÎ Ø ƒú„¬Åßù‡æúdæ +rÐR*æ å´’ƒJקÌETD¯¢ºØ­ù[­-Tê;Û)ù~J~’wRònJ®_É»[)¹’’« •, Â1’”×&Ñ’.†›@ˆoÙŠVã[èzj ©M^þïû¯ûÆ#eÎ0Eë˜"ƒæάl%Ò•®šHµDÚN$š}ƒÌ ²2ÈÈHlèpt2 :F]gÍþ6âÙq¾ˆèDXd~4ÿÏHi…#È+‡Ññp2}ý!QùY8 7û#t!Qi4|½IÉbÍø~šÍfS]»üÑ€òðÒ gnm/data/backPain.rda0000644000176000001440000000113712515664670014212 0ustar ripleyusers‹ÍÕÝnÚ0à8tEš6u½Düïó]À´£žf`6Ô„ €Ú]}7Cãw&J+ìàk_Û‰;|ýüÀ&“$IÒ„¦þ/õ1ó!!I–Üúÿï¾óÇ/Åj$ôÓ~Ð×{_îpÌÛ*j¨ÿE®\}s§ÿYõÝ›¡ëBÛºftà˜¡yN“ î}ßúɉ}¥çîqh¿ôÌ÷æoÏ×9ïÐÐ<ÿr­¯ñŽÄís×õVƒ\ø<ù¢áÞìÛ#_ãè†vÖæxšæ¶©Ÿ§ñTáb0 ‘0!ÈT:‚ ëÈgHXeŽUæI I$…¤‘  ƒÁ`0 ƒÁ`0 ƒÁà08 ƒÃà08 ƒÃà08 CÀ0 CÀ0 CÀ0$ CÂ0$ CÂ0$ CÁP0 CÁP0 CÁÐ04 CÃÐ04 CÃÐ04 ÃÀ00 ÃÀ00 ÃÀ°0, ð0, ð0l0h>›ý‰ùë·€þj_ýðM˜,Š]1]6û/ãáÓ¼üɪ%&ü gnm/data/House2001.rda0000644000176000001440000001201412515664670014064 0ustar ripleyusers‹í][sÛÈrl’º8[u*©­$§R¹¿$UËXö®ÏžGI”DY+¤vuÎãX‚ÉÔþUþÝI”Ð=Ô @HJ–kí’…æ æÒ·¯»1„{?½ÛüÓ¦eY¯¬×¯àß×pÙ€ ˶ÖüÞèª$’ï޾ݲ¬×| ¿¿öÿuºûiÜÿèôc¿‹ÚŠî/ÿá}’~vÁ¸?Ecµ7ü°KÖݨ1—]Òn—ð©lβ1LY>vÏ+Ƙ †î4JdY¶'ŽÌªöÚ`êWæUëàØEÃ`ÌÇÖþºdŽ-UÝ×`ê´‰=4 äÉÕo®ü«tƒë³l†Ž½2ô/U¾±QÃç”ébƒ¹¶FMÈñávÅúªäû ï¶|y-ý®~7k½ûÔîä‡ùÅi~Ñ{0BÓñE”€Ã^ 'V!\ý~TSkîO¾°üÚJKÚ´ßÖ#´]ÐþØçç+·èÞ¹?iñœEóíç±½>äkiùZìþTí·èÞÂ}¥Å2«úœ»«DOÊälU苽 >غXGïæÆI«m;oÕÞæîOùsí©¨oá=iµý”íù1ßbŒÁ’}…¿âÈãÑyÒbݲË|Ã÷TÚ]Ê·?+eîÇ€—ƒ_]äø»cÊdø0qäðÜ®!3»Â¯”ù>»®O³êcœUC¹Xµ,Œ3õexËÝ—eñ°·J—,&~ÛOlÃu- ãl™VÙ…I,Xˆ=i5®Ù û\ãLõšk£&1œm(‹©£&ñ±Íäåª0ÎfæVA.Ãm¦\Æqì»ãR¾Ì–q&{´­Åb7‹ákåqO„q&ØS'3Ÿ¥æqKÆ8â­E1Ž‹÷‹`Ü£ëYã8öò¥c\ù›èè21nÙyœ]S—áѯx˜šÕ·Lò8#¹}Ay'Ö_´V¹LŒ«ƒ;ÜXqgš_=Æqü‹]Q)WÇ>jWéâvbŠqOU«ä`Áªó¸'øtžïO‰qsyIÊÏý«|47YÆYOˆqœúR<®*7îuân»®æ¯Ë8sR$w‹9‡ÅäG‘à`V%_Kü¥é™“2a­²(HËëM#ãÄ}•µJ®›ærœzCaÜýXM)]r­2åמ¤V™.X«LÍcH?\´¯EãÑ9ÿ2â Fö¼ï¬}UÇ>ƒ˜ï§Èÿ<†q¦µ{N3'i¹nÛ%>ÍÄ?ØVõ÷±Lê©E¼[U­Ò$žµ|(Ô»´8Geçqæ\Egßš|wà)ŸÇ})gNVú<ΪÎãLꌦgïWy®r‘Z%÷,§æi‚quôñ‹z—òôoiÏãR¾mšàÇ"ÏãLò¸ºµ±*LåèŒ)Æ-û\%×ÏÖ9çhR«4ŽÓ¹tʈµR¾¿]EWgÏ–Á˜å¦–¡¿äè·nUÄÖ–¡~°y–ò0Î42õc¦937®5Ña“s¾…ó¥üx‹»žJþ¦×1êžUâÞk×Øgå´‹è<«V’òcΪ<®Ê§˜Ôcjù ΘqHÝçq«Îã¸5™/õûq‹<~¶<.åÇk¦¶´ÌïTÅ2ËÊãŒ1®Fúœy\Z¥‘-äfœ3'Ïñý8‹)ÿeäqUv´’3'gW`\Õy7.Æ­äÌÉ‚W”Ǚ愦k®Š“8¾¯ÊX†¾Ò¤^mkrž©—Õ¬«önUÔ½95JnLÆ> ËÄNÎYçœnÝ8Ø^ O]û5­©²ýNjVK¨²?Žÿ)zg’×›Ö™MÏmsdkz¾Ùä½1Ugƒ¸¾Œ‹q¥ú–òózÓ³™&XUç=\ߪ÷áäqUk6OY~#ååA‹äq¦ØµhW÷\¥IŒdYüó`¬ó–ÙsHÓØvQ|5y^dâS9çõVy®²4§)ó邼²W7>\ÆY5ý£iŽÉi7yÏÀ²Þ eqtñy™Ø*·^cZ«4}ÆòÒÞWùRk•¿9Yð='†±àS¿çÄTæOR«LW[«|ÉgNêȵʺgNLbèR›Iy¾…óŽ§8sò\µÊß0î?[ÆYulí+Ã8N|Ug}%Ç}Vð5`GÇž ã ×øÆ-Œq¦ß·|ôþ'Â8“gM&gïVõpÓ³¹‹Ô*MÎé›âæc± ã8¼äÔ*MkŒœZ‹I Êô;à¬gȆµJvn’–èʊ󸪳ñeïæøO“8Œ«{¬÷œ”ÕvµG‹ÉÆÕ©U²ž§­+5ó /ñ]^\ŒãúûºøÆþ>RÊð/+ÎãJc«ÔŒÿŒ³˜òjŒ³¬Å0Λªd[¥µÞåÅÐe`\FØ1g)Æ•Ìo¥<Û­‹á:ÒúÇÍÓ9>æ91®R7S¾sxUãÒêû–Žqé‚y\Z£V™>­Òä]=uêV_M­2]N­’;ÆÒj•éWV«´VðNæ´F­2}¤VÉȾ´ÿ?îÙj•)ÿþ*;Y4[Y­25³³—R«4Â8F|ÿbÎœ¤«Íã¾–3'lŒ«°—Œq «4¼gåÏã¬ÕÖ*ëê÷—þ¤.ŠqvM[YÆÕοŸãLÏ–Õµ}îû –…qußq»HgÕ°eî3!níË´VÉÅÒ:ß«©“ÇYL_^º&Æq÷üyÜ"1’És>Vü2}NZí›Mñ«ª÷Ì…:ŸVûÖ*½/«ç>Æ™Ô6ë|œSqjúuÏpê¢%õâÆ«Ý°6àw3cÁÅïàçÛüÉãé=ñÝÞû÷ïÛ;"…É$v¦Ÿ®eèŠÄoÿn¼ëóoý=}hï…C¨±çD½Ä—/“I¨®…ß>U÷]ÿ†kŽËsñ¹'‰{* 0߇ö‰CO ç»Ð|ÛíýØÞ‹bßõð=y¥5ÿ]·÷ÃÛ÷í} ÿ,T±|0Å·ÝÞV»ï¸Jù?«ÄqeQã¿Bãï?´aàÝ$îIGŽç{"bÇóµ­üS·÷îýÚ»b<Þ0Ø÷8²'¯T8¦n¿ïfÌ÷}1ìÈH:¡øÕ–ÿÐÞ^ìîˆH/Îæ»–Úžþc6ÊoÛ».Ls 3ß»•ƒóP `J6d4Ï£ïÛÛ§½N«ö‚!HPkþwXüÞ·ÏàFÄQϺqÔ=ùtì‰KØo<¥õýヾ>]ÝÝA½þzýñûïÛm{Î(Q ÃÈ+Oïø_Ùp?~hÿÔ?É&áô $ ‚@ú™nERDÛ—*|¸¡ÎÆÿÃíí öÎeªÐ‹ÆÛIìÂ…¾Ú߃†½ýãÛöžs§‹ýØ'~¢-à?³o߶g¼«l.j_D149£ìŽ™¹¼þ¿{ ÉÍfs bѾ ÁxÔF¨nÚšQA4y×ðfûR†LvéÉûÖ·‘AHÀ´”ßÓí‘—·5·}_æÄÚv0åM„máXæöÚØŽ¸¿nÁµ›`·á…"`î|âá„ãxùø}ƒso “±?Œ:ž Ãî CÇ4pûDg7Æ* £o”U Š†^ß‘¡LbIÓÈpäãž[;:›@…Sìé aá1nT.#/Â[½ÈU“\H;`£êyí9n.ê?Ë@$8y> h‡Jº>M°–ѶP€µ@€T!î4”:ª[Út £ÉCE7Ò÷‰œy-fJb&¶dÄ5©?#¶æ¨wDMn°±“„$÷D_s'™’ í‚Ž7°ô5±¥‘9É|”]8JFÅÈ &^œ3¥ Ô$¢'‰@†AÏpài£d.„ºfÔ5F …H¹âR¡B5¾ô=TÄÆ®/Py²ëXÖ—  1‘ÓË$DvìªKœdmWeN—Öv^— ‚G㨩õU£u{}WE1È>ß÷ë]õ™fœ´­Ì¡Ž…M뻡 "WÜà n²žænrI¶¾›ø—Rãåún2|äËÛÜM‚h`[~G\“=͈­9êÝõ~Žú>g{G ÔïŽÜ·ž"ò\î¦hç’Bó±HBÔ™ŒD èÈOï™ÄZö›Ž'n¿ß&PM›Ï¡ :°MÍê: < ¹²„ÈÃŒòbÒ²fGMIå:¡ôÈ{v’ÀACi… ¶7¸­F^îež—³ç†>ù¥µ½±.¦æ^¶Vl † Ò.¶‹Ëù³±ƒëðC\éÞµ EÝ£¥M6öɵ·öE‹|Ìõ}š¿µïùäöšëŒP`û€š¿jîkŒkí«ð1¾ÔoSQ$Éw6÷A¥G÷Ä7û¡Ì‚7™J@…6¶~ÖÐlj`—¸¦æ‰»Ê‰ |G=9ð./òdãÀó7$ó…vß“ÏRñÜ _ÞâLJ $Äj¨Ì­Ø0âB?Ô:ÙØP:y\k„™Ðš[8aFÝÀ¬ÔœÐo$ÒbßjŒ鸨]Ð̮Ѵ»ÀOšµ+4†›»à| Ö3Rs'9½E“LqgÐ{z¡ ‹ÓÈ+Mwº ƒƒ]Wºž]KQÀŒöêÖZ× @#µæ@ý¢"„鮺ÔФ«äÕ•¤±•Eq¨uö’¶Nñ[³ n^릩=tCàØìf>ó!M‘@pMÆÝU„´kÝÄ{¾ÂqÁbãf7Sà!êNQýZ‡AäKĨÃHŒ4 >ŒB»lFÄŽÃ(LÊoû¡1ÝöæžüîÞø˜ñKÇã26>*—šàZC›{rëýîƒHó£ P[Z3UÿH¿èQTëHLâMè| ù¤ŒBØ?’Œ`µ~O¾£¾!î¥uäôóo5q(’ïõÈ ÚõÍ”^‹…¼9|äKçv”³ÿÍQ ²ø`yˆKP€Ò8Tâx† Çb_ø ýXtÉþ¡q<Ñø}œ¹•kŠ£Ž³h,¢Î``ˆŸÐj&¬‡\@Åä®6ŹJB µ›ÇRà _“®Àç4;7Ï­9 #‰cà*™Ë±7íBA¯«Hè@««aH0q¬n(ø9NASfM™ÄjObñ‰ðA×PCîÉ-¢ƒÛ„b¶Ö‰GØnŽþ:Ã4'ŽÜˉÁlìb ‘Ó[ZP‘èP†Zÿ$GŠR²ç@ÈÓtNœ Ý}á]DƒxJê|$Iy€ò‚@Û˜sšøX9jHI A¨\Á›Ù5Í! ҼŎ&°}’ùñLÔßÁ>Ti)îiæ»vGmÍ“ïæÉ÷4Æk'Jh‰äú °RÑ¢›'Jáæ)A›4!tÓҼȓ\¢¦šChŠ-|óTL²À‘òšÆ)›§2Ó@'Aé½>%nŸfÀ˜L4RÇùÓ$Š0]ÿ”Åô±@¬-àS–¨!Æ^.‰µOÑ%ÀT>ÈëO²âSk÷ÜHô‡ÍOŸ‰kg0(ÊCÎ@ñCŠ¨[@Sê×8 FgBËlΤ¯"›dàЈ2Öf#§·¨7xcÈ•>æ“g^V~!—Ü<ƒ]ÁY¦á®Î Cw„ mä UÎ@Ø’ˆ)9â³$(kúïÄÃÀ³' ZÙ¡ÕƒèržƒøÔ¨Eû­ž&¨{k=éêHÑ“Z8` &Íž§ES=ïš²ž„Þ0A³lõ”¤ü¨¡–šÜQ9Óßô”ŠK¡ÅüÕSD.„z*õ(Ú…b—˜d2’ã|OßôÔôRøßmƒ‰Q >DÞ6z ¦;Þ”¼Nošàu_ û" o5r í'#}ž¸¡ø ¨Ï„ÛoúàÊÁJùÐõ¾ãŠ,Ê…›Ñ#uÕð‰wu…³Av¡ÐA4ûyëoú2‹²/Ã,Ao×Ï*eè1Öû2Š<ÊWÖú®Èá·M™®§Z¿¹êcFjš½ñˆª‡}WÝh·&‘VŽë{ã±¾O1€Œ<Ú|³?¢le Ÿ¸¹Ñ÷Eù’oé=ŠþgÄÖõnŽÂÚB?˜´AÀÅ¢a÷U2Є9ÑÜ P¡@Þ7ûà0(ú1ä¸Åu WùðôgAŸ/0؃»*y1ý$€¿8ò ðƒ è43”¼´€&ÑCkiv­É-)깘úèP×î(t…çZY´uîª1%À¥Il#§ó{7áƒ0 @q€æ¹› ‡^BÓ¤Ö¹'\Ìø€‚±”yù&í¶u®Õ®›çê†ÒîóP\A(@áçyj%†Ÿä[3wúÓDS©Ÿ¥/nÿ‡ÙÆÏ^ä 5þìihÖºZ:׊˅Vļ±æ0€Òœ¬QËÀ.4ºÐPº}Ö¸v!½@[†ô©0°vGmi¾Þõ3Q®_y’"¥‹ÎåëKS>º¡ HP©îøê"ÁNSªæŸU‚€Ùš[põ—ìçÿŸxá9´­gnm/data/wheat.rda0000644000176000001440000000562212515664670013615 0ustar ripleyusers‹íilUiÇß.ÀÀ03-”î-·¥Ä á¾g§Ì´d J¥” Ž‰¡B`ZÀ–%ÄmÌD£#&ê#úÅ-î1Œ‰Fã¸M\˜¸ŒÜ>Ì3.•8.Q#ø/ç9Ïs‡`Æ%Mç?¹7yû{Îú?ïö<ï{zν{‡øe–9çj]]-þÖÁ¬‡áj\½[ .:wtrâ´su+±ÐƒtÒçVÓ¹ö5Îuüùf¬j‰kºÓ¹Æ/:wÇŸ[[Æ>ÿt®ws¥ÂÞè\ç óãZ;×ÿZœî«Øïuàì÷ö»öï±ßOœëÂEuüëܪCgνù8w×[°ût5 Á18®ã/ ¶u<†k{#Òë‘þ ½.¬ÿÒýÎucÿ§q­¯Áyp}+·†qp×6žDú.®s7öÿD~MÝý8æÖüÜ]Kpþ÷ã:·a=ôK˱y_aû»°®<žÃúVðs¸†kþ-o@Ú‰ëù¬sk2ûÎó#;wʤûQóm$ì×û¶¿EÃ~/Â2òYú$ö¯Áy¾…óâZ›QßGù4Á¾ yù’sËqLû—q(¯Nä½ã•8ç®ävç‡q=(‹.Tpû/°î HÃÁ¶N¬Ç9ºJr,òÜò&œ×ÓŒëjDù7 ¬>‹ôy쇲èz+öý”ý\9üyy ¶£œ{çòùn,÷æuÔ5—_è–æ¶Íµ‰B÷3yÝvB³éqÔÅ{ëCy¯E÷ü må{8º¥‹àßq®OcûÛ°íW(GÔg í£t éCØŽ\ú5Î÷”ÉÇÀ38'òÕ€2n¸iMøòtõ°Ê¹•ØÕ(h· Ý4£ü[P®+qM¯O(÷¶`=Ú{3Ú{ëqäãÖÍåõÐ…<·âšš§ÀËØí½eÖùGpƒäûçRöèG]6Ú_ÇW°ücäÿ*øu×Ññj$ÔA7®³m°ý¤© e‚òlD[»íz9òׇ¶Û“ ÿo7äípõ7.æí®÷SXÿ3lǵôìqΞ‡ÀËyùÔI?FGŸßT;Ï©nžSý<§EóœÏsZ2Ïé–yNKç9-›ßôŒ`½xjòìäÔ,¬†ë[óµõå,M+ìÌìlS…]®°}…TØa…UØq…Ü8„8451[\T­¬\rræðäÌäáâÊL:}>Ï]­ðMµ>¤Ê*«¤âœ¨½vCǯ9]ûò=êŠ=jžÕÔÞd\Q]®.W—ŸóË7÷³…1ö?ù‚5ÿÝœ£º½º½ºý9³ýæ¾`º0Fÿ+_ðÿÞx¶ùõüÕóWÏ?oç†/¨+\@q[¢æDaìºÁ)ü'÷Üís>b(½8÷y/w^Æç1^>ydîÃËàú‡– ;¥X™¹ëZÆ’Vn–|ÐRü()5gÒŸYÉÛ+ÙãXÚ&ã!RÆ’V²C³"“2} ¬vÄJòö¯qx@ƬdG¤’VHXÉÇÖI>XÙ'~”•ÝÒH©q˜}>Ó'ó2Vö’“½üÙçÃ}E<&%û8¨¿¸¿NÊ5ÒXIÞ5oöDKiO¬\OÎuÒX¹œ/y+×K?fåš"ž‘²Ú~”‡WHfe9›¤?°²CÆw¬\]üŸ‰”äÏ© õJ=ÐRê•·I>ª\j¾]â3+—H{beƒäƒ•-Ò®XY}Þ~aÙ&ý€•äãèÁÇó|°2}ù!¥Æaöù ûó–·È¼€•ÝÅsߤ¬>¯»ÀÏë’ãØŸïh%'ùÏ h™A-÷<‘烕;¤_°rô’-q”ãwÊûФÜóÓ¼³r|¹ä‡”c¨iqXÞçcå¸Ì‹iù°ø#Rî—xÆÊ=³2¯'åKe<ÄÊñ•ÒžH¹÷Ñ<¬<ðfrJ\#¥ÆáÉ÷äùaåA¹?Êʉwäñ€•áæËe<ÁÊCÏXyPƬ|Õ=Ü<,ÿcå!‰Ç¤Ô8|\òØXÞžXy\¾gŽ•Ç$°ò¾¹y\îO³ò¨ÄZž•|ò~ñC¬¯íë?ŒuÐsº_Sï×ÕýŠªU}zàÇë»çbÝsjž›·éÛ‡žc¾yõ|±;°ßYºSu¬suÅ8rùñ æ©„¼vµêRëýª*ݧý¬´_ õé»ÛÍWm³žZh|øýp]Ê5.¼·ªÎ½Gû÷]ûꇪ´wçºqò£5=b‹Elñˆ-±uFl]Û¶ˆ-­mý•ÉgKÙ|ÑO»O›w;ú(£ÐOa€Âq '( R8Ia(Ä%Fón‘4¬!çŽz“?m„NƒÞ²C‰–ÿ}­ñÜߎ ¶GßT;bо¥(0 œ‚IÁ¢ (H 6…4…LSŠmþiï·•KîÅ,•ÍSV!^œð²*w•ÜÂDÖ+·!Œ1×s{rº‘,L^éie†ÿÕ÷ë©^¤CâH&’…$$’ƒÁ`0 ƒÁ`0 ƒÁ`08 ƒÃà08 ƒÃà08 ÄaÂ0a˜0L& † Äa°`X0, †ÂaÁ°`X0 CÀ0 CÀ0 CÂ0$ CÂ0$ CÂ0l6 † ÆaðaØ0l6Œtoó'@Ûø Oÿï gnm/data/friend.rda0000644000176000001440000000253312515664670013752 0ustar ripleyusers‹íYÝnÓH>¤ìj@Zdi}á KkY–eEQ”æ§4-lÓ&-‘M¶©BB¢@WB.€È£±o²°O€ÖЙ1þf¾¸7{G¤vÆΙ33çç;ßLÎfOû·ŸÞ‘–lµÊÿ[e÷fÙ‘rSn•íÏ}xsùþµÈÖòïïÉÑZ¾~&¹\µÝu›ê¶­ÛD·=ÝîÀ{=þà_ý¼­ÛXêóˆn#ÝõÖŒ·zÄ/gßgºU~=γ‘_èýdŸm˜ÏŒë‚½Œ\Hì—Ã÷ èËA~vË`¾ümðê7û‰ßÿ°Nk·ôD¤ È~SX?~Öç3|oõÇðœB\˜ý«`¿(‡þïÂs r…ß/v½°?;è[§Á¾¸oUoç푸1û=‚udÄ_)±¯"Ï诜ةÁßNÞc~%ÄŸ}’_/'õï×Æ=y£÷pÝ9É—ÉŸô þL¾2½wI¼Ëæ|²~U$"¿ŸEüþÆ“ìYÀÞ9ìñùöµü\·CâK`<àŸ Bh Hþ ‰“˜àˆÀ:3’Çè‡!Øëì„øÜ%x4àD^ß'Æ‹?‰l®Ã‘yÍ÷{0þ>ä÷ ß&xßûöIfdþýÍya¿ïÀøÃu]¾r1ÁÕèïÿÍøëlæeˆ·¾³|E\WP眼Iý|Ä>õùz‚y²Mp4hà©àwÄ-Ê_Õf¾‹8éÔ¯ö^‡ùœ}Çk|„DoÜÀG°Ž!ïÙƒuĤ~6ñàkŽÃ>¢ÀŽÁõæÃy˜ßÿGÇ©#Ì{G°î.9·0^4ƒxè“xÇu§¤î!^;ˆ#b?å—§uEHÞáú‡›ù,;2?Óó‰"vŒÿðC 8’óß6á¥cØï>Œ´öÛñ&‡y‰“éÚŽÜ#xÐ!õp@p(!¸$~Þåäð’ÉŸšÿ݃uª¾\=ìßÀ^ó¾€áõ‰—‚œûÂóÉקxïµ y€ýï Io7Üoæ$ÿ2r"ã¤ÙøU„wɾÆ伞“:læËN]6Ç_MxÇîÛ¦Äßì<”ÜJHýgvU~»:÷ ±žûBR/ÉyÜêûÔ…„Ü0¾JökíëMÉùEüüʉƃٹJ‘ásxï8€º4½‰ççë:~!ÁUEêmLüÔ%öP$¯B†`‡±ÿ¾ŒÞg Øáœ;äþ>%÷qi_Ò²{~vNÀ{Ÿ‚C7œõÞþ‡ñüW~¿9yž:_3.r8„8’uã=ö©ù½øÚ‹üõ8'~ùý)Œ‡KþÄ„wîB3Ü Éï{ø~¶öÇÏÔ¬¼ö+ñÖë7ïÊæùúC²ÈoW5UŠ¼¿xwù±ùUËß䮾o­æFÙ|¹2ÝÅ|aº懦»œYÙóå±é®ŽÏ®Ó¥yy¸|XuO«î‘í.¬Ú?¦SÓ=9³ÝÅô‰a6­&«ºóïºóªkVÞ² Z§VÕ´Òº\-«®}{´:³–©D+¥çÖ0gOŒ¹Zÿ°êÿcÕZ,ÿ¤ù›©[úå¦óªü÷‡¼z{ñÑ ±/?]¼|{y%-_þ¥é¥!gnm/data/voting.rda0000644000176000001440000000062112515664670014005 0ustar ripleyusers‹­“QJÃ@†·I*4¢>õ""¥Ù6mEЈ‚úZ Š°¤Ûˆ‰¦¡¾öÅûx„É´n»3C³­‚h`ú3»;ÿI»W=îö\ƘÅlKýÚ ¬ÄVQº5Nó(2fï/–T쪨lù-PôÔ7ò6+—ßNð*zžìS×»°~ Z‡s÷ßMµž±¢?*ÖŸŒ>]8wë÷ŒùPÍ}¯ZCè÷8)ú¼Oa{g1Ïò ®‡µ!ì álˆòz¿”XŽe‚‰ûÆ[ç;´óU¶Þ¾3gΙ3ósf éJl¶2«” ('àÿ;¾ô5¥‚jÆÇéfÏ{Óïv”r®øÃߺÿV|º1XUKÀu`fhð½6X\À|UÛþ»oÃ. ÿ{˜ÿ€¸¯ ¿ }öyèKÚŽŸ‚Ý}à o°eëô«À80,“ÀÛ²_àðùG)nûyLûw1. íý\ÎCGÆ•‘­¿œnaÝ=Œ÷Øã]Ì¿~ÿbðkëÃ.«ìýïç<ôÛ‚Ð3„êÏû°7~k¿”…:<ÓÄçð)pqž mûù'©.ãGò+çÖ[ÊΫíz—|­(ÛOxYnB_„YÛõ› üqS„I:OŒü"Ê®³;Êî§2â>¡uWÉu&ߘ7EûûôÓ®Ãñ"çÛ¾Ùy)ëàéÐ4šnWíÚþ âAxÕðÓÔÏK”×ò[¤¸rÏm í¸‹”¿yŠ3Gý,}ò€îFEu¢ú¦¸ãzQú;BõÊu¼lïwÒwî)Óyù¾ŒN¸Ï%/kò>aÜÖl;©Z7N<Õä¾¢w.óožä<çÎw—ÎC<Ò³í4¼¶—ODÿ7ƒŸCf!߬㶛}ÿé¿ Ûð_[3?•ƒȉääΈ%÷T¬ŠuˆÁÒ­Rn,çrî¹öÿC9!S– YŠÄÌÐb9^ìö¼–×ÁèR£ÙxwàuE5]ïv½¾xÌ Öܾ,8VÜÚAÓX«ã? ”ÂŽ gnm/data/mentalHealth.rda0000644000176000001440000000054112515664670015106 0ustar ripleyusers‹­‘OOÂ0À»v ,š˜>€ODÊ`<ˆ‚rÁƒ\8™4¬Ä%ÝfÆ”OáWß ïz3ñð[¯ëúþìm4—Á<`Œq&8<¨Â<泬g©ÎJe&Z™òƒ1q {Õ¡s  <ïÀÐ6Ž€'À¸&À0µç®eÏ¿WÀ‹a˜ÇsàÂÁw¨9Ô9ê¹nô·6+°‹ÝÛý®7DyDyB¡ŒQž;k £Vx%·›'yëBǘw©e^€mœÎùAWÿWUò­-Ì·•økm zš¬ð4Í¡`UjŒ“ôS%U»{Äö 'S©ÆñšYþ••6³ñ u:™9ß7Š|Ý:¼£‰?¢"Q:(!J%Bé¡ôQVxû–¬M&É:d!Y—,"ë‘õÉ(‡¤’rHÊ!)‡ Áã<‚X•ªµ,`"ûi³Í/Twžcìgnm/data/cautres.rda0000644000176000001440000000074212515664670014151 0ustar ripleyusers‹•“ÉNA†f@Á]q‰¢¨c\bXÄ].уkÔ  ਔ#ŸBŸÀ÷ðäÁGðyŒ9 ”䣺§«ªÿúaÖR1ʯ”r+ÃoK åR¦ò!ÖeÓvÑ*)e4aû"À:$ 0vz0|Œ^ Œs\Œr>Z¥·Èia¿v‘À~=¼kÌ€Qp­œù=‚Ó_GŸ¼G4Á1Ÿ'X;Åš=j¼!Þ!÷Ï™=\1·è™¤¦S0M’`ôƒæ¬r_&&÷=!nWxâ#m¼_<ÙÝ`“ž\‚%úÙÅ8H@'Ø¥öý¦þ>ö Ðã1žKþýíQÖ,²gœwfÀ x†nÙ§—é—üœ'™6@#ØçÙ!ˆUüo<üÝ%g…³oÑ÷÷aΩÿwƒœgžþŸð·ç AÖg8s‚¹Eð^Á;°=Ô,Þ4Sç,}:ô¡œ×¼ÆEî–rÌ +L‡ªÔz¤Þ¥oåUrëW§õ»Æ9wEõ"Æï/'\Äõ"¡Óz‘dµù§êåVÍãá0U ÌrÁ¶tB6Ÿ.•ôôE+Ÿ»ÊîôÞÊ[Y[öJ}Övv Î?ít&o9Ùêó ¡wI@¹gnm/R/0000755000176000001440000000000012515534061011266 5ustar ripleyusersgnm/R/dummy.coef.gnm.R0000644000176000001440000000144212515147523014243 0ustar ripleyusers# Copyright (C) 2005 Heather Turner # # This program is free software; you can redistribute it and/or modify # it under the terms of the GNU General Public License as published by # the Free Software Foundation; either version 2 or 3 of the License # (at your option). # # This program is distributed in the hope that it will be useful, # but WITHOUT ANY WARRANTY; without even the implied warranty of # MERCHANTABILITY or FITNESS FOR A PARTICULAR PURPOSE. See the # GNU General Public License for more details. # # A copy of the GNU General Public License is available at # http://www.r-project.org/Licenses/ dummy.coef.gnm <- function (object, ...) { if (inherits(object, "gnm", TRUE) == 1) stop("dummy.coef is not implemented for gnm objects") else NextMethod } gnm/R/Logistic.R0000644000176000001440000000206112515147523013170 0ustar ripleyusers# Copyright (C) 2007 Heather Turner # # This program is free software; you can redistribute it and/or modify # it under the terms of the GNU General Public License as published by # the Free Software Foundation; either version 2 or 3 of the License # (at your option). # # This program is distributed in the hope that it will be useful, # but WITHOUT ANY WARRANTY; without even the implied warranty of # MERCHANTABILITY or FITNESS FOR A PARTICULAR PURPOSE. See the # GNU General Public License for more details. # # A copy of the GNU General Public License is available at # http://www.r-project.org/Licenses/ Logistic <- function(x, inst = NULL){ list(predictors = list(Asym = 1, xmid = 1, scal = 1), variables = list(substitute(x)), term = function(predLabels, varLabels) { paste(predLabels[1], "/(1 + exp((", predLabels[2], "-", varLabels[1], ")/", predLabels[3], "))") }, start = function(theta){ c(NA, mean(x), sd(x)) } ) } class(Logistic) <- "nonlin" gnm/R/effects.gnm.R0000644000176000001440000000143412515147523013615 0ustar ripleyusers# Copyright (C) 2005 Heather Turner # # This program is free software; you can redistribute it and/or modify # it under the terms of the GNU General Public License as published by # the Free Software Foundation; either version 2 or 3 of the License # (at your option). # # This program is distributed in the hope that it will be useful, # but WITHOUT ANY WARRANTY; without even the implied warranty of # MERCHANTABILITY or FITNESS FOR A PARTICULAR PURPOSE. See the # GNU General Public License for more details. # # A copy of the GNU General Public License is available at # http://www.r-project.org/Licenses/ effects.gnm <- function (object, ...) { if (inherits(object, "gnm", TRUE) == 1) stop("effects is not implemented for gnm objects") else NextMethod } gnm/R/gnm.R0000755000176000001440000003464612515147523012215 0ustar ripleyusers# Designed to take similar arguments to glm from the stats package from R; # some of the code to handle the arguments is copied/modified from glm. # # Copyright (C) 1995-2005 The R Core Team # Copyright (C) 2005-2010, 2012, 2013 Heather Turner and David Firth # # This program is free software; you can redistribute it and/or modify # it under the terms of the GNU General Public License as published by # the Free Software Foundation; either version 2 or 3 of the License # (at your option). # # This program is distributed in the hope that it will be useful, # but WITHOUT ANY WARRANTY; without even the implied warranty of # MERCHANTABILITY or FITNESS FOR A PARTICULAR PURPOSE. See the # GNU General Public License for more details. # # A copy of the GNU General Public License is available at # http://www.r-project.org/Licenses/ gnm <- function(formula, eliminate = NULL, ofInterest = NULL, constrain = numeric(0), #index of non-eliminated parameters constrainTo = numeric(length(constrain)), family = gaussian, data = NULL, subset, weights, na.action, method = "gnmFit", checkLinear = TRUE, offset, start = NULL, etastart = NULL, mustart = NULL, tolerance = 1e-6, iterStart = 2, iterMax = 500, trace = FALSE, verbose = TRUE, model = TRUE, x = TRUE, termPredictors = FALSE, ridge = 1e-8, ...) { call <- match.call() modelTerms <- gnmTerms(formula, substitute(eliminate), data) modelData <- as.list(match.call(expand.dots = FALSE)) if (inherits(data, "table") && missing(na.action)) modelData$na.action <- "na.exclude" argPos <- match(c("eliminate", "data", "subset", "weights", "na.action", "offset", "etastart", "mustart"), names(modelData), 0) modelData <- as.call(c(as.name("model.frame"), formula = modelTerms, modelData[argPos], drop.unused.levels = TRUE)) modelData <- eval(modelData, parent.frame()) if (!missing(eliminate)) { eliminate <- modelData$`(eliminate)` if (!is.factor(eliminate)) stop("'eliminate' must be a factor") xtf <- xtfrm(modelData$`(eliminate)`) ord <- order(xtf) if (ordTRUE <- !identical(ord, xtf)) { modelData <- modelData[ord, , drop = FALSE] eliminate <- modelData$`(eliminate)` } nElim <- nlevels(eliminate) } else nElim <- 0 if (method == "model.frame") return(modelData) else if (!method %in% c("gnmFit", "coefNames", "model.matrix") && !is.function(get(method))) { warning("function ", method, " can not be found. Using \"gnmFit\".\n", call. = FALSE) method <- "gnmFit" } nobs <- nrow(modelData) y <- model.response(modelData, "numeric") if (is.null(y)) y <- rep(0, nobs) weights <- as.vector(model.weights(modelData)) if (!is.null(weights) && any(weights < 0)) stop("negative weights are not allowed") if (is.null(weights)) weights <- rep.int(1, nobs) offset <- as.vector(model.offset(modelData)) if (is.null(offset)) offset <- rep.int(0, nobs) mustart <- model.extract(modelData, "mustart") etastart <- model.extract(modelData, "etastart") if (is.character(family)) family <- get(family, mode = "function", envir = parent.frame()) if (is.function(family)) family <- family() if (is.null(family$family)) { print(family) stop("`family' not recognized") } if (family$family == "binomial") { if (is.factor(y) && NCOL(y) == 1) y <- y != levels(y)[1] else if (NCOL(y) == 2) { n <- y[, 1] + y[, 2] y <- ifelse(n == 0, 0, y[, 1]/n) weights <- weights * n } } if (is.empty.model(modelTerms) && missing(eliminate)) { if (method == "coefNames") return(numeric(0)) else if (method == "model.matrix") return(model.matrix(modelTerms, data = modelData)) if (!family$valideta(offset)) stop("invalid predictor values in empty model") mu <- family$linkinv(offset) if (!family$validmu(mu)) stop("invalid fitted values in empty model") dmu <- family$mu.eta(offset) dev <- sum(family$dev.resids(y, mu, weights)) modelAIC <- suppressWarnings(family$aic(y, rep.int(1, nobs), mu, weights, dev)) fit <- list(coefficients = numeric(0), constrain = numeric(0), constrainTo = numeric(0), eliminate = NULL, predictors = offset, fitted.values = mu, deviance = dev, aic = modelAIC, iter = 0, weights = weights*dmu^2/family$variance(mu), residuals = (y - mu)/dmu, df.residual = nobs, rank = 0, family = family, prior.weights = weights, y = y, converged = NA) if (x) fit <- c(fit, x = model.matrix(modelTerms, data = modelData)) if (termPredictors) fit <- c(fit, termPredictors = matrix(, nrow(modelData), 0)) } else { onlyLin <- checkLinear && all(attr(modelTerms, "type") == "Linear") if (onlyLin) { if (nElim) { X <- model.matrix(update(modelTerms, . ~ . + 1), modelData) asgn <- attr(X, "assign") X <- X[,-1, drop = FALSE] attr(X, "assign") <- asgn[-1] } else X <- model.matrix(modelTerms, modelData) coefNames <- colnames(X) } else { modelTools <- gnmTools(modelTerms, modelData, method == "model.matrix" | x) coefNames <- names(modelTools$start) } if (method == "coefNames") return(coefNames) nParam <- length(coefNames) if (identical(constrain, "[?]")) call$constrain <- constrain <- unlist(pickFrom(coefNames, edit.setlabels = FALSE, title = "Constrain one or more gnm coefficients", items.label = "Model coefficients:", warningText = "No parameters were specified to constrain", return.indices = TRUE)) if (is.character(constrain)) { res <- match(constrain, coefNames, 0) if (res == 0 && length(constrain) == 1){ constrain <- match(grep(constrain, coefNames), seq_len(nParam), 0) } else constrain <- res } ## dropped logical option if (!all(constrain %in% seq_len(nParam))) stop(" cannot match 'constrain' to non-eliminated parameters. ") if (is.null(start)) start <- rep.int(NA, nElim + nParam) else if (length(start) != nElim + nParam) { if (!missing(eliminate) && length(start) == nParam) start <- c(rep.int(NA, nElim), start) else stop("length(start) must either equal the no. of parameters\n", "or the no. of non-eliminated parameters.") } if (onlyLin) { if (length(constrain)) { offset <- drop(offset + X[, constrain, drop = FALSE] %*% constrainTo) X[, constrain] <- 0 } if (method == "model.matrix") return(X) } else if (method == "model.matrix"){ theta <- modelTools$start theta[!is.na(start)] <- start[!is.na(start)] theta[constrain] <- constrainTo theta[is.na(theta)] <- seq(start)[is.na(theta)] varPredictors <- modelTools$varPredictors(theta) X <- modelTools$localDesignFunction(theta, varPredictors) attr(X, "assign") <- modelTools$termAssign return(X) } if (!is.numeric(tolerance) || tolerance <= 0) stop("value of 'tolerance' must be > 0") if (!is.numeric(iterMax) || iterMax < 0) stop("maximum number of iterations must be >= 0") if (onlyLin) { if (any(is.na(start))) start <- NULL fit <- glm.fit.e(X, y, weights = weights, start = start, etastart = etastart, mustart = mustart, offset = offset, family = family, control = glm.control(tolerance, iterMax, trace), intercept = attr(modelTerms, "intercept"), eliminate = eliminate) if (sum(is.na(coef(fit))) > length(constrain)) { extra <- setdiff(which(is.na(coef(fit))), constrain) ind <- order(c(constrain, extra)) constrain <- c(constrain, extra)[ind] constrainTo <- c(constrainTo, numeric(length(extra)))[ind] } if (!is.null(fit$null.deviance)) { extra <- match(c("effects", "R", "qr", "null.deviance", "df.null", "boundary"), names(fit)) fit <- fit[-extra] } names(fit)[match("linear.predictors", names(fit))] <- "predictors" fit$constrain <- constrain fit$constrainTo <- constrainTo if (x) { fit$x <- X } if (termPredictors) { modelTools <- gnmTools(modelTerms, modelData) varPredictors <- modelTools$varPredictors(naToZero(coef(fit))) fit$termPredictors <- modelTools$predictor(varPredictors, term = TRUE) } } else if (method != "gnmFit") fit <- do.call(method, list(modelTools = modelTools, y = y, constrain = constrain, constrainTo = constrainTo, eliminate = eliminate, family = family, weights = weights, offset = offset, nobs = nobs, start = start, etastart = etastart, mustart = mustart, tolerance = tolerance, iterStart = iterStart, iterMax = iterMax, trace = trace, verbose = verbose, x = x, termPredictors = termPredictors, ridge = ridge, ...)) else fit <- gnmFit(modelTools = modelTools, y = y, constrain = constrain, constrainTo = constrainTo, eliminate = eliminate, family = family, weights = weights, offset = offset, nobs = nobs, start = start, etastart = etastart, mustart = mustart, tolerance = tolerance, iterStart = iterStart, iterMax = iterMax, trace = trace, verbose = verbose, x = x, termPredictors = termPredictors, ridge = ridge) } if (is.null(fit)) { warning("Algorithm failed - no model could be estimated", call. = FALSE) return() } if (is.null(ofInterest) && !missing(eliminate)) ofInterest <- seq_len(nParam) if (identical(ofInterest, "[?]")) call$ofInterest <- ofInterest <- pickCoef(fit, warningText = paste("No subset of coefficients selected", "- assuming all are of interest.")) if (is.character(ofInterest)) { if (length(ofInterest) == 1) ofInterest <- match(grep(ofInterest, coefNames), seq_len(nParam), 0) else ofInterest <- match(ofInterest, coefNames, 0) if (!sum(ofInterest)) ofInterest <- seq_len(nParam) } if (!is.null(ofInterest)) { if (!all(ofInterest %in% seq_len(nParam))) stop("'ofInterest' does not specify a subset of the ", "non.eliminated coefficients.") names(ofInterest) <- coefNames[ofInterest] } if (missing(data)) data <- environment(formula) fit <- c(list(call = call, formula = formula, terms = modelTerms, data = data, eliminate = eliminate, ofInterest = ofInterest, na.action = attr(modelData, "na.action"), xlevels = .getXlevels(modelTerms, modelData), offset = offset, tolerance = tolerance, iterStart = iterStart, iterMax = iterMax), fit) if (!missing(eliminate) && ordTRUE) { reorder <- order(ord) fit <- within(fit, { y <- y[reorder] fitted.values <- fitted.values[reorder] predictors <- predictors[reorder] residuals <- residuals[reorder] weights <- weights[reorder] prior.weights <- prior.weights[reorder] eliminate <- eliminate[reorder] offset <- offset[reorder] }) modelData <- modelData[reorder, , drop = FALSE] y <- y[reorder] if (x) { asgn <- attr(fit$x, "assign") fit$x <- fit$x[reorder, , drop = FALSE] attr(fit$x, "assign") <- asgn } } asY <- c("predictors", "fitted.values", "residuals", "prior.weights", "weights", "y", "offset") if (inherits(data, "table") && (is.null(fit$na.action) | inherits(fit$na.action, "exclude"))) { attr <- attributes(data) if (!missing(subset)) { ind <- as.numeric(names(y)) lev <- do.call("expand.grid", attr$dimnames)[ind,, drop = FALSE] attr$dimnames <- apply(lev, 2, unique) attr$dim <- unname(sapply(attr$dimnames, length)) } fit$table.attr <- attr } fit[asY] <- lapply(fit[asY], structure, dim = NULL, names = names(y)) if (termPredictors) rownames(fit$termPredictors) <- names(y) if (model) fit$model <- modelData class(fit) <- c("gnm", "glm", "lm") attr(fit, ".Environment") <- environment(gnm) fit } gnm/R/print.summary.gnm.R0000644000176000001440000000720012515257256015030 0ustar ripleyusers# Modification of print.summary.glm from the stats package for R. # # Copyright (C) 1995-2006 The R Core Team # Copyright (C) 2006, 2008, 2009, 2015 Heather Turner # # This program is free software; you can redistribute it and/or modify # it under the terms of the GNU General Public License as published by # the Free Software Foundation; either version 2 or 3 of the License # (at your option). # # This program is distributed in the hope that it will be useful, # but WITHOUT ANY WARRANTY; without even the implied warranty of # MERCHANTABILITY or FITNESS FOR A PARTICULAR PURPOSE. See the # GNU General Public License for more details. # # A copy of the GNU General Public License is available at # http://www.r-project.org/Licenses/ print.summary.gnm <- function (x, digits = max(3, getOption("digits") - 3), signif.stars = getOption("show.signif.stars"), symbolic.cor = x$symbolic.cor, ...) { cat("\nCall:\n", deparse(x$call), "\n", sep = "", fill = TRUE) cat("Deviance Residuals: \n") if (length(x$deviance.resid) > 5) { x$deviance.resid <- quantile(x$deviance.resid, na.rm = TRUE) names(x$deviance.resid) <- c("Min", "1Q", "Median", "3Q", "Max") } print.default(x$deviance.resid, digits = digits, na.print = "", print.gap = 2) tidy.zeros <- function(vec) ifelse(abs(vec) < 100 * .Machine$double.eps, 0, vec) coefs <- tidy.zeros(coef(x)) if (length(ofInterest(x)) > 0) coefs <- coefs[ofInterest(x), , drop = FALSE] non.elim <- length(coefs) elim <- length(x$eliminated) if (non.elim | elim) { cat("\nCoefficients", " of interest"[!is.null(ofInterest(x))], ":\n", sep = "") printCoefmat(coefs, digits = digits, signif.stars = signif.stars, signif.legend = !elim, na.print = "NA", ...) if (elim){ cat("\nEliminated coefficients:\n", sep = "") printCoefmat(x$eliminated, digits = digits, signif.stars = signif.stars, na.print = "NA", ...) } coefs <- c(coefs[,2], x$eliminated[,2]) if (any(!is.na(coefs))) cat("\n(Dispersion parameter for ", x$family$family, " family taken to be ", format(x$dispersion), ")\n", sep = "") if (any(is.na(coefs))) cat("\nStd. Error is NA where coefficient has been constrained or", "is unidentified\n") } else cat("\nNo coefficients", " of interest"[!is.null(ofInterest(x))], ". \n\n", sep = "") cat("\nResidual deviance: ", format(x$deviance, digits = max(5, digits + 1)), " on ", format(x$df.residual, digits = max(5, digits + 1)), " degrees of freedom\n", "AIC: ", format(x$aic, digits = max(4, digits + 1)), "\n\n", "Number of iterations: ", x$iter, "\n", sep = "") correl <- x$correlation if (!is.null(correl)) { if (attr(x$cov.scaled, "eliminate")) { eliminate <- seq(attr(x$cov.scaled, "eliminate")) correl <- correl[-eliminate, -eliminate] } p <- NCOL(correl) if (p > 1) { cat("\nCorrelation of Coefficients:\n") if (is.logical(symbolic.cor) && symbolic.cor) { print(symnum(correl, abbr.colnames = NULL)) } else { correl <- format(round(correl, 2), nsmall = 2, digits = digits) correl[!lower.tri(correl)] <- "" print(correl[-1, -p, drop = FALSE], quote = FALSE) } } } cat("\n") invisible(x) } gnm/R/labels.gnm.R0000755000176000001440000000157312515147523013447 0ustar ripleyusers# Copyright (C) 2005, 2006 Heather Turner # # This program is free software; you can redistribute it and/or modify # it under the terms of the GNU General Public License as published by # the Free Software Foundation; either version 2 or 3 of the License # (at your option). # # This program is distributed in the hope that it will be useful, # but WITHOUT ANY WARRANTY; without even the implied warranty of # MERCHANTABILITY or FITNESS FOR A PARTICULAR PURPOSE. See the # GNU General Public License for more details. # # A copy of the GNU General Public License is available at # http://www.r-project.org/Licenses/ labels.gnm <- function(object, ...) { labels <- attr(terms(object), "term.labels") termAssign <- attr(model.matrix(object), "assign") if (length(object$constrain)) termAssign <- termAssign[-object$constrain] unique(labels[termAssign]) } gnm/R/asGnm.default.R0000644000176000001440000000137312515147523014110 0ustar ripleyusers# Copyright (C) 2006 Heather Turner # # This program is free software; you can redistribute it and/or modify # it under the terms of the GNU General Public License as published by # the Free Software Foundation; either version 2 or 3 of the License # (at your option). # # This program is distributed in the hope that it will be useful, # but WITHOUT ANY WARRANTY; without even the implied warranty of # MERCHANTABILITY or FITNESS FOR A PARTICULAR PURPOSE. See the # GNU General Public License for more details. # # A copy of the GNU General Public License is available at # http://www.r-project.org/Licenses/ asGnm.default <- function (object, ...) { stop("\nCannot coerce objects of class \"", class(object), "\" to class \"gnm\".") } gnm/R/hashSplit.R0000644000176000001440000000226512515147523013360 0ustar ripleyusers# Copyright (C) 2006 Heather Turner # # This program is free software; you can redistribute it and/or modify # it under the terms of the GNU General Public License as published by # the Free Software Foundation; either version 2 or 3 of the License # (at your option). # # This program is distributed in the hope that it will be useful, # but WITHOUT ANY WARRANTY; without even the implied warranty of # MERCHANTABILITY or FITNESS FOR A PARTICULAR PURPOSE. See the # GNU General Public License for more details. # # A copy of the GNU General Public License is available at # http://www.r-project.org/Licenses/ hashSplit <- function(string){ ## An adaptation of some Python code by 'tim' ## http://forum.textdrive.com/viewtopic.php?id=3095 if (!length(string) || !nchar(string)) return(string) s <- strsplit(string, "")[[1]] a <- 0 ans <- vector("list", length(s)) iq <- FALSE for (z in seq(s)) { if (s[z] == "#" & !iq) { ans[z] <- paste(s[a:(z - 1)], collapse = "") a <- z + 1 } else if (s[z] == "\""){ iq <- !iq } } ans[z] <- paste(s[a:z], collapse = "") unlist(ans) } gnm/R/instances.R0000644000176000001440000000176212515147523013411 0ustar ripleyusers# Copyright (C) 2006 Heather Turner # # This program is free software; you can redistribute it and/or modify # it under the terms of the GNU General Public License as published by # the Free Software Foundation; either version 2 or 3 of the License # (at your option). # # This program is distributed in the hope that it will be useful, # but WITHOUT ANY WARRANTY; without even the implied warranty of # MERCHANTABILITY or FITNESS FOR A PARTICULAR PURPOSE. See the # GNU General Public License for more details. # # A copy of the GNU General Public License is available at # http://www.r-project.org/Licenses/ instances <- function(term, instances = 1){ term <- match.call()$term if (!"inst" %in% names(formals(match.fun(term[[1]])))) stop(term[[1]], " has no inst argumnt") termList <- vector(mode = "list", length = instances) for (i in seq(instances)) { termList[[i]] <- term termList[[i]]$inst <- i } paste(unlist(termList), collapse = " + ") } gnm/R/summary.gnm.R0000644000176000001440000001003412515257233013667 0ustar ripleyusers# Modification of summary.glm from the stats package for R. # # Copyright (C) 1995-2005 The R Core Team # Copyright (C) 2005, 2006, 2010, 2015 Heather Turner # # This program is free software; you can redistribute it and/or modify # it under the terms of the GNU General Public License as published by # the Free Software Foundation; either version 2 or 3 of the License # (at your option). # # This program is distributed in the hope that it will be useful, # but WITHOUT ANY WARRANTY; without even the implied warranty of # MERCHANTABILITY or FITNESS FOR A PARTICULAR PURPOSE. See the # GNU General Public License for more details. # # A copy of the GNU General Public License is available at # http://www.r-project.org/Licenses/ summary.gnm <- function (object, dispersion = NULL, correlation = FALSE, symbolic.cor = FALSE, with.eliminate = FALSE, ...) { est.disp <- (!object$family$family %in% c("poisson", "binomial") && is.null(dispersion) && object$df.residual > 0) coefs <- parameters(object) if (with.eliminate) coefs <- c(attr(coef(object), "eliminated"), coefs) if (object$rank > 0) { cov.scaled <- vcov(object, dispersion = dispersion, with.eliminate = with.eliminate) ## non-eliminated par only if (nrow(cov.scaled)) { estimable <- checkEstimable(object, ...) estimable[is.na(estimable)] <- FALSE } if (is.matrix(cov.scaled)) sterr <- sqrt(diag(cov.scaled)) else sterr <- diag(cov.scaled) if (length(sterr)) is.na(sterr[!estimable]) <- TRUE if (with.eliminate){ ## check estimability of eliminated coefficients X <- cbind(1, model.matrix(object)[,!is.na(coef(object))]) estimable2 <- sapply(split(1:nrow(X), object$eliminate), function(i) { quickRank(X[i, , drop = FALSE]) == quickRank(X[i, -1, drop = FALSE]) + 1}) sterr <- c(ifelse(estimable2, sqrt(attr(cov.scaled, "varElim")), NA), sterr) } tvalue <- coefs/sterr dn <- c("Estimate", "Std. Error") if (!est.disp) { pvalue <- 2 * pnorm(-abs(tvalue)) coef.table <- cbind(coefs, sterr, tvalue, pvalue) dimnames(coef.table) <- list(names(coefs), c(dn, "z value", "Pr(>|z|)")) } else if (object$df.residual > 0) { pvalue <- 2 * pt(-abs(tvalue), object$df.residual) coef.table <- cbind(coefs, sterr, tvalue, pvalue) dimnames(coef.table) <- list(names(coefs), c(dn, "t value", "Pr(>|t|)")) } else { coef.table <- cbind(coefs, Inf) dimnames(coef.table) <- list(names(coefs), dn) } } else { coef.table <- matrix(, 0, 4) dimnames(coef.table) <- list(NULL, c("Estimate", "Std. Error", "t value", "Pr(>|t|)")) cov.scaled <- matrix(, 0, 0) } df.f <- nrow(coef.table) non.elim <- seq(object$coef) + nlevels(object$eliminate)*with.eliminate elim <- seq(length.out = nlevels(object$eliminate)*with.eliminate) ans <- c(object[c("call", "ofInterest", "family", "deviance", "aic", "df.residual", "iter")], list(deviance.resid = residuals(object, type = "deviance"), coefficients = coef.table[non.elim, , drop = FALSE], eliminated = coef.table[elim, , drop = FALSE], dispersion = attr(cov.scaled, "dispersion"), df = c(object$rank, object$df.residual, df.f), cov.scaled = as.matrix(cov.scaled))) if (correlation & object$rank > 0) { dd <- sqrt(diag(cov.scaled)) ans$correlation <- cov.scaled/outer(dd, dd) ans$symbolic.cor <- symbolic.cor } class(ans) <- "summary.gnm" ans } gnm/R/gnmStart.R0000644000176000001440000000134212515147523013213 0ustar ripleyusers# Copyright (C) 2006 Heather Turner # # This program is free software; you can redistribute it and/or modify # it under the terms of the GNU General Public License as published by # the Free Software Foundation; either version 2 or 3 of the License # (at your option). # # This program is distributed in the hope that it will be useful, # but WITHOUT ANY WARRANTY; without even the implied warranty of # MERCHANTABILITY or FITNESS FOR A PARTICULAR PURPOSE. See the # GNU General Public License for more details. # # A copy of the GNU General Public License is available at # http://www.r-project.org/Licenses/ gnmStart <- function(n, scale = 0.1) { theta <- runif(n, -1, 1) * scale theta + (-1)^(theta < 0) * scale } gnm/R/sumExpression.R0000644000176000001440000000142212515147523014277 0ustar ripleyusers# Copyright (C) 2006 Heather Turner # # This program is free software; you can redistribute it and/or modify # it under the terms of the GNU General Public License as published by # the Free Software Foundation; either version 2 or 3 of the License # (at your option). # # This program is distributed in the hope that it will be useful, # but WITHOUT ANY WARRANTY; without even the implied warranty of # MERCHANTABILITY or FITNESS FOR A PARTICULAR PURPOSE. See the # GNU General Public License for more details. # # A copy of the GNU General Public License is available at # http://www.r-project.org/Licenses/ sumExpression <- function(exprList) { expr <- exprList[[1]] for (i in seq(exprList)[-1]) { expr <- call("+", expr, exprList[[i]]) } expr } gnm/R/model.matrix.gnm.R0000755000176000001440000000226412515147523014606 0ustar ripleyusers# Copyright (C) 2005, 2006 Heather Turner # # This program is free software; you can redistribute it and/or modify # it under the terms of the GNU General Public License as published by # the Free Software Foundation; either version 2 or 3 of the License # (at your option). # # This program is distributed in the hope that it will be useful, # but WITHOUT ANY WARRANTY; without even the implied warranty of # MERCHANTABILITY or FITNESS FOR A PARTICULAR PURPOSE. See the # GNU General Public License for more details. # # A copy of the GNU General Public License is available at # http://www.r-project.org/Licenses/ model.matrix.gnm <- function(object, coef = NULL, ...) { if (!"x" %in% names(object) || !is.null(coef)) { xcall <- object$call xcall$method <- "model.matrix" xcall$constrain <- object$constrain xcall$constrainTo <- object$constrainTo xcall$data <- model.frame(object) xcall[c("weights", "offset")] <- NULL xcall$verbose <- FALSE if (!is.null(coef)) xcall$start <- coef else xcall$start <- coef(object) eval(xcall) } else object[[match("x", names(object))]] } gnm/R/residuals.gnm.R0000755000176000001440000000163412515147523014176 0ustar ripleyusers# Copyright (C) 2005, 2008, 2013 Heather Turner # # This program is free software; you can redistribute it and/or modify # it under the terms of the GNU General Public License as published by # the Free Software Foundation; either version 2 or 3 of the License # (at your option). # # This program is distributed in the hope that it will be useful, # but WITHOUT ANY WARRANTY; without even the implied warranty of # MERCHANTABILITY or FITNESS FOR A PARTICULAR PURPOSE. See the # GNU General Public License for more details. # # A copy of the GNU General Public License is available at # http://www.r-project.org/Licenses/ residuals.gnm <- function(object, type = "deviance", ...) { if (type == "partial") stop("type = \"partial\" not implemented for gnm objects.") else res <- NextMethod("residuals") if (!is.null(object$table.attr)) attributes(res) <- object$table.attr res } gnm/R/dfbeta.gnm.R0000644000176000001440000000143012515147523013417 0ustar ripleyusers# Copyright (C) 2005 Heather Turner # # This program is free software; you can redistribute it and/or modify # it under the terms of the GNU General Public License as published by # the Free Software Foundation; either version 2 or 3 of the License # (at your option). # # This program is distributed in the hope that it will be useful, # but WITHOUT ANY WARRANTY; without even the implied warranty of # MERCHANTABILITY or FITNESS FOR A PARTICULAR PURPOSE. See the # GNU General Public License for more details. # # A copy of the GNU General Public License is available at # http://www.r-project.org/Licenses/ dfbeta.gnm <- function (model, ...) { if (inherits(model, "gnm", TRUE) == 1) stop("dfbeta is not implemented for gnm objects") else NextMethod } gnm/R/dfbetas.gnm.R0000644000176000001440000000143212515147523013604 0ustar ripleyusers# Copyright (C) 2005 Heather Turner # # This program is free software; you can redistribute it and/or modify # it under the terms of the GNU General Public License as published by # the Free Software Foundation; either version 2 or 3 of the License # (at your option). # # This program is distributed in the hope that it will be useful, # but WITHOUT ANY WARRANTY; without even the implied warranty of # MERCHANTABILITY or FITNESS FOR A PARTICULAR PURPOSE. See the # GNU General Public License for more details. # # A copy of the GNU General Public License is available at # http://www.r-project.org/Licenses/ dfbetas.gnm <- function (model, ...) { if (inherits(model, "gnm", TRUE) == 1) stop("dfbetas is not implemented for gnm objects") else NextMethod } gnm/R/Topo.R0000644000176000001440000000234712515147523012343 0ustar ripleyusers# Copyright (C) 2005 David Firth # # This program is free software; you can redistribute it and/or modify # it under the terms of the GNU General Public License as published by # the Free Software Foundation; either version 2 or 3 of the License # (at your option). # # This program is distributed in the hope that it will be useful, # but WITHOUT ANY WARRANTY; without even the implied warranty of # MERCHANTABILITY or FITNESS FOR A PARTICULAR PURPOSE. See the # GNU General Public License for more details. # # A copy of the GNU General Public License is available at # http://www.r-project.org/Licenses/ "Topo" <- function (..., spec = NULL) { if (is.null(spec)) stop("No spec given") dots <- list(...) factorLengths <- sapply(dots, length) lengthsEqual <- {if (length(factorLengths) == 1) TRUE else sd(factorLengths) == 0} if (!lengthsEqual) stop("Factors have different lengths") specDim <- if (is.vector(spec)) length(spec) else dim(spec) dots <- lapply(dots, as.factor) facMat <- cbind(...) spec.ok <- identical(sapply(dots, nlevels), specDim) if (!spec.ok) stop( "Dimensions of spec do not match the factor arguments") return(as.factor(spec[facMat])) } gnm/R/gnm-defunct.R0000644000176000001440000000225312515147523013625 0ustar ripleyusers# Copyright (C) 2012 Heather Turner # # This program is free software; you can redistribute it and/or modify # it under the terms of the GNU General Public License as published by # the Free Software Foundation; either version 2 or 3 of the License # (at your option). # # This program is distributed in the hope that it will be useful, # but WITHOUT ANY WARRANTY; without even the implied warranty of # MERCHANTABILITY or FITNESS FOR A PARTICULAR PURPOSE. See the # GNU General Public License for more details. # # A copy of the GNU General Public License is available at # http://www.r-project.org/Licenses/ Nonlin <- function(functionCall){ .Defunct(msg = paste("'Nonlin' is defunct.", "\nUse functions of class \"nonlin\" instead.", "\nSee ?nonlin.function for more details.")) } class(Nonlin) <- "nonlin" getModelFrame <- function() { .Defunct(msg = paste("'getModelFrame' is deprecated as it was designed to ", "work with the old plug-in architecture for nonlinear terms.")) } qrSolve <- function(A, b, rank = NULL, ...) { .Defunct(msg = paste("'qrSolve' is deprecated as it is no longer used ", "by gnm.")) } gnm/R/pickCoef.R0000644000176000001440000000323012515147523013135 0ustar ripleyusers# Copyright (C) 2006, 2010, 2012, 2013 Heather Turner # # This program is free software; you can redistribute it and/or modify # it under the terms of the GNU General Public License as published by # the Free Software Foundation; either version 2 or 3 of the License # (at your option). # # This program is distributed in the hope that it will be useful, # but WITHOUT ANY WARRANTY; without even the implied warranty of # MERCHANTABILITY or FITNESS FOR A PARTICULAR PURPOSE. See the # GNU General Public License for more details. # # A copy of the GNU General Public License is available at # http://www.r-project.org/Licenses/ pickCoef <- function(object, pattern = NULL, value = FALSE, ...){ coefs <- names(coef(object)) if (is.null(coefs)) stop("Coefficient names cannot be extracted from 'object'") if (missing(pattern)) { default <- list(setlabels = "Selected coefficients", title = "Select coefficients of interest", items.label = "Model coefficients:", return.indices = TRUE, edit.setlabels = FALSE, warningText = "No subset of coefficients selected") dots <- list(...) dotArgs <- match(names(default), names(dots)) allArgs <- c(list(coefs), dots, default[is.na(dotArgs)]) selection <- unname(unlist(do.call(pickFrom, allArgs))) } else { selection <- grep(pattern, coefs, value = FALSE, ...) } if (!length(selection)) selection <- NULL else if (!value) names(selection) <- coefs[selection] else selection <- parameters(object)[selection] selection } gnm/R/vcov.gnm.R0000755000176000001440000000641612515147523013163 0ustar ripleyusers# Code to estimate dispersion from summary.glm from the stats package for R. # # Copyright (C) 1995-2005 The R Core Team # Copyright (C) 2005, 2006, 2010 Heather Turner # # This program is free software; you can redistribute it and/or modify # it under the terms of the GNU General Public License as published by # the Free Software Foundation; either version 2 or 3 of the License # (at your option). # # This program is distributed in the hope that it will be useful, # but WITHOUT ANY WARRANTY; without even the implied warranty of # MERCHANTABILITY or FITNESS FOR A PARTICULAR PURPOSE. See the # GNU General Public License for more details. # # A copy of the GNU General Public License is available at # http://www.r-project.org/Licenses/ ## returns vcov for the non-eliminated parameters vcov.gnm <- function(object, dispersion = NULL, with.eliminate = FALSE, ...){ if (is.null(dispersion)) { if (any(object$family$family == c("poisson", "binomial"))) dispersion <- 1 else if (object$df.residual > 0) { if (any(object$weights == 0)) warning("observations with zero weight ", "not used for calculating dispersion") dispersion <- sum(object$weights * object$residuals^2)/ object$df.residual } else dispersion <- Inf } constrain <- object$constrain eliminate <- object$eliminate nelim <- nlevels(eliminate) w <- as.vector(object$weights) X <- model.matrix(object) ind <- !(seq_len(ncol(X)) %in% constrain) cov.unscaled <- array(0, dim = rep(ncol(X), 2), dimnames = list(colnames(X), colnames(X))) if (!length(ind)) { if (nelim && with.eliminate) { Ti <- 1/sapply(split(w, eliminate), sum) attr(cov.unscaled, "varElim") <- dispersion * Ti } return(structure(cov.unscaled, dispersion = dispersion, ofInterest = NULL, class = "vcov.gnm")) } if (length(constrain)) X <- X[, -constrain, drop = FALSE] W.X <- sqrt(w) * X if (object$rank == ncol(W.X)) { cov.unscaled[ind, ind] <- chol2inv(chol(crossprod(W.X))) } else { if (is.null(eliminate)) { cov.unscaled[ind, ind] <- MPinv(crossprod(W.X), method = "chol", rank = object$rank) } else { ## try without ridge and generalized inverse of Q Ti <- 1/sapply(split(w, eliminate), sum) U <- rowsum(sqrt(w) * W.X, eliminate) W <- crossprod(W.X) Ti.U <- Ti * U UTU <- crossprod(U, Ti.U) cov.unscaled[ind, ind] <- MPinv(W - UTU, method = "chol", rank = object$rank - nelim) if (with.eliminate) { rownames(Ti.U) <- names(attr(coef(object), "eliminated")) attr(cov.unscaled, "covElim") <- dispersion * -Ti.U %*% cov.unscaled[ind, ind] attr(cov.unscaled, "varElim") <- dispersion * -rowSums(attr(cov.unscaled, "covElim") * Ti.U) + Ti } } } structure(dispersion * cov.unscaled, dispersion = dispersion, ofInterest = ofInterest(object), class = "vcov.gnm") } gnm/R/getContrasts.R0000644000176000001440000001267412515147523014106 0ustar ripleyusers# Copyright (C) 2005, 2006, 2008, 2010, 2013 David Firth and Heather Turner # # This program is free software; you can redistribute it and/or modify # it under the terms of the GNU General Public License as published by # the Free Software Foundation; either version 2 or 3 of the License # (at your option). # # This program is distributed in the hope that it will be useful, # but WITHOUT ANY WARRANTY; without even the implied warranty of # MERCHANTABILITY or FITNESS FOR A PARTICULAR PURPOSE. See the # GNU General Public License for more details. # # A copy of the GNU General Public License is available at # http://www.r-project.org/Licenses/ getContrasts <- function(model, set = NULL, ref = "first", scaleRef = "mean", scaleWeights = NULL, dispersion = NULL, check = TRUE, ...){ coefs <- parameters(model) l <- length(coefs) if (!l) stop("Model has no non-eliminated parameters") of.interest <- ofInterest(model) if (!length(of.interest)) of.interest <- seq(l) coefNames <- names(coefs) if (is.null(set)) set <- unlist(pickFrom(coefNames[of.interest], 1, ...)) setLength <- length(set) if (setLength == 0) stop( "No non-empty parameter set specified") if (setLength < 1.5) stop( "For contrasts, at least 2 parameters are needed in a set") if (is.numeric(set)) set <- coefNames[set] for (refName in c("ref", "scaleRef"[!is.null(scaleWeights)])) { refSpec <- c(get(refName)) if (is.numeric(refSpec)){ assign(refName, refSpec) if (length(refSpec) == 1){ if (refSpec %in% seq(setLength)) { temp <- rep(0, setLength) temp[refSpec] <- 1 assign(refName, temp) } else stop("The specified ", refName, " is out of range") } if (length(refSpec) != setLength) stop("The specified ", refName, " has the wrong length") if ((sum(refSpec) - 1) ^ 2 > 1e-10) stop("The ", refName, " weights do not sum to 1") } else assign(refName, switch(refSpec, "first" = c(1, rep.int(0, setLength - 1)), "last" = c(rep.int(0, setLength - 1), 1), "mean"= rep.int(1/setLength, setLength), stop("Specified ", refName, " is not an opton."))) } setCoefs <- coefs[coefNames %in% set] contr <- setCoefs - ref %*% setCoefs grad <- diag(rep(1, setLength)) grad <- grad - ref rownames(grad) <- set if (!is.null(scaleWeights)) { if (is.numeric(scaleWeights)) { scaleWeights <- c(scaleWeights) if (length(scaleWeights) != setLength) stop("The specified scaleWeights has the wrong length") } else scaleWeights <- switch(scaleWeights, unit = rep.int(1, setLength), setLength = rep.int(1/setLength, setLength), stop("Specified scaleWeights is not an opton.")) d <- setCoefs - scaleRef %*% setCoefs vd <- scaleWeights * d vdd <- sqrt(drop(vd %*% d)) contr <- contr/vdd grad <- ((scaleRef * sum(vd) - vd) %o% contr/vdd + grad)/vdd } combMatrix <- matrix(0, l, setLength) combMatrix[match(set, coefNames), ] <- grad colnames(combMatrix) <- set Vcov <- vcov(model, dispersion = dispersion) if (!is.logical(check) && !(all(check %in% seq(setLength)))) { stop("check must be TRUE or FALSE or a suitable numeric index vector") } iden <- rep(TRUE, ncol(combMatrix)) ## all unchecked as yet names(iden) <- colnames(combMatrix) if (is.logical(check)) { if (check) iden <- checkEstimable(model, combMatrix) } else iden[check] <- checkEstimable(model, combMatrix[, check]) if (any(!na.omit(iden))) { if (all(!na.omit(iden))) { warning("None of the specified contrasts is estimable", call. = FALSE) return(NULL) } cat("Note: the following contrasts are unestimable:\n") print(names(iden)[iden %in% FALSE]) } not.unestimable <- iden | is.na(iden) combMatrix <- combMatrix[, not.unestimable, drop = FALSE] V <- crossprod(combMatrix, crossprod(Vcov, combMatrix)) result <- data.frame(contr[not.unestimable], sqrt(diag(V))) dimnames(result) <- list(set[not.unestimable], c("Estimate", "Std. Error")) relerrs <- NULL if (sum(not.unestimable) > 2 && is.null(scaleWeights)) { estimable.names <- names(not.unestimable)[not.unestimable] Vcov <- Vcov[estimable.names, estimable.names, drop = FALSE] QVs <- try(qvcalc(Vcov), silent = TRUE) if (inherits(QVs, "try-error")) message("Quasi-variances could not be computed") else { quasiSE <- sqrt(QVs$qvframe$quasiVar) result <- cbind(result, quasiSE) names(result)[1:2] <- c("estimate", "SE") result$quasiVar <- QVs$qvframe$quasiVar relerrs <- QVs$relerrs } } return(structure(list(covmat = Vcov, qvframe = result, relerrs = relerrs, modelcall = model$call), class = "qv") ) } gnm/R/plot.gnm.R0000755000176000001440000002522512515147523013163 0ustar ripleyusers# Modification of plot.lm from the stats package for R. # # Copyright (C) 1995-2005 The R Core Team # Copyright (C) 2005, 2006, 2008 Heather Turner # # This program is free software; you can redistribute it and/or modify # it under the terms of the GNU General Public License as published by # the Free Software Foundation; either version 2 or 3 of the License # (at your option). # # This program is distributed in the hope that it will be useful, # but WITHOUT ANY WARRANTY; without even the implied warranty of # MERCHANTABILITY or FITNESS FOR A PARTICULAR PURPOSE. See the # GNU General Public License for more details. # # A copy of the GNU General Public License is available at # http://www.r-project.org/Licenses/ plot.gnm <- function (x, which = c(1:3, 5), caption = c("Residuals vs Fitted", "Normal Q-Q", "Scale-Location", "Cook's distance", "Residuals vs Leverage"), panel = if (add.smooth) panel.smooth else points, sub.caption = NULL, main = "", ask = prod(par("mfcol")) < length(which) && dev.interactive(), ..., id.n = 3, labels.id = names(residuals(x)), cex.id = 0.75, qqline = TRUE, cook.levels = c(0.5, 1.0), add.smooth = getOption("add.smooth"), label.pos = c(4, 2), cex.caption = 1) { if (!is.numeric(which) || any(which < 1) || any(which > 5)) stop("'which' must be in 1:5") show <- rep(FALSE, 5) show[which] <- TRUE r <- residuals(x) yh <- predict(x) # != fitted() for glm w <- weights(x) if (!is.null(w)) { # drop obs with zero wt: PR#6640 wind <- w != 0 r <- r[wind] yh <- yh[wind] w <- w[wind] labels.id <- labels.id[wind] } n <- length(r) if (any(show[2:5])) { s <- sqrt(deviance(x)/df.residual(x)) hii <- c(hatvalues(x)) if (any(show[4:5])) { cook <- c(cooks.distance(x)) } } if (any(show[2:3])) { ylab23 <- "Std. deviance resid." r.w <- if (is.null(w)) r else sqrt(w) * r } if (show[5]) { ylab5 <- "Std. Pearson resid." r.w <- residuals(x, "pearson") if(!is.null(w)) r.w <- r.w[wind] # drop 0-weight cases r.hat <- range(hii, na.rm = TRUE) # though should never have NA isConst.hat <- all(r.hat == 0) || diff(r.hat) < 1e-10 * mean(hii) } dropInf <- function(x) { if(any(isInf <- is.infinite(x))) { warning("Not plotting observations with leverage one:\n ", paste(which(isInf), collapse=", ")) x[isInf] <- NaN } x } if (any(show[c(2:3,5)])) rs <- dropInf( r.w/(s * sqrt(1 - hii)) ) if (any(show[c(1, 3)])) l.fit <- "Predicted values" if (is.null(id.n)) id.n <- 0 else { id.n <- as.integer(id.n) if (id.n < 0 || id.n > n) stop(gettextf("'id.n' must be in {1,..,%d}", n), domain = NA) } if (id.n > 0) { ## label the largest residuals if (is.null(labels.id)) labels.id <- paste(1:n) iid <- 1:id.n show.r <- sort.list(abs(r), decreasing = TRUE)[iid] if (any(show[2:3])) show.rs <- sort.list(abs(rs), decreasing = TRUE)[iid] text.id <- function(x, y, ind, adj.x = TRUE) { labpos <- if (adj.x) label.pos[1 + as.numeric(x > mean(range(x)))] else 3 text(x, y, labels.id[ind], cex = cex.id, xpd = TRUE, pos = labpos, offset = 0.25) } } getCaption <- function(k) # allow caption = "" , plotmath etc as.graphicsAnnot(unlist(caption[k])) if (is.null(sub.caption)) { ## construct a default: cal <- x$call if (!is.na(m.f <- match("formula", names(cal)))) { cal <- cal[c(1, m.f)] names(cal)[2] <- "" # drop " formula = " } cc <- deparse(cal, 80) # (80, 75) are ``parameters'' nc <- nchar(cc[1], "c") abbr <- length(cc) > 1 || nc > 75 sub.caption <- if (abbr) paste(substr(cc[1], 1, min(75, nc)), "...") else cc[1] } one.fig <- prod(par("mfcol")) == 1 if (ask) { oask <- devAskNewPage(TRUE) on.exit(devAskNewPage(oask)) } ##---------- Do the individual plots : ---------- if (show[1]) { ylim <- range(r, na.rm = TRUE) if (id.n > 0) ylim <- extendrange(r = ylim, f = 0.08) plot(yh, r, xlab = l.fit, ylab = "Residuals", main = main, ylim = ylim, type = "n", ...) panel(yh, r, ...) if (one.fig) title(sub = sub.caption, ...) mtext(getCaption(1), 3, 0.25, cex = cex.caption) if (id.n > 0) { y.id <- r[show.r] y.id[y.id < 0] <- y.id[y.id < 0] - strheight(" ")/3 text.id(yh[show.r], y.id, show.r) } abline(h = 0, lty = 3, col = "gray") } if (show[2]) { ## Normal ylim <- range(rs, na.rm = TRUE) ylim[2] <- ylim[2] + diff(ylim) * 0.075 qq <- qqnorm(rs, main = main, ylab = ylab23, ylim = ylim, ...) if (qqline) qqline(rs, lty = 3, col = "gray50") if (one.fig) title(sub = sub.caption, ...) mtext(getCaption(2), 3, 0.25, cex = cex.caption) if (id.n > 0) text.id(qq$x[show.rs], qq$y[show.rs], show.rs) } if (show[3]) { sqrtabsr <- sqrt(abs(rs)) ylim <- c(0, max(sqrtabsr, na.rm = TRUE)) yl <- as.expression(substitute(sqrt(abs(YL)), list(YL = as.name(ylab23)))) yhn0 <- if (is.null(w)) yh else yh[w != 0] plot(yhn0, sqrtabsr, xlab = l.fit, ylab = yl, main = main, ylim = ylim, type = "n", ...) panel(yhn0, sqrtabsr, ...) if (one.fig) title(sub = sub.caption, ...) mtext(getCaption(3), 3, 0.25, cex = cex.caption) if (id.n > 0) text.id(yhn0[show.rs], sqrtabsr[show.rs], show.rs) } if (show[4]) { if (id.n > 0) { show.r <- order(-cook)[iid]# index of largest 'id.n' ones ymx <- cook[show.r[1]] * 1.075 } else ymx <- max(cook, na.rm = TRUE) plot(cook, type = "h", ylim = c(0, ymx), main = main, xlab = "Obs. number", ylab = "Cook's distance", ...) if (one.fig) title(sub = sub.caption, ...) mtext(getCaption(4), 3, 0.25, cex = cex.caption) if (id.n > 0) text.id(show.r, cook[show.r], show.r, adj.x = FALSE) } if (show[5]) { ylim <- range(rs, na.rm = TRUE) if (id.n > 0) { ylim <- extendrange(r = ylim, f = 0.08) show.r <- order(-cook)[iid] } do.plot <- TRUE if (isConst.hat) {## leverages are all the same caption[5] <- "Constant Leverage:\n Residuals vs Factor Levels" ## plot against factor-level combinations instead aterms <- attributes(terms(x)) ## classes w/o response dcl <- aterms$dataClasses[-aterms$response] facvars <- names(dcl)[dcl %in% c("factor", "ordered")] mf <- model.frame(x)[facvars]# better than x$model if(ncol(mf) > 0) { ## now re-order the factor levels *along* factor-effects ## using a "robust" method {not requiring dummy.coef}: effM <- mf for(j in seq_len(ncol(mf))) effM[, j] <- sapply(split(yh, mf[, j]), mean)[mf[, j]] ord <- do.call(order, effM) dm <- data.matrix(mf)[ord, , drop = FALSE] ## #{levels} for each of the factors: nf <- length(nlev <- unlist(unname(lapply(x$xlevels, length)))) ff <- if(nf == 1) 1 else rev(cumprod(c(1, nlev[nf:2]))) facval <- ((dm-1) %*% ff) ## now reorder to the same order as the residuals facval[ord] <- facval xx <- facval # for use in do.plot section. plot(facval, rs, xlim = c(-1/2, sum((nlev-1) * ff) + 1/2), ylim = ylim, xaxt = "n", main = main, xlab = "Factor Level Combinations", ylab = ylab5, type = "n", ...) axis(1, at = ff[1]*(1:nlev[1] - 1/2) - 1/2, labels= x$xlevels[[1]][order(sapply(split(yh,mf[,1]), mean))]) mtext(paste(facvars[1],":"), side = 1, line = 0.25, adj=-.05) abline(v = ff[1]*(0:nlev[1]) - 1/2, col="gray", lty="F4") panel(facval, rs, ...) abline(h = 0, lty = 3, col = "gray") } else { # no factors message("hat values (leverages) are all = ", format(mean(r.hat)), "\n and there are no factor predictors; no plot no. 5") frame() do.plot <- FALSE } } else { ## Residual vs Leverage xx <- hii ## omit hatvalues of 1. xx[xx >= 1] <- NA plot(xx, rs, xlim = c(0, max(xx, na.rm = TRUE)), ylim = ylim, main = main, xlab = "Leverage", ylab = ylab5, type = "n", ...) panel(xx, rs, ...) abline(h = 0, v = 0, lty = 3, col = "gray") if (one.fig) title(sub = sub.caption, ...) if (length(cook.levels)) { p <- length(coef(x)) usr <- par("usr") hh <- seq.int(min(r.hat[1], r.hat[2]/100), usr[2], length.out = 101) for (crit in cook.levels) { cl.h <- sqrt(crit * p * (1 - hh)/hh) lines(hh, cl.h, lty = 2, col = 2) lines(hh, -cl.h, lty = 2, col = 2) } legend("bottomleft", legend = "Cook's distance", lty = 2, col = 2, bty = "n") xmax <- min(0.99, usr[2]) ymult <- sqrt(p * (1 - xmax)/xmax) aty <- c(-sqrt(rev(cook.levels)) * ymult, sqrt(cook.levels) * ymult) axis(4, at = aty, labels = paste(c(rev(cook.levels), cook.levels)), mgp = c(.25, .25, 0), las = 2, tck = 0, cex.axis = cex.id, col.axis = 2) } } # if(const h_ii) .. else .. if (do.plot) { mtext(getCaption(5), 3, 0.25, cex = cex.caption) if (id.n > 0) { y.id <- rs[show.r] y.id[y.id < 0] <- y.id[y.id < 0] - strheight(" ")/3 text.id(xx[show.r], y.id, show.r) } } } if (!one.fig && par("oma")[3] >= 1) mtext(sub.caption, outer = TRUE, cex = 1.25) invisible() } gnm/R/checkEstimable.R0000644000176000001440000000371512515147523014325 0ustar ripleyusers# Copyright (C) 2005, 2006, 2008, 2010 David Firth and Heather Turner # # This program is free software; you can redistribute it and/or modify # it under the terms of the GNU General Public License as published by # the Free Software Foundation; either version 2 or 3 of the License # (at your option). # # This program is distributed in the hope that it will be useful, # but WITHOUT ANY WARRANTY; without even the implied warranty of # MERCHANTABILITY or FITNESS FOR A PARTICULAR PURPOSE. See the # GNU General Public License for more details. # # A copy of the GNU General Public License is available at # http://www.r-project.org/Licenses/ checkEstimable <- function(model, combMatrix = diag(length(coef(model))), tolerance = NULL) { if (!inherits(model, "gnm")) stop("model not of class gnm") coefs <- coef(model) l <- length(coefs) combMatrix <- as.matrix(combMatrix) if (nrow(combMatrix) != l) stop( "dimensions of combMatrix do not match coef(model)") X <- model.matrix(model)[, !is.na(coefs), drop = FALSE] combMatrix <- scale(combMatrix[!is.na(coefs), ], center = FALSE) resultNA <- apply(combMatrix, 2, function(col) any(is.na(col))) result <- logical(ncol(combMatrix)) is.na(result) <- resultNA eliminate <- model$eliminate if (!is.null(eliminate)) { ## sweeps needed to get the rank right subtracted <- rowsum(X, eliminate)/tabulate(eliminate) if (attr(terms(model), "intercept") == 1) subtracted[,1] <- 0 X <- X - subtracted[eliminate, , drop = FALSE] } rankX <- model$rank - nlevels(eliminate) check.1 <- function(comb){ Xc <- rbind(X, comb) rankXc <- quickRank(Xc, tol = tolerance) return(rankXc == rankX) } result[!resultNA] <- apply(combMatrix[, !resultNA, drop = FALSE], 2, check.1) names(result) <- colnames(combMatrix) return(result) } gnm/R/Raise.R0000644000176000001440000000165712515147523012470 0ustar ripleyusers# Copyright (C) 2006, 2008 Heather Turner # # This program is free software; you can redistribute it and/or modify # it under the terms of the GNU General Public License as published by # the Free Software Foundation; either version 2 or 3 of the License # (at your option). # # This program is distributed in the hope that it will be useful, # but WITHOUT ANY WARRANTY; without even the implied warranty of # MERCHANTABILITY or FITNESS FOR A PARTICULAR PURPOSE. See the # GNU General Public License for more details. # # A copy of the GNU General Public License is available at # http://www.r-project.org/Licenses/ Raise <- function(expression, power = 1, inst = NULL){ list(predictors = list(substitute(expression)), term = function(predLabels, ...) { paste("(", predLabels, ")^", power, sep = "") }, call = as.expression(match.call()), match = 1) } class(Raise) <- "nonlin" gnm/R/exitInfo.R0000644000176000001440000000214312515147523013201 0ustar ripleyusers# Copyright (C) 2006 Heather Turner # # This program is free software; you can redistribute it and/or modify # it under the terms of the GNU General Public License as published by # the Free Software Foundation; either version 2 or 3 of the License # (at your option). # # This program is distributed in the hope that it will be useful, # but WITHOUT ANY WARRANTY; without even the implied warranty of # MERCHANTABILITY or FITNESS FOR A PARTICULAR PURPOSE. See the # GNU General Public License for more details. # # A copy of the GNU General Public License is available at # http://www.r-project.org/Licenses/ exitInfo <- function(object){ conv <- object$converged if (conv) cat("Algorithm converged\n") else { cat("\nTolerance: ", object$tolerance, "\n") cat("\nAbsolute scores >= ", "tolerance * sqrt(tolerance + diag(information matrix)):\n\n") score <- abs(attr(conv, "score")) fail <- score >= attr(conv, "criterion") print(data.frame(abs.score = score, criterion = attr(conv, "criterion"))[fail,]) } } gnm/R/kappa.gnm.R0000644000176000001440000000141612515147523013272 0ustar ripleyusers# Copyright (C) 2005 Heather Turner # # This program is free software; you can redistribute it and/or modify # it under the terms of the GNU General Public License as published by # the Free Software Foundation; either version 2 or 3 of the License # (at your option). # # This program is distributed in the hope that it will be useful, # but WITHOUT ANY WARRANTY; without even the implied warranty of # MERCHANTABILITY or FITNESS FOR A PARTICULAR PURPOSE. See the # GNU General Public License for more details. # # A copy of the GNU General Public License is available at # http://www.r-project.org/Licenses/ kappa.gnm <- function (z, ...) { if (inherits(z, "gnm", TRUE) == 1) stop("kappa is not implemented for gnm objects") else NextMethod } gnm/R/residSVD.R0000644000176000001440000000377112515147523013107 0ustar ripleyusers# Copyright (C) 2005, 2012 David Firth and Heather Turner # # This program is free software; you can redistribute it and/or modify # it under the terms of the GNU General Public License as published by # the Free Software Foundation; either version 2 or 3 of the License # (at your option). # # This program is distributed in the hope that it will be useful, # but WITHOUT ANY WARRANTY; without even the implied warranty of # MERCHANTABILITY or FITNESS FOR A PARTICULAR PURPOSE. See the # GNU General Public License for more details. # # A copy of the GNU General Public License is available at # http://www.r-project.org/Licenses/ residSVD <- function(model, fac1, fac2, d = 1) { if (!is.null(model$call$data)) { Data <- as.data.frame(eval(model$call$data, parent.frame())) fac1 <- eval(match.call()$fac1, Data) fac2 <- eval(match.call()$fac2, Data) } if (!inherits(model, "glm") && class(model) != "lm") stop( "model not of class lm, glm or gnm") if (!is.factor(fac1)) stop("fac1 must be a factor") if (!is.factor(fac2)) stop("fac2 must be a factor") Data <- data.frame(fac1, fac2) if (!is.null(model$na.action)) Data <- Data[-model$na.action, ] weights <- if (!is.null(model$weights)) as.vector(model$weights) else 1 X <- data.frame(rw = as.vector(model$residuals) * weights, w = weights) X <- lapply(X, tapply, Data, sum, simplify = TRUE) X <- X$rw/X$w X <- svd(naToZero(X), d, d) uPart <- sqrt(X$d[seq(d)]) * t(X$u) vPart <- sqrt(X$d[seq(d)]) * t(X$v) # uPartNegative <- apply(uPart, 1, function(row) all(row < 0)) # vPartNegative <- apply(vPart, 1, function(row) all(row < 0)) # multiplier <- ifelse(uPartNegative + vPartNegative == 1, -1, 1) multiplier <- 1 result <- t(cbind(uPart, vPart) * multiplier) rownames(result) <- c(paste("fac1", levels(fac1), sep = "."), paste("fac2", levels(fac2), sep = ".")) colnames(result) <- 1:d drop(result) } gnm/R/expandCategorical.R0000644000176000001440000000376712515147523015046 0ustar ripleyusers# Copyright (C) 2006, 2009, 2013, 2014 Heather Turner # # This program is free software; you can redistribute it and/or modify # it under the terms of the GNU General Public License as published by # the Free Software Foundation; either version 2 or 3 of the License # (at your option). # # This program is distributed in the hope that it will be useful, # but WITHOUT ANY WARRANTY; without even the implied warranty of # MERCHANTABILITY or FITNESS FOR A PARTICULAR PURPOSE. See the # GNU General Public License for more details. # # A copy of the GNU General Public License is available at # http://www.r-project.org/Licenses/ expandCategorical <- function(data, catvar, sep = ".", countvar = "count", idvar = "id", as.ordered = FALSE, group = TRUE) { cat <- interaction(data[catvar], sep = sep) ncat <- nlevels(cat) covar <- data[, -match(catvar, names(data)), drop = FALSE] catvar <- paste(catvar, collapse = sep) if (group == TRUE) { if (length(covar)) { ord <- do.call("order", covar) vars <- covar[ord, , drop = FALSE] dupvars <- duplicated(vars) d <- diff(c(which(!dupvars), length(dupvars) + 1)) n <- sum(!dupvars) id <- factor(rep(seq(n), d)) counts <- as.data.frame(table(list(cat = cat[ord], id = id))) newData <- vars[which(!dupvars)[counts$id], , drop = FALSE] rownames(newData) <- NULL newData[c(catvar, idvar, countvar)] <- counts } else { newData <- data.frame(table(cat)) colnames(newData) <- c(catvar, countvar) newData[[idvar]] <- factor(1) } } else { n <- nrow(covar) id <- gl(n, ncat) newData <- covar[id, , drop = FALSE] newData[[catvar]] <- gl(ncat, 1, n * ncat, labels = levels(cat), ordered = as.ordered) newData[[countvar]] <- as.vector(t(class.ind(cat))) newData[[idvar]] <- id } newData } gnm/R/weights.gnm.R0000644000176000001440000000147712515147523013657 0ustar ripleyusers# Copyright (C) 2008 Heather Turner # # This program is free software; you can redistribute it and/or modify # it under the terms of the GNU General Public License as published by # the Free Software Foundation; either version 2 or 3 of the License # (at your option). # # This program is distributed in the hope that it will be useful, # but WITHOUT ANY WARRANTY; without even the implied warranty of # MERCHANTABILITY or FITNESS FOR A PARTICULAR PURPOSE. See the # GNU General Public License for more details. # # A copy of the GNU General Public License is available at # http://www.r-project.org/Licenses/ weights.gnm <- function(object, type = c("prior", "working"), ...) { weights <- NextMethod("weights") if (!is.null(object$table.attr)) attributes(weights) <- object$table.attr weights } gnm/R/cooks.distance.gnm.R0000755000176000001440000000220112515147523015101 0ustar ripleyusers# Modification of cooks.distance.glm from the stats package for R. # # Copyright (C) 1995-2005 The R Core Team # Copyright (C) 2005, 2006 Heather Turner # # This program is free software; you can redistribute it and/or modify # it under the terms of the GNU General Public License as published by # the Free Software Foundation; either version 2 or 3 of the License # (at your option). # # This program is distributed in the hope that it will be useful, # but WITHOUT ANY WARRANTY; without even the implied warranty of # MERCHANTABILITY or FITNESS FOR A PARTICULAR PURPOSE. See the # GNU General Public License for more details. # # A copy of the GNU General Public License is available at # http://www.r-project.org/Licenses/ cooks.distance.gnm <- function(model, hat = hatvalues(model), dispersion = attr(vcov(model), "dispersion"), ...){ p <- model$rank res <- na.omit(residuals(model, type = "pearson"))[model$prior.weights != 0] res <- naresid(model$na.action, res) res <- (res/(1 - hat))^2 * hat/(dispersion * p) res[is.infinite(res)] <- NaN res } gnm/R/anova.gnm.R0000644000176000001440000000763512515147523013313 0ustar ripleyusers# Modification of anova.glm from the stats package for R. # # Copyright (C) 1995-2005 The R Core Team # Copyright (C) 2005, 2006, 2008, 2012 Heather Turner # # This program is free software; you can redistribute it and/or modify # it under the terms of the GNU General Public License as published by # the Free Software Foundation; either version 2 or 3 of the License # (at your option). # # This program is distributed in the hope that it will be useful, # but WITHOUT ANY WARRANTY; without even the implied warranty of # MERCHANTABILITY or FITNESS FOR A PARTICULAR PURPOSE. See the # GNU General Public License for more details. # # A copy of the GNU General Public License is available at # http://www.r-project.org/Licenses/ anova.gnm <- function (object, ..., dispersion = NULL, test = NULL) { dotargs <- list(...) named <- if (is.null(names(dotargs))) rep(FALSE, length(dotargs)) else (names(dotargs) != "") if (any(named)) warning("the following arguments to 'anova.gnm' are invalid and dropped: ", paste(deparse(dotargs[named]), collapse = ", ")) dotargs <- dotargs[!named] is.gnm <- unlist(lapply(dotargs, function(x) inherits(x, c("gnm", "glm")))) dotargs <- dotargs[is.gnm] if (length(dotargs) > 0) return(anova(structure(c(list(object), dotargs), class="glmlist"), dispersion = dispersion, test = test)) x <- model.matrix(object) varlist <- attr(terms(object), "term.labels") varseq <- attr(x, "assign") pars <- setdiff(unique(varseq), c(0, varseq[object$constrain])) nvars <- length(varlist) nonlinear <- match(TRUE, attr(terms(object), "type") != "Linear") if (is.na(nonlinear)) nonlinear <- nvars + 1 resdev <- resdf <- fit <- NULL origConstrain <- object$constrain origConstrainTo <- object$constrainTo if (nvars > 0) { for (i in pars) { if (i < nonlinear && is.null(object$eliminate)){ fit <- glm.fit(x = x[, varseq < i, drop = FALSE], y = c(object$y), offset = c(object$offset), start = object$start, weights = c(object$prior.weights), family = object$family) } else { f <- update.formula(formula(object), paste(". ~ . -", paste(varlist[i:nvars], collapse = " - "))) f <- update.formula(formula(object), f) fit <- update(object, formula = f, verbose = FALSE) } resdev <- c(resdev, fit$deviance) resdf <- c(resdf, fit$df.residual) } resdf <- c(resdf, object$df.residual) resdev <- c(resdev, object$deviance) table <- data.frame(c(NA, -diff(resdf)), c(NA, pmax(0, -diff(resdev))), resdf, resdev) } else table <- data.frame(NA, NA, object$df.residual, object$deviance) dimnames(table) <- list(c("NULL", labels(object)), c("Df", "Deviance", "Resid. Df", "Resid. Dev")) title <- paste("Analysis of Deviance Table", "\n\nModel: ", object$family$family, ", link: ", object$family$link, "\n\nResponse: ", as.character(formula(object)[[2]]), "\n\nTerms added sequentially (first to last)\n\n", sep = "") df.dispersion <- Inf if (is.null(dispersion)) { dispersion <- attr(vcov(object), "dispersion") df.dispersion <- if (dispersion == 1) Inf else object$df.residual } if (!is.null(test)) table <- stat.anova(table = table, test = test, scale = dispersion, df.scale = df.dispersion, n = NROW(x)) structure(table, heading = title, class = c("anova", "data.frame")) } gnm/R/gnmTools.R0000755000176000001440000002363512515525407013233 0ustar ripleyusers# Copyright (C) 2005-2012 Heather Turner and David Firth # # This program is free software; you can redistribute it and/or modify # it under the terms of the GNU General Public License as published by # the Free Software Foundation; either version 2 or 3 of the License # (at your option). # # This program is distributed in the hope that it will be useful, # but WITHOUT ANY WARRANTY; without even the implied warranty of # MERCHANTABILITY or FITNESS FOR A PARTICULAR PURPOSE. See the # GNU General Public License for more details. # # A copy of the GNU General Public License is available at # http://www.r-project.org/Licenses/ "gnmTools" <- function(modelTerms, gnmData = NULL, x = TRUE) { eliminate <- attr(modelTerms, "eliminate") unitLabels <- attr(modelTerms, "unitLabels") common <- attr(modelTerms, "common") prefixLabels <- attr(modelTerms, "prefixLabels") match <- attr(modelTerms, "match") varLabels <- attr(modelTerms, "varLabels") block <- attr(modelTerms, "block") type <- attr(modelTerms, "type") nFactor <- length(varLabels) seqFactor <- seq(nFactor) termTools <- factorAssign <- thetaID <- vector(mode = "list", length = nFactor) blockID <- unique(block) adj <- 1 for (i in blockID) { b <- block == i if (all(common[b])) { ## get full set of levels facs <- sapply(unitLabels[b], function(x) { is.factor(eval(parse(text = x), gnmData))}) if (!all(facs)) stop(paste(c("The following should be factors:", unitLabels[b][!facs]), collapse = "\n")) allLevels <- lapply(unitLabels[b], function(x) levels(eval(parse(text = x), gnmData))) labels <- unique(unlist(allLevels)) if (!all(mapply(identical, allLevels, list(labels)))) { labels <- sort(labels) } nLevels <- length(labels) ## create design matrices termTools[b] <- lapply(unitLabels[b], function(x) { class.ind(factor(eval(parse(text = x), gnmData), levels = labels)) }) ## create labels i <- which(b) nm <- paste(prefixLabels[i], labels, sep = "") factorAssign[b] <- lapply(i, function(x, nLevels, nm) structure(rep(x, nLevels), names = nm), nLevels, nm) adj <- adj + nLevels } else { intercept <- as.numeric(i == 0 && eliminate) tmp <- model.matrix(terms(reformulate(c(intercept, unitLabels[b])), keep.order = TRUE), data = gnmData) tmpAssign <- attr(tmp, "assign") if (intercept) { tmp <- tmp[,-1, drop = FALSE] tmpAssign <- tmpAssign[-1] } tmpAssign <- which(b)[tmpAssign + !tmpAssign[1]] ## don't paste "(Intercept)" if non-empty prefix and only parameter prefixOnly <- {identical(colnames(tmp), "(Intercept)") && prefixLabels[tmpAssign] != ""} nm <- paste(prefixLabels[tmpAssign], colnames(tmp)[!prefixOnly], sep = "") names(tmpAssign) <- nm termTools[b] <- lapply(split(1:ncol(tmp), tmpAssign), function(i, M) M[, i , drop = FALSE], tmp) factorAssign[b] <- split(tmpAssign, tmpAssign) adj <- adj + length(tmpAssign) } } factorAssign <- unlist(factorAssign) uniq <- !(duplicated(block) & common)[factorAssign] parLabels <- names(factorAssign) nTheta <- length(factorAssign) thetaID <- numeric(nTheta) thetaID[uniq] <- seq(sum(uniq)) thetaID[!uniq] <- thetaID[common[factorAssign] & uniq] nr <- dim(gnmData)[1] tmp <- seq(factorAssign) * nr first <- c(0, tmp[-nTheta]) firstX <- first[thetaID] last <- tmp - 1 lastX <- last[thetaID] + 1 nc <- tabulate(factorAssign) tmp <- cumsum(nc) a <- c(1, tmp[-nFactor] + 1) z <- tmp lt <- last[z] - first[a] + 1 storage.mode(firstX) <- storage.mode(first) <- storage.mode(lastX) <- storage.mode(last) <- storage.mode(a) <- storage.mode(z) <- storage.mode(lt) <- "integer" baseMatrix <- matrix(1, nrow = nr, ncol = nTheta) for (i in seq(termTools)) if (is.matrix(termTools[[i]])) baseMatrix[, factorAssign == i] <- termTools[[i]] X <- baseMatrix colID <- match(thetaID, thetaID) thetaID <- split(thetaID, factorAssign) names(thetaID) <- varLabels if (any(duplicated(parLabels[uniq]))){ parLabels[uniq] <- make.unique(parLabels[uniq]) warning("Using make.unique() to make default parameter labels unique", call. = FALSE) } colnames(X) <- parLabels X <- X[, uniq, drop = FALSE] ## check for zero columns constrain <- which(colSums(X) == 0) theta <- rep(NA, nTheta) for (i in blockID) { b <- block == i if (sum(b) == 1 && is.list(termTools[[which(b)]]) && !is.null(termTools[[which(b)]]$start)){ theta[unlist(thetaID[b])] <- termTools[[which(b)]]$start } } names(theta) <- parLabels termAssign <- attr(modelTerms, "assign")[factorAssign] block <- block[factorAssign] for (i in seq(attr(modelTerms, "predictor"))) { if (!is.null(attr(modelTerms, "start")[[i]])) { termID <- termAssign == i & uniq split <- block[termID] split <- match(split, unique(split)) theta[termID] <- attr(modelTerms, "start")[[i]](structure(theta[termID], assign = split)) } } theta <- theta[uniq] if (attr(modelTerms, "intercept")) termAssign <- termAssign - 1 prodList <- vector(mode = "list", length = nFactor) names(prodList) <- varLabels type <- type == "Special" varPredictors <- function(theta) { for (i in seqFactor) { prodList[[i]] <- .Call("submatprod", baseMatrix, theta[thetaID[[i]]], first[a[i]], nr, nc[i], PACKAGE = "gnm", NAOK = TRUE) } prodList } predictor <- function(varPredictors, term = FALSE) { if (term) { es <- lapply(attr(modelTerms, "predictor"), function(x) { do.call("bquote", list(x, gnmData))}) tp <- matrix(sapply(es, eval, c(varPredictors, gnmData)), nr) colnames(tp) <- c("(Intercept)"[attr(modelTerms, "intercept")], attr(modelTerms, "term.labels")) tp } else eval(e, c(varPredictors, gnmData)) } gnmData <- lapply(gnmData[, !names(gnmData) %in% varLabels, drop = FALSE], drop) e <- sumExpression(attr(modelTerms, "predictor")) varDerivs <- lapply(varLabels, deriv, expr = e) commonAssign <- factorAssign[colID] nCommon <- table(commonAssign[!duplicated(factorAssign)]) tmpID <- unique(commonAssign) tmpID <- tmpID[type[tmpID]] nCommon <- as.integer(nCommon[as.character(tmpID)]) if (any(type)) specialVarDerivs <- deriv(e, varLabels[type]) convID <- colID[uniq] vID <- cumsum(c(1, nCommon))[seq(nCommon)] localDesignFunction <- function(theta, varPredictors, ind = NULL) { if (!any(common)) { if (!is.null(ind)){ i1 <- convID[ind] tmpID <- commonAssign[i1] } for (i in tmpID) { fi <- unique(factorAssign[commonAssign == i]) if (is.null(ind)){ i1 <- a[fi][1] i2 <- z[fi][1] } else { i2 <- i1 a <- ind if (factorAssign[ind] > 1) ind <- ind - z[factorAssign[ind] - 1] } if (type[fi]) { v <- attr(eval(varDerivs[[fi]], c(varPredictors, gnmData)), "gradient") .Call("subprod", X, baseMatrix, as.double(v), first[i1], last[i2], nr, PACKAGE = "gnm") } } if(!is.null(ind)) X[, a, drop = FALSE] else X } else { if (is.null(ind)){ v <- attr(eval(specialVarDerivs, c(varPredictors, gnmData)), "gradient") .Call("newsubprod", baseMatrix, as.double(v), X, first[a[tmpID]], first[vID], firstX[a[tmpID]], as.integer(length(nCommon)), lt[tmpID], lastX[z[tmpID]], nr, nCommon, max(nCommon), PACKAGE = "gnm") } else { i1 <- convID[ind] fi <- unique(factorAssign[commonAssign == commonAssign[i1]]) v <- list() for(j in fi) v[[j]] <- attr(eval(varDerivs[[j]], c(varPredictors, gnmData)), "gradient") .Call("onecol", baseMatrix, as.double(unlist(v[fi])), first[i1], lt[fi[1]], nr, as.integer(length(fi)), PACKAGE = "gnm") } } } toolList <- list(start = theta, constrain = constrain, varPredictors = varPredictors, predictor = predictor, localDesignFunction = localDesignFunction) if (x) toolList$termAssign <- termAssign[uniq] toolList } gnm/R/glm.fit.e.R0000644000176000001440000001441112515147523013200 0ustar ripleyusers# This fits a glm with eliminated factor, and should be much quicker # than glm.fit when the number of levels of the eliminated factor is large. # # Copyright (C) 2009, 2010, 2012 David Firth and Heather Turner # # This program is free software; you can redistribute it and/or modify # it under the terms of the GNU General Public License as published by # the Free Software Foundation; either version 2 or 3 of the License # (at your option). # # This program is distributed in the hope that it will be useful, # but WITHOUT ANY WARRANTY; without even the implied warranty of # MERCHANTABILITY or FITNESS FOR A PARTICULAR PURPOSE. See the # GNU General Public License for more details. # # A copy of the GNU General Public License is available at # http://www.r-project.org/Licenses/ glm.fit.e <- function( x, y, weights = rep(1, NROW(y)), start = NULL, etastart = NULL, mustart = NULL, offset = rep(0, NROW(y)), family = gaussian(), control = glm.control(), ## only for compatibility with glm.fit intercept = TRUE, ## only for compatibility with glm.fit eliminate = NULL, ## alternatively a factor ridge = 1e-8, coefonly = FALSE) { if (is.null(eliminate)) { ## just revert to glm.fit ## can make a difference in timing! tmp <- glm.fit(x, y, weights = weights, start = start, etastart = etastart, mustart = mustart, offset = offset, family = family, control = control, intercept = intercept) if (coefonly) return(tmp$coef) else return(tmp) } ## The rest handles the case of an eliminated factor names(y) <- rownames(x) <- NULL nobs <- NROW(y) non.elim <- ncol(x) if (is.null(weights)) weights <- rep.int(1, nobs) if (is.null(offset)) offset <- rep.int(0, nobs) link <- family$linkfun linkinv <- family$linkinv linkder <- family$mu.eta variance <- family$variance dev.resids <- family$dev.resids aic <- family$aic ## sort data to help compute group means quickly ord <- order(xtfrm(eliminate)) if (ordTRUE <- !identical(ord, seq(eliminate))) { y <- as.numeric(y[ord]) weights <- weights[ord] offset <- offset[ord] if (non.elim) x <- x[ord, , drop = FALSE] eliminate <- eliminate[ord] } size <- tabulate(eliminate) end <- cumsum(size) nelim <- rank <- nlevels(eliminate) elim <- seq.int(nelim) if (is.null(start)) { # use either y or etastart or mustart if (!is.null(etastart)) mustart <- linkinv(etastart) if (!is.null(mustart)) z <- mustart else z <- y elim.means <- grp.sum(z, end)/size os.by.level <- link(0.999 * elim.means + 0.001 * mean(z)) - grp.sum(offset, end)/size } else os.by.level <- start[elim] os.vec <- os.by.level[eliminate] eta.stored <- eta <- offset + os.vec mu <- linkinv(eta) mu.eta <- linkder(eta) z <- eta - offset + (y - mu) / mu.eta w <- weights * (mu.eta)^2/variance(mu) counter <- 0 devold <- 0 if (intercept) x <- x[, -1, drop = FALSE] #non-null eliminate if (non.elim) { ## sweeps needed to get the rank right subtracted <- rowsum.default(x, eliminate, reorder = FALSE)/size x <- x - subtracted[eliminate, , drop = FALSE] ## initial fit to drop aliased columns model <- lm.wfit(x, z, w, offset = os.vec) full.theta <- model$coefficients eta <- model$fitted + offset rank <- model$rank + nelim rm(model) mu <- linkinv(eta) mu.eta <- linkder(eta) z <- eta - offset + (y - mu) / mu.eta w <- weights * (mu.eta)^2/variance(mu) est <- !is.na(full.theta) x <- x[, est, drop = FALSE] theta <- full.theta[est] } Z <- cbind(z, x) I1 <- numeric(ncol(Z)) I1[1] <- 1 for (i in 1:control$maxit) { ## try without scaling etc - already of full rank Tvec <- sqrt(grp.sum(w, end)) Umat <- rowsum.default(w * Z, eliminate, reorder = FALSE) Umat <- Umat/Tvec Wmat <- crossprod(sqrt(w) * Z) diag(Wmat) <- diag(Wmat) + ridge Qi <- solve(Wmat - crossprod(Umat), I1) theta <- -Qi[-1]/Qi[1] os.by.level <- ((Umat %*% Qi)/Qi[1])/Tvec if (non.elim) eta <- drop(x %*% theta + offset + os.by.level[eliminate]) else eta <- offset + os.by.level[eliminate] mu <- linkinv(eta) dev <- sum(dev.resids(y, mu, weights)) if (control$trace) cat("Deviance =", dev, "Iterations -", i, "\n") if (abs(dev - devold)/(0.1 + abs(dev)) < control$epsilon) { conv <- TRUE break } devold <- dev mu.eta <- linkder(eta) Z[,1] <- eta - offset + (y - mu) / mu.eta w <- weights * (mu.eta)^2/variance(mu) } converged <- !(i == control$maxit) if (!converged) warning(paste("The convergence criterion was not met after", control$maxit, "iterations.")) names(os.by.level) <- paste("(eliminate)", elim, sep = "") if (non.elim) { full.theta[est] <- theta os.by.level <- os.by.level - subtracted %*% naToZero(full.theta) } else full.theta <- numeric(0) if (ordTRUE) { reorder <- order(ord) y <- y[reorder] mu <- mu[reorder] eta <- eta[reorder] weights <- weights[reorder] } mu.eta <- linkder(eta) w <- weights * (mu.eta)^2/variance(mu) if (coefonly) return(structure(full.theta, eliminated = c(os.by.level))) aic.model <- aic(y, sum(weights > 0), mu, weights, dev) + 2 * rank eliminated <- structure(c(os.by.level), names = levels(eliminate)) list(coefficients = structure(full.theta, eliminated = eliminated), residuals = (y - mu) / linkder(eta), fitted.values = mu, rank = rank, family = family, linear.predictors = eta, deviance = dev, aic = aic.model, iter = i, weights = w, prior.weights = weights, df.residual = nobs - sum(weights == 0) - rank, y = y, converged = converged) ## NB: some components of the result of glm.fit are missing from this list } gnm/R/Log.R0000644000176000001440000000163312515147523012140 0ustar ripleyusers# Copyright (C) 2006, 2008 Heather Turner # # This program is free software; you can redistribute it and/or modify # it under the terms of the GNU General Public License as published by # the Free Software Foundation; either version 2 or 3 of the License # (at your option). # # This program is distributed in the hope that it will be useful, # but WITHOUT ANY WARRANTY; without even the implied warranty of # MERCHANTABILITY or FITNESS FOR A PARTICULAR PURPOSE. See the # GNU General Public License for more details. # # A copy of the GNU General Public License is available at # http://www.r-project.org/Licenses/ Log <- function(expression, inst = NULL){ list(predictors = list(substitute(expression)), term = function(predLabels, ...) { paste("log(", predLabels, ")", sep = "") }, call = as.expression(match.call()), match = 1) } class(Log) <- "nonlin" gnm/R/se.R0000644000176000001440000000653412515147523012033 0ustar ripleyusers# Copyright (C) 2005, 2006, 2010 David Firth and Heather Turner # # This program is free software; you can redistribute it and/or modify # it under the terms of the GNU General Public License as published by # the Free Software Foundation; either version 2 or 3 of the License # (at your option). # # This program is distributed in the hope that it will be useful, # but WITHOUT ANY WARRANTY; without even the implied warranty of # MERCHANTABILITY or FITNESS FOR A PARTICULAR PURPOSE. See the # GNU General Public License for more details. # # A copy of the GNU General Public License is available at # http://www.r-project.org/Licenses/ ## now only computes se for non-eliminated parameters se <- function(model, estimate = NULL, checkEstimability = TRUE, Vcov = NULL, dispersion = NULL, ...){ if (!inherits(model, "gnm")) stop("model is not of class \"gnm\"") if (!is.null(Vcov) && !is.null(dispersion)){ Vcov <- Vcov * dispersion } else { Vcov <- vcov(model, dispersion = dispersion, use.eliminate = FALSE) } if (!length(Vcov)) return("Model has no non-eliminated parameters") coefs <- coef(model) coefNames <- names(coefs) eliminate <- model$eliminate nelim <- nlevels(eliminate) l <- length(coefs) if (identical(estimate, "[?]")) estimate <- pickCoef(model, title = paste("Estimate standard errors", "for one or more gnm coefficients")) if (is.null(estimate)){ if (!is.null(model$ofInterest)) estimate <- ofInterest(model) else estimate <- seq(model$coefficients) } if (is.character(estimate)) estimate <- match(estimate, coefNames, 0) if (is.vector(estimate) && all(estimate %in% seq(coefs))) { if (!length(estimate)) stop("no non-eliminated coefficients specified by 'estimate'", "argument") comb <- naToZero(coefs[estimate]) var <- Vcov[estimate, estimate] coefMatrix <- matrix(0, l, length(comb)) coefMatrix[cbind(estimate, seq(length(comb)))] <- 1 colnames(coefMatrix) <- names(comb) } else { coefMatrix <- as.matrix(estimate) if (!is.numeric(coefMatrix)) stop("'estimate' should specify parameters using ", "\"pick\" or a vector of \n names/indices; ", "or specify linear combinations using ", "a numeric vector/matrix.") if (nrow(coefMatrix) != l) stop("NROW(estimate) should equal ", "length(coef(model)) - nlevels(model$eliminate)") comb <- drop(crossprod(coefMatrix, naToZero(coefs))) var <- crossprod(coefMatrix, crossprod(Vcov, coefMatrix)) } estimable <- rep(TRUE, ncol(coefMatrix)) if (checkEstimability) { estimable <- checkEstimable(model, coefMatrix, ...) if (any(!na.omit(estimable))) cat("Std. Error is NA where estimate is fixed or unidentified\n") } if (is.matrix(var)) sterr <- sqrt(diag(var)) else sterr <- sqrt(var) is.na(sterr[estimable %in% c(FALSE, NA)]) <- TRUE result <- data.frame(comb, sterr) rowNames <- colnames(coefMatrix) if (is.null(rowNames)) rowNames <- paste("Combination", ncol(coefMatrix)) dimnames(result) <- list(rowNames, c("Estimate", "Std. Error")) result } gnm/R/boxcox.gnm.R0000644000176000001440000000143212515147523013476 0ustar ripleyusers# Copyright (C) 2005 Heather Turner # # This program is free software; you can redistribute it and/or modify # it under the terms of the GNU General Public License as published by # the Free Software Foundation; either version 2 or 3 of the License # (at your option). # # This program is distributed in the hope that it will be useful, # but WITHOUT ANY WARRANTY; without even the implied warranty of # MERCHANTABILITY or FITNESS FOR A PARTICULAR PURPOSE. See the # GNU General Public License for more details. # # A copy of the GNU General Public License is available at # http://www.r-project.org/Licenses/ boxcox.gnm <- function (object, ...) { if (inherits(object, "gnm", TRUE) == 1) stop("boxcox is not implemented for gnm objects") else NextMethod } gnm/R/Inv.R0000644000176000001440000000163312515147523012153 0ustar ripleyusers# Copyright (C) 2006, 2008 Heather Turner # # This program is free software; you can redistribute it and/or modify # it under the terms of the GNU General Public License as published by # the Free Software Foundation; either version 2 or 3 of the License # (at your option). # # This program is distributed in the hope that it will be useful, # but WITHOUT ANY WARRANTY; without even the implied warranty of # MERCHANTABILITY or FITNESS FOR A PARTICULAR PURPOSE. See the # GNU General Public License for more details. # # A copy of the GNU General Public License is available at # http://www.r-project.org/Licenses/ Inv <- function(expression, inst = NULL){ list(predictors = list(substitute(expression)), term = function(predLabels, ...) { paste("(", predLabels, ")^-1", sep = "") }, call = as.expression(match.call()), match = 1) } class(Inv) <- "nonlin" gnm/R/print.meanResiduals.R0000644000176000001440000000203412515147523015342 0ustar ripleyusers# Copyright (C) 2010-2012 Heather Turner # # This program is free software; you can redistribute it and/or modify # it under the terms of the GNU General Public License as published by # the Free Software Foundation; either version 2 or 3 of the License # (at your option). # # This program is distributed in the hope that it will be useful, # but WITHOUT ANY WARRANTY; without even the implied warranty of # MERCHANTABILITY or FITNESS FOR A PARTICULAR PURPOSE. See the # GNU General Public License for more details. # # A copy of the GNU General Public License is available at # http://www.r-project.org/Licenses/ print.meanResiduals <- function (x, digits = max(3, getOption("digits") - 3), ...) { cat("\nModel call:\n", deparse(attr(x, "call"), width.cutoff = options()$width), "\n", sep = "", fill = TRUE) cat("Mean residuals by ", attr(x, "by"), ":\n\n", sep = "") if (!inherits(x, "table")) x <- as.numeric(x) NextMethod(object = x, digits = digits, print.gap = 2, ...) } gnm/R/Symm.R0000644000176000001440000000224412515147523012343 0ustar ripleyusers# Copyright (C) 2005, 2006, 2008 David Firth and Heather Turner # # This program is free software; you can redistribute it and/or modify # it under the terms of the GNU General Public License as published by # the Free Software Foundation; either version 2 or 3 of the License # (at your option). # # This program is distributed in the hope that it will be useful, # but WITHOUT ANY WARRANTY; without even the implied warranty of # MERCHANTABILITY or FITNESS FOR A PARTICULAR PURPOSE. See the # GNU General Public License for more details. # # A copy of the GNU General Public License is available at # http://www.r-project.org/Licenses/ Symm <- function(...){ dots <- list(...) if (any(diff(sapply(dots, length)) != 0)) stop( "arguments to symm() must all have same length") dots <- lapply(dots, as.factor) facMatrix <- sapply(dots, as.character) f <- function(row){ string <- paste(sort(row), collapse = "") if (any(is.na(row))) is.na(string) <- TRUE string } if (inherits(facMatrix, "matrix")) result <- factor(apply(facMatrix, 1, f)) else result <- factor(f(facMatrix)) result } gnm/R/Exp.R0000755000176000001440000000165312515147523012160 0ustar ripleyusers# Copyright (C) 2005, 2006 Heather Turner and David Firth # # This program is free software; you can redistribute it and/or modify # it under the terms of the GNU General Public License as published by # the Free Software Foundation; either version 2 or 3 of the License # (at your option). # # This program is distributed in the hope that it will be useful, # but WITHOUT ANY WARRANTY; without even the implied warranty of # MERCHANTABILITY or FITNESS FOR A PARTICULAR PURPOSE. See the # GNU General Public License for more details. # # A copy of the GNU General Public License is available at # http://www.r-project.org/Licenses/ Exp <- function(expression, inst = NULL){ list(predictors = list(substitute(expression)), term = function(predLabels, ...) { paste("exp(", predLabels, ")", sep = "") }, call = as.expression(match.call()), match = 1) } class(Exp) <- "nonlin" gnm/R/psum.R0000755000176000001440000000206112515147523012402 0ustar ripleyusers# Copyright (C) 2005 Heather Turner # # This program is free software; you can redistribute it and/or modify # it under the terms of the GNU General Public License as published by # the Free Software Foundation; either version 2 or 3 of the License # (at your option). # # This program is distributed in the hope that it will be useful, # but WITHOUT ANY WARRANTY; without even the implied warranty of # MERCHANTABILITY or FITNESS FOR A PARTICULAR PURPOSE. See the # GNU General Public License for more details. # # A copy of the GNU General Public License is available at # http://www.r-project.org/Licenses/ psum <- function(...) { summandList <- list(...) nSummands <- length(summandList) if (nSummands == 0) return(0) else if (nSummands == 1) return(summandList[[1]]) else { trySum <- try(summandList[[1]] + do.call("Recall", summandList[-1]), silent = TRUE) if (inherits(trySum, "try-error")) stop("addition not implemented for types of argument supplied") else trySum } } gnm/R/coef.gnm.R0000644000176000001440000000143212515147523013110 0ustar ripleyusers# Copyright (C) 2005, 2006, 2010, 2012 Heather Turner # # This program is free software; you can redistribute it and/or modify # it under the terms of the GNU General Public License as published by # the Free Software Foundation; either version 2 or 3 of the License # (at your option). # # This program is distributed in the hope that it will be useful, # but WITHOUT ANY WARRANTY; without even the implied warranty of # MERCHANTABILITY or FITNESS FOR A PARTICULAR PURPOSE. See the # GNU General Public License for more details. # # A copy of the GNU General Public License is available at # http://www.r-project.org/Licenses/ coef.gnm <- function(object, ...) { structure(object$coefficients, ofInterest = object$ofInterest, class = c("coef.gnm", "numeric")) } gnm/R/ofInterestReplacement.R0000644000176000001440000000167112515147523015723 0ustar ripleyusers# Copyright (C) 2006 Heather Turner # # This program is free software; you can redistribute it and/or modify # it under the terms of the GNU General Public License as published by # the Free Software Foundation; either version 2 or 3 of the License # (at your option). # # This program is distributed in the hope that it will be useful, # but WITHOUT ANY WARRANTY; without even the implied warranty of # MERCHANTABILITY or FITNESS FOR A PARTICULAR PURPOSE. See the # GNU General Public License for more details. # # A copy of the GNU General Public License is available at # http://www.r-project.org/Licenses/ "ofInterest<-" <- function(object, value = NULL) { coefNames <- names(coef(object)) if (!is.null(value)) { if (!any(value %in% seq(coefNames))) stop("One or more replacement value is invalid.") names(value) <- coefNames[value] } object$ofInterest <- value print(value) object } gnm/R/quickRank.R0000644000176000001440000000174612515147523013354 0ustar ripleyusers# as tolNorm2 method in rankMatrix from the Matrix package, but avoids validity # checks - much faster if need to do repeated rank calculations # # Copyright (C) 2007 Martin Maechler # Copyright (C) 2010 Heather Turner # # This program is free software; you can redistribute it and/or modify # it under the terms of the GNU General Public License as published by # the Free Software Foundation; either version 2 or 3 of the License # (at your option). # # This program is distributed in the hope that it will be useful, # but WITHOUT ANY WARRANTY; without even the implied warranty of # MERCHANTABILITY or FITNESS FOR A PARTICULAR PURPOSE. See the # GNU General Public License for more details. # # A copy of the GNU General Public License is available at # http://www.r-project.org/Licenses/ quickRank <- function(X, tol = NULL) { sval <- svd(X, 0, 0)$d if (is.null(tol)) sum(sval >= max(dim(X)) * .Machine$double.eps * sval[1]) else sum(sval >= tol) } gnm/R/Dref.R0000755000176000001440000000366112515147523012305 0ustar ripleyusers# Copyright (C) 2005-2007, 2012 Heather Turner # # This program is free software; you can redistribute it and/or modify # it under the terms of the GNU General Public License as published by # the Free Software Foundation; either version 2 or 3 of the License # (at your option). # # This program is distributed in the hope that it will be useful, # but WITHOUT ANY WARRANTY; without even the implied warranty of # MERCHANTABILITY or FITNESS FOR A PARTICULAR PURPOSE. See the # GNU General Public License for more details. # # A copy of the GNU General Public License is available at # http://www.r-project.org/Licenses/ Dref <- function(..., delta = ~ 1){ preds <- match.call(expand.dots = FALSE)[["..."]] n <- length(preds) preds <- c(delta = rep(list(delta), n), preds) common <- c(1:n, rep(n + 1, n)) extra <- setdiff(names(match.call()[-1]), c("", "delta")) if (length(extra)) stop(paste(c("invalid argument passed to Dref:", extra), collapse = " ")) nf <- match(c("delta"), names(match.call()[-1]), 0) if ("formula" %in% names(match.call()[-1])) stop("formula argument of old plug-in has been renamed ", "\"delta\" in this function.") match <- c(rep(nf, n), 1:n) names(preds) <- c(rep("delta", n), rep("", n)) list(predictors = preds, common = common, match = match, term = function(predLabels, ...){ delta <- predLabels[1:n] gamma <- predLabels[-c(1:n)] paste("(((exp(", delta, "))/(", paste("exp(", delta, ")", collapse = " + "), "))*", gamma, ")", sep = "", collapse = " + ")}, start = function(theta) { ifelse(attr(theta, "assign") == n + 1, 0.5, runif(length(theta)) - 0.5) }, call = as.expression(match.call())) } class(Dref) <- "nonlin" gnm/R/quick.glm.fit.R0000644000176000001440000000645212515147523014076 0ustar ripleyusers# Copyright (C) 2006 David Firth # # This program is free software; you can redistribute it and/or modify # it under the terms of the GNU General Public License as published by # the Free Software Foundation; either version 2 or 3 of the License # (at your option). # # This program is distributed in the hope that it will be useful, # but WITHOUT ANY WARRANTY; without even the implied warranty of # MERCHANTABILITY or FITNESS FOR A PARTICULAR PURPOSE. See the # GNU General Public License for more details. # # A copy of the GNU General Public License is available at # http://www.r-project.org/Licenses/ "quick.glm.fit" <- ## A wrapper for glm.fit, which is much faster when a large number ## of parameters can be eliminated, but which typically (if nIter is small) ## stops before convergence. Useful for getting gnm starting values. ## ## The eliminate argument is assumed numeric (no. of columns in X). No ## check is done here on "eliminability" of the specified columns. ## ## The non-eliminated columns are assumed not to include the intercept (ie ## no column of ones). ## ## When eliminate is used, only the "coefficients" component is returned. ## (for reasons of speed/laziness). This is fine for gnm purposes, but if ## quick.glm.fit is made into a `method' for glm() fits then the result ## needs to have various other components added. ## ## No account is taken of NAs -- will that be a problem, or have they gone by ## the time glm.fit gets called? ## function (x, y, weights = rep(1, length(y)), offset = rep(0, length(y)), family = gaussian(), eliminate = 0, nIter = 2, verbose = FALSE) { if (eliminate == 0) return(suppressWarnings(glm.fit(x, y, weights = weights, offset = offset, family = family)$coef)) ## The rest handles the case of eliminated columns in X xElim <- x[ , seq(eliminate), drop = FALSE] if (eliminate < ncol(x)) xNotElim <- cbind(1, x[ , (eliminate + 1):ncol(x), drop = FALSE]) else xNotElim <- matrix(1, nrow(x), 1) os.by.level <- numeric(eliminate) model <- suppressWarnings(glm.fit(xNotElim, y, weights = weights, offset = offset, family = family, control = glm.control(maxit = 1))) for (i in 1:nIter) { if (verbose) cat("quick.glm.fit iteration", i, "deviance =", deviance(model), "\n") w <- xElim * model$weights wz <- w * model$residuals os.by.level <- os.by.level + colSums(wz)/colSums(w) + coef(model)[1] os.vec <- offset + colSums(os.by.level * t(xElim)) model <- suppressWarnings(glm.fit(xNotElim, y, weights = weights, offset = os.vec, etastart = model$linear.predictors, family = family, control = glm.control(maxit = 2))) } structure(c(os.by.level + coef(model)[1], coef(model)[-1]), names = colnames(x)) } gnm/R/gnmFit.R0000755000176000001440000004406212515147523012651 0ustar ripleyusers# Copyright (C) 2005-2013 Heather Turner and David Firth # # This program is free software; you can redistribute it and/or modify # it under the terms of the GNU General Public License as published by # the Free Software Foundation; either version 2 or 3 of the License # (at your option). # # This program is distributed in the hope that it will be useful, # but WITHOUT ANY WARRANTY; without even the implied warranty of # MERCHANTABILITY or FITNESS FOR A PARTICULAR PURPOSE. See the # GNU General Public License for more details. # # A copy of the GNU General Public License is available at # http://www.r-project.org/Licenses/ gnmFit <- function (modelTools, y, constrain = numeric(0), # index of non-elimindated parameters constrainTo = numeric(length(constrain)), eliminate = NULL, # now a factor family = poisson(), weights = rep.int(1, length(y)), offset = rep.int(0, length(y)), nobs = length(y), start = rep.int(NA, length(modelTools$start) + nlevels(eliminate)), etastart = NULL, mustart = NULL, tolerance = 1e-6, iterStart = 2, iterMax = 500, trace = FALSE, verbose = FALSE, x = FALSE, termPredictors = FALSE, ridge = 1e-8) { names(y) <- NULL eps <- 100*.Machine$double.eps ridge <- 1 + ridge if (verbose) width <- as.numeric(options("width")) nTheta <- length(modelTools$start) nelim <- nlevels(eliminate) non.elim <- seq.int(nelim + 1, length(start)) ## add constraints specified by modelTools and glm tmpTheta <- as.double(rep.int(NA, nTheta)) varPredictors <- modelTools$varPredictors(tmpTheta) X <- modelTools$localDesignFunction(tmpTheta, varPredictors) isLinear <- unname(!is.na(colSums(X))) tmpTheta[constrain] <- constrainTo unspecified <- unname(is.na(tmpTheta)) if (any(isLinear & unspecified)) { tmpTheta[isLinear & unspecified] <- suppressWarnings(glm.fit.e(X[, isLinear & unspecified, drop = FALSE], y, family = family, intercept = FALSE, eliminate = if (nelim) eliminate else NULL, coefonly = TRUE, control = glm.control(maxit = 1))) extraLin <- which(isLinear & is.na(tmpTheta)) } else extraLin <- numeric() extra <- setdiff(c(modelTools$constrain, extraLin), constrain) ind <- order(c(constrain, extra)) constrain <- c(constrain, extra)[ind] constrainTo <- c(constrainTo, numeric(length(extra)))[ind] notConstrained <- !seq.int(nTheta) %in% constrain status <- "not.converged" unspecifiedNonlin <- FALSE dev <- numeric(2) if (nelim) { elim <- seq.int(nelim) alpha <- start[elim] } else { eliminate <- 1 alpha <- 0 } if (any(is.na(start))) { if (verbose == TRUE) prattle("Initialising", "\n", sep = "") ## only use start for elim par if all specified initElim <- any(is.na(alpha)) if (initElim) alpha[] <- numeric(nelim) theta <- start[non.elim] theta[is.na(theta)] <- modelTools$start[is.na(theta)] names(theta) <- names(modelTools$start) theta[constrain] <- constrainTo ## update any unspecified linear parameters unspecified <- unname(is.na(theta)) unspecifiedLin <- unspecified & isLinear unspecifiedNonlin <- unspecified & !isLinear if (!is.null(mustart)) etastart <- family$linkfun(mustart) if (any(unspecifiedNonlin) && is.null(etastart)){ theta[unspecifiedNonlin] <- gnmStart(sum(unspecifiedNonlin)) } if (any(unspecifiedLin) || initElim) { ## offset nonLin terms (currently NA if using etastart) ## plus offset contribution of any specified lin par if (!is.null(etastart)) z <- family$linkinv(etastart) else z <- y varPredictors <- modelTools$varPredictors(theta) tmpOffset <- modelTools$predictor(varPredictors, term = TRUE) tmpOffset <- rowSums(naToZero(tmpOffset)) tmpOffset <- offset + alpha[eliminate] + tmpOffset ## starting values for elim ignored here tmpTheta <- suppressWarnings({ glm.fit.e(X[, unspecifiedLin, drop = FALSE], z, weights = weights, etastart = etastart, offset = tmpOffset, family = family, intercept = FALSE, eliminate = if (nelim) eliminate else NULL, coefonly = TRUE)}) ## if no starting values for elim, use result of above if (initElim) alpha <- unname(attr(tmpTheta, "eliminated")) theta[unspecifiedLin] <- naToZero(tmpTheta) } if (any(unspecifiedNonlin) && !is.null(etastart)){ ## offset linear terms ## plus contribution of specified nonlin terms varPredictors <- modelTools$varPredictors(theta) tmpOffset <- modelTools$predictor(varPredictors, term = TRUE) tmpOffset <- rowSums(naToZero(tmpOffset)) tmpOffset <- offset + alpha[eliminate] + tmpOffset if (any(isLinear) && isTRUE(all.equal(unname(etastart), tmpOffset))) { etastart <- mustart <- NULL eval(family$initialize) etastart <- family$linkfun(mustart) } tmpOffset <- offset + alpha[eliminate] rss <- function(par) { theta[unspecifiedNonlin] <<- par varPredictors <<- modelTools$varPredictors(theta) eta <<- tmpOffset + modelTools$predictor(varPredictors) sum((etastart - eta)^2) } gr.rss <- function(par) { X <- modelTools$localDesignFunction(theta, varPredictors) -2 * t(X[, unspecifiedNonlin]) %*% ((etastart - eta)) } theta[unspecifiedNonlin] <- optim(gnmStart(sum(unspecifiedNonlin)), rss, gr.rss, method = c("L-BFGS-B"), control = list(maxit = iterStart), lower = -10, upper = 10)$par } varPredictors <- modelTools$varPredictors(theta) tmpOffset <- offset + alpha[eliminate] eta <- tmpOffset + modelTools$predictor(varPredictors) mu <- family$linkinv(eta) dev[1] <- sum(family$dev.resids(y, mu, weights)) if (trace) prattle("Initial Deviance = ", format(dev[1], nsmall = 6), "\n", sep = "") niter <- iterStart * (any(unspecifiedNonlin) && is.null(etastart)) for (iter in seq_len(niter)) { if (verbose) { if (iter == 1) prattle("Running start-up iterations", "\n"[trace], sep = "") if ((iter + 25)%%width == (width - 1)) cat("\n") } round <- 1 pmsh <- FALSE do <- seq_len(nTheta)[unspecifiedNonlin] maxDo <- max(do) for (i in rep.int(do, 2)) { dmu <- family$mu.eta(eta) vmu <- family$variance(mu) Xi <- modelTools$localDesignFunction(theta, varPredictors, i) wXi <- weights * (abs(dmu) >= eps) * dmu * dmu/vmu * Xi step <- sum((abs(y - mu) >= eps) * (y - mu)/dmu * wXi)/sum(wXi * Xi) otheta <- theta[i] theta[i] <- as.vector(otheta + step) if (!is.finite(theta[i])) { status <- "bad.param" break } varPredictors <- modelTools$varPredictors(theta) eta <- tmpOffset + modelTools$predictor(varPredictors) mu <- family$linkinv(eta) if (iter == 1 && (round == 1 || pmsh)) { dev[2] <- dev[1] dev[1] <- sum(family$dev.resids(y, mu, weights)) if (!is.finite(dev[1])) { status <- "bad.param" break } ## poor man's step-halving if (dev[1] > dev[2]) { pmsh <- TRUE theta[i] <- otheta + step/4 varPredictors <- modelTools$varPredictors(theta) eta <- tmpOffset + modelTools$predictor(varPredictors) mu <- family$linkinv(eta) dev[1] <- sum(family$dev.resids(y, mu, weights)) } } if (iter == 1 && i == maxDo) round <- 2 } if (status == "not.converged" && any(isLinear)) { if (iter == 1) { which <- which(isLinear & notConstrained) if(!exists("X")) X <- modelTools$localDesignFunction(theta, varPredictors) } tmpTheta <- updateLinear(which, theta, y, mu, eta, offset, weights, family, modelTools, X, if(nelim) eliminate else NULL) if (nelim){ alpha <- unname(attr(tmpTheta, "eliminated")) tmpOffset <- offset + alpha[eliminate] } theta[which] <- tmpTheta varPredictors <- modelTools$varPredictors(theta) eta <- tmpOffset + modelTools$predictor(varPredictors) mu <- family$linkinv(eta) } dev[1] <- sum(family$dev.resids(y, mu, weights)) if (!is.finite(dev[1])) { status <- "bad.param" break } if (trace) prattle("Start-up iteration ", iter, ". Deviance = ", format(dev[1], nsmall = 6), "\n", sep = "") else if (verbose) prattle(".") cat("\n"[iter == iterStart & verbose & !trace]) } } else { theta <- structure(replace(start[non.elim], constrain, constrainTo), names = names(modelTools$start)) varPredictors <- modelTools$varPredictors(theta) eta <- offset + alpha[eliminate] + modelTools$predictor(varPredictors) if (any(!is.finite(eta))) { stop("Values of 'start' and 'constrain' produce non-finite ", "predictor values") } mu <- family$linkinv(eta) dev[1] <- sum(family$dev.resids(y, mu, weights)) if (trace) prattle("Initial Deviance = ", format(dev[1], nsmall = 6), "\n", sep = "") } if (status == "not.converged") { X <- modelTools$localDesignFunction(theta, varPredictors) X <- X[, notConstrained, drop = FALSE] np <- ncol(X) + 1 ZWZ <- array(dim = c(np, np)) I1 <- numeric(np) I1[1] <- 1 if (nelim) Umat <- array(dim = c(nelim, np)) if (nelim){ grp.size <- tabulate(eliminate) grp.end <- cumsum(grp.size) } tmpAlpha <- 0 for (iter in seq_len(iterMax + 1)) { if (verbose) { if (iter == 1) prattle("Running main iterations", "\n"[trace], sep = "") if ((iter + 21)%%width == (width - 1)) cat("\n") } dmu <- family$mu.eta(eta) vmu <- family$variance(mu) w <- sqrt(weights * (abs(dmu) >= eps) * dmu * dmu/vmu) X <- w * X z <- w * (abs(dmu) >= eps) * (y - mu)/dmu ZWZ[-1,-1] <- crossprod(X) score <- ZWZ[1,-1] <- ZWZ[-1,1] <- crossprod(z, X) ZWZ[1,1] <- sum(z * z) diagInfo <- diag(ZWZ) ## only check for non-eliminated coefficients if (any(!is.finite(diagInfo))) { status <- "fail" break } if (all(diagInfo < 1e-20) || all(abs(score) < tolerance * sqrt(tolerance + diagInfo[-1]))) { status <- "converged" break } Zscales <- sqrt(diagInfo) Zscales[Zscales < 1e-3] <- 1e-3 ## to allow for zeros if (iter > iterMax) break if (nelim){ elimXscales <- grp.sum(w * w, grp.end) elimXscales <- sqrt(elimXscales * ridge) Umat[,1] <- rowsum.default(w * z, eliminate, reorder = FALSE) Umat[,-1] <- rowsum.default(w * X, eliminate, reorder = FALSE) Umat <- Umat/(elimXscales %o% Zscales) ZWZ <- ZWZ/(Zscales %o% Zscales) diag(ZWZ) <- ridge z <- solve(ZWZ - crossprod(Umat), I1, tol = .Machine$double.eps) thetaChange <- -z[-1]/z[1] * Zscales[1]/Zscales[-1] alphaChange <- c(Umat %*% (z * sqrt(ridge)))/z[1] * Zscales[1]/elimXscales } else { ZWZ <- ZWZ/(Zscales %o% Zscales) diag(ZWZ) <- ridge z <- solve(ZWZ, I1, tol = .Machine$double.eps)/Zscales thetaChange <- -z[-1]/z[1] } dev[2] <- dev[1] j <- scale <- 1 while (!is.nan(dev[1]) && dev[1] >= dev[2] && j < 11) { if (nelim) tmpAlpha <- alpha + alphaChange/scale tmpTheta <- replace(theta, notConstrained, theta[notConstrained] + thetaChange/scale) varPredictors <- modelTools$varPredictors(tmpTheta) eta <- offset + tmpAlpha[eliminate] + modelTools$predictor(varPredictors) mu <- family$linkinv(eta) dev[1] <- sum(family$dev.resids(y, mu, weights)) scale <- scale*2 j <- j + 1 } if (!is.finite(dev[1])) { status <- "no.deviance" break } if (trace){ prattle("Iteration ", iter, ". Deviance = ", format(dev[1], nsmall = 6), "\n", sep = "") } else if (verbose) prattle(".") if (nelim) alpha <- tmpAlpha theta <- tmpTheta X <- modelTools$localDesignFunction(theta, varPredictors) X <- X[, notConstrained, drop = FALSE] } } if (status %in% c("converged", "not.converged")) { if (verbose) prattle("\n"[!trace], "Done\n", sep = "") } else { if (any(!is.finite(eta))) status <- "eta.not.finite" if (exists("w") && any(!is.finite(w))) status <- "w.not.finite" if (any(is.infinite(X))) status <- "X.not.finite" if (verbose) message("\n"[!trace], switch(status, bad.param = "Bad parameterisation", eta.not.finite = "Predictors are not all finite", w.not.finite = "Iterative weights are not all finite", X.not.finite = "Local design matrix has infinite elements", no.deviance = "Deviance is not finite")) return() } theta[constrain] <- NA X <- modelTools$localDesignFunction(theta, varPredictors) X <- X[, notConstrained, drop = FALSE] ## suppress warnings in rankMatrix re coercion to dense matrix if (nelim) { ## sweeps needed to get the rank right subtracted <- rowsum.default(X, eliminate, reorder = FALSE)/grp.size if (modelTools$termAssign[1] == 0) subtracted[,1] <- 0 theRank <- suppressWarnings( rankMatrix(X - subtracted[eliminate, , drop = FALSE])) + nelim names(alpha) <- paste("(eliminate)", elim, sep = "") } else theRank <- suppressWarnings(rankMatrix(X)) modelAIC <- suppressWarnings(family$aic(y, rep.int(1, nobs), mu, weights, dev[1]) + 2 * theRank) fit <- list(coefficients = structure(theta, eliminated = alpha), constrain = constrain, constrainTo = constrainTo, residuals = z/w, fitted.values = mu, rank = theRank, family = family, predictors = eta, deviance = dev[1], aic = modelAIC, iter = iter - (iter != iterMax), weights = w * w, prior.weights = weights, df.residual = c(nobs - theRank), y = y) if (status == "not.converged") { warning("Fitting algorithm has either not converged or converged\n", "to a non-solution of the likelihood equations.\n", "Use exitInfo() for numerical details of last iteration.\n") fit$converged <- structure(FALSE, score = score, criterion = tolerance * sqrt(tolerance + diagInfo[-1])) } else fit$converged <- TRUE if (x) { X <- modelTools$localDesignFunction(theta, varPredictors) fit$x <- structure(X, assign = modelTools$termAssign) } if (termPredictors) { theta[is.na(theta)] <- 0 varPredictors <- modelTools$varPredictors(theta) fit$termPredictors <- modelTools$predictor(varPredictors, term = TRUE) } fit } gnm/R/termPredictors.gnm.R0000644000176000001440000000267612515147523015215 0ustar ripleyusers# Copyright (C) 2005, 2008, 2010, 2012 Heather Turner # # This program is free software; you can redistribute it and/or modify # it under the terms of the GNU General Public License as published by # the Free Software Foundation; either version 2 or 3 of the License # (at your option). # # This program is distributed in the hope that it will be useful, # but WITHOUT ANY WARRANTY; without even the implied warranty of # MERCHANTABILITY or FITNESS FOR A PARTICULAR PURPOSE. See the # GNU General Public License for more details. # # A copy of the GNU General Public License is available at # http://www.r-project.org/Licenses/ termPredictors.gnm <- function(object, ...) { if (is.null(object$termPredictors)){ modelData <- model.frame(object) modelTerms <- terms(object) if (!is.empty.model(modelTerms)) { modelTools <- gnmTools(modelTerms, modelData) theta <- parameters(object) varPredictors <- modelTools$varPredictors(theta) termPredictors <- modelTools$predictor(varPredictors, term = TRUE) rownames(termPredictors) <- rownames(modelData) } else termPredictors <- modelData[,0] if (!is.null(object$eliminate)) termPredictors <- cbind("(eliminate)" = as.vector(attr(coef(object), "eliminated")[object$eliminate]), termPredictors) termPredictors } else object$termPredictors } gnm/R/weighted.MM.R0000644000176000001440000000214612515147523013527 0ustar ripleyusers# Copyright (C) 2006 Heather Turner # # This program is free software; you can redistribute it and/or modify # it under the terms of the GNU General Public License as published by # the Free Software Foundation; either version 2 or 3 of the License # (at your option). # # This program is distributed in the hope that it will be useful, # but WITHOUT ANY WARRANTY; without even the implied warranty of # MERCHANTABILITY or FITNESS FOR A PARTICULAR PURPOSE. See the # GNU General Public License for more details. # # A copy of the GNU General Public License is available at # http://www.r-project.org/Licenses/ weighted.MM <- function(resp, conc){ list(predictors = list(Vm = substitute(conc), K = 1), variables = list(substitute(resp), substitute(conc)), term = function(predLabels, varLabels) { pred <- paste("(", predLabels[1], "/(", predLabels[2], " + ", varLabels[2], "))", sep = "") pred <- paste("(", varLabels[1], " - ", pred, ")/sqrt(", pred, ")", sep = "") }) } class(weighted.MM) <- "nonlin" gnm/R/prattle.R0000644000176000001440000000125012515147523013065 0ustar ripleyusers# Copyright (C) 2005 David Firth # # This program is free software; you can redistribute it and/or modify # it under the terms of the GNU General Public License as published by # the Free Software Foundation; either version 2 or 3 of the License # (at your option). # # This program is distributed in the hope that it will be useful, # but WITHOUT ANY WARRANTY; without even the implied warranty of # MERCHANTABILITY or FITNESS FOR A PARTICULAR PURPOSE. See the # GNU General Public License for more details. # # A copy of the GNU General Public License is available at # http://www.r-project.org/Licenses/ prattle <- function(...) { cat(...) flush.console() } gnm/R/updateLinear.R0000644000176000001440000000250712515147523014035 0ustar ripleyusers# Copyright (C) 2005, 2006, 2010 Heather Turner # # This program is free software; you can redistribute it and/or modify # it under the terms of the GNU General Public License as published by # the Free Software Foundation; either version 2 or 3 of the License # (at your option). # # This program is distributed in the hope that it will be useful, # but WITHOUT ANY WARRANTY; without even the implied warranty of # MERCHANTABILITY or FITNESS FOR A PARTICULAR PURPOSE. See the # GNU General Public License for more details. # # A copy of the GNU General Public License is available at # http://www.r-project.org/Licenses/ updateLinear <- function(which, theta, y, mu, eta, offset, weights, family, modelTools, X, eliminate) { dmu <- family$mu.eta(eta) vmu <- family$variance(mu) w <- weights * dmu * dmu / vmu theta[which] <- 0 offsetVarPredictors <- modelTools$varPredictors(theta) offset <- offset + modelTools$predictor(offsetVarPredictors) z <- eta - offset + (y - mu)/dmu if (is.null(eliminate)) naToZero(lm.wfit(X[,which, drop = FALSE], z, w)$coef) else suppressWarnings(glm.fit.e(X[,which, drop = FALSE], z, weights = w, intercept = FALSE, eliminate = eliminate, coefonly = TRUE)) } gnm/R/termPredictors.default.R0000644000176000001440000000210512515147523016043 0ustar ripleyusers# Copyright (C) 2005 Heather Turner # # This program is free software; you can redistribute it and/or modify # it under the terms of the GNU General Public License as published by # the Free Software Foundation; either version 2 or 3 of the License # (at your option). # # This program is distributed in the hope that it will be useful, # but WITHOUT ANY WARRANTY; without even the implied warranty of # MERCHANTABILITY or FITNESS FOR A PARTICULAR PURPOSE. See the # GNU General Public License for more details. # # A copy of the GNU General Public License is available at # http://www.r-project.org/Licenses/ termPredictors.default <- function(object, ...) { if (is.null(object$termPredictors)){ X <- model.matrix(object) termPredictors <- t(rowsum(t(X %*% diag(naToZero(coef(object)))), attr(X, "assign"))) colnames(termPredictors) <- c("(Intercept)"[0 %in% attr(X, "assign")], attr(object$terms, "term.labels")) termPredictors } else object$termPredictors } gnm/R/asGnm.glm.R0000644000176000001440000000306712515147523013245 0ustar ripleyusers# Copyright (C) 2006, 2008, 2010 Heather Turner # # This program is free software; you can redistribute it and/or modify # it under the terms of the GNU General Public License as published by # the Free Software Foundation; either version 2 or 3 of the License # (at your option). # # This program is distributed in the hope that it will be useful, # but WITHOUT ANY WARRANTY; without even the implied warranty of # MERCHANTABILITY or FITNESS FOR A PARTICULAR PURPOSE. See the # GNU General Public License for more details. # # A copy of the GNU General Public License is available at # http://www.r-project.org/Licenses/ asGnm.glm <- function(object, ...) { glmExtra <- match(c("effects", "R", "qr", "null.deviance", "df.null", "boundary", "control", "contrasts"), names(object)) modelData <- model.frame(object) object[glmExtra] <- NULL object$call[[1]] <- as.name("gnm") constrain <- which(is.na(coef(object))) object$terms <- gnmTerms(object$formula, data = modelData) object <- c(list(eliminate = NULL, ofInterest = NULL, na.action = na.action(modelData), constrain = constrain, constrainTo = numeric(length(constrain))), object) names(object)[match("linear.predictors", names(object))] <- "predictors" if (is.null(object$offset)) object$offset <- rep.int(0, length(coef(object))) object$tolerance <- object$iterStart <- object$iterMax <- "Not available - model fitted by glm()" class(object) <- c("gnm", "glm", "lm") object } gnm/R/Const.R0000644000176000001440000000160012515147523012477 0ustar ripleyusers# Copyright (C) 2006, 2008 Heather Turner # # This program is free software; you can redistribute it and/or modify # it under the terms of the GNU General Public License as published by # the Free Software Foundation; either version 2 or 3 of the License # (at your option). # # This program is distributed in the hope that it will be useful, # but WITHOUT ANY WARRANTY; without even the implied warranty of # MERCHANTABILITY or FITNESS FOR A PARTICULAR PURPOSE. See the # GNU General Public License for more details. # # A copy of the GNU General Public License is available at # http://www.r-project.org/Licenses/ Const <- function(const) { badCall <- !"nonlinTerms" %in% lapply(sys.calls(), "[[", 1) if (any(badCall)) stop("Const terms are only valid in the predictors of \"nonlin\" ", "functions.") call("rep", substitute(const), quote(nObs)) } gnm/R/Logit.R0000644000176000001440000000167212515147523012500 0ustar ripleyusers# Copyright (C) 2006, 2008 Heather Turner # # This program is free software; you can redistribute it and/or modify # it under the terms of the GNU General Public License as published by # the Free Software Foundation; either version 2 or 3 of the License # (at your option). # # This program is distributed in the hope that it will be useful, # but WITHOUT ANY WARRANTY; without even the implied warranty of # MERCHANTABILITY or FITNESS FOR A PARTICULAR PURPOSE. See the # GNU General Public License for more details. # # A copy of the GNU General Public License is available at # http://www.r-project.org/Licenses/ Logit <- function(expression, inst = NULL){ list(predictors = list(substitute(expression)), term = function(predLabels, ...) { paste("log((", predLabels, ")/(1 - (", predLabels, ")))", sep = "") }, call = as.expression(match.call()), match = 1) } class(Logit) <- "nonlin" gnm/R/logtrans.gnm.R0000644000176000001440000000143612515147523014031 0ustar ripleyusers# Copyright (C) 2005 Heather Turner # # This program is free software; you can redistribute it and/or modify # it under the terms of the GNU General Public License as published by # the Free Software Foundation; either version 2 or 3 of the License # (at your option). # # This program is distributed in the hope that it will be useful, # but WITHOUT ANY WARRANTY; without even the implied warranty of # MERCHANTABILITY or FITNESS FOR A PARTICULAR PURPOSE. See the # GNU General Public License for more details. # # A copy of the GNU General Public License is available at # http://www.r-project.org/Licenses/ logtrans.gnm <- function (object, ...) { if (inherits(object, "gnm", TRUE) == 1) stop("logtrans is not implemented for gnm objects") else NextMethod } gnm/R/proj.gnm.R0000644000176000001440000000142612515147523013151 0ustar ripleyusers# Copyright (C) 2005 Heather Turner # # This program is free software; you can redistribute it and/or modify # it under the terms of the GNU General Public License as published by # the Free Software Foundation; either version 2 or 3 of the License # (at your option). # # This program is distributed in the hope that it will be useful, # but WITHOUT ANY WARRANTY; without even the implied warranty of # MERCHANTABILITY or FITNESS FOR A PARTICULAR PURPOSE. See the # GNU General Public License for more details. # # A copy of the GNU General Public License is available at # http://www.r-project.org/Licenses/ proj.gnm <- function (object, ...) { if (inherits(object, "gnm", TRUE) == 1) stop("proj is not implemented for gnm objects") else NextMethod } gnm/R/drop1.gnm.R0000644000176000001440000000767212515147523013235 0ustar ripleyusers# Modification of drop1.glm from the stats package for R. # # Copyright (C) 1994-8 W. N. Venables and B. D. Ripley # Copyright (C) 1998-2005 The R Core Team # Copyright (C) 2005, 2010, 2013 Heather Turner # # This program is free software; you can redistribute it and/or modify # it under the terms of the GNU General Public License as published by # the Free Software Foundation; either version 2 or 3 of the License # (at your option). # # This program is distributed in the hope that it will be useful, # but WITHOUT ANY WARRANTY; without even the implied warranty of # MERCHANTABILITY or FITNESS FOR A PARTICULAR PURPOSE. See the # GNU General Public License for more details. # # A copy of the GNU General Public License is available at # http://www.r-project.org/Licenses/ drop1.gnm <- function (object, scope, scale = 0, test = c("none", "Chisq", "F"), k = 2, ...) { if (any(attr(terms(object), "type") != "Linear")) stop("add1 is not implemented for gnm objects with nonlinear terms.") x <- model.matrix(object) n <- nrow(x) asgn <- attr(x, "assign") tl <- attr(object$terms, "term.labels") if (missing(scope)) scope <- drop.scope(object) else { if (!is.character(scope)) scope <- attr(terms(update.formula(object, scope)), "term.labels") if (!all(match(scope, tl, 0L) > 0L)) stop("scope is not a subset of term labels") } ndrop <- match(scope, tl) ns <- length(scope) rdf <- object$df.residual chisq <- object$deviance dfs <- numeric(ns) dev <- numeric(ns) y <- object$y if (is.null(y)) { y <- model.response(model.frame(object)) if (!is.factor(y)) storage.mode(y) <- "double" } wt <- object$prior.weights if (is.null(wt)) wt <- rep.int(1, n) for (i in 1L:ns) { ii <- seq_along(asgn)[asgn == ndrop[i]] jj <- setdiff(seq(ncol(x)), ii) z <- glm.fit.e(x[, jj, drop = FALSE], y, wt, offset = object$offset, family = object$family, eliminate = object$eliminate) dfs[i] <- z$rank dev[i] <- z$deviance } scope <- c("", scope) dfs <- c(object$rank, dfs) dev <- c(chisq, dev) dispersion <- if (is.null(scale) || scale == 0) summary(object, dispersion = NULL)$dispersion else scale fam <- object$family$family loglik <- if (fam == "gaussian") { if (scale > 0) dev/scale - n else n * log(dev/n) } else dev/dispersion aic <- loglik + k * dfs dfs <- dfs[1L] - dfs dfs[1L] <- NA aic <- aic + (extractAIC(object, k = k)[2L] - aic[1L]) aod <- data.frame(Df = dfs, Deviance = dev, AIC = aic, row.names = scope, check.names = FALSE) if (all(is.na(aic))) aod <- aod[, -3] test <- match.arg(test) if (test == "Chisq") { dev <- pmax(0, loglik - loglik[1L]) dev[1L] <- NA nas <- !is.na(dev) LRT <- if (dispersion == 1) "LRT" else "scaled dev." aod[, LRT] <- dev dev[nas] <- pchisq(dev[nas], aod$Df[nas], lower.tail = FALSE) aod[, "Pr(Chi)"] <- dev } else if (test == "F") { if (fam == "binomial" || fam == "poisson") warning(gettextf("F test assumes 'quasi%s' family", fam), domain = NA) dev <- aod$Deviance rms <- dev[1L]/rdf dev <- pmax(0, dev - dev[1L]) dfs <- aod$Df rdf <- object$df.residual Fs <- (dev/dfs)/rms Fs[dfs < 1e-04] <- NA P <- Fs nas <- !is.na(Fs) P[nas] <- pf(Fs[nas], dfs[nas], rdf, lower.tail = FALSE) aod[, c("F value", "Pr(F)")] <- list(Fs, P) } head <- c("Single term deletions", "\nModel:", deparse(as.vector(formula(object))), if (!is.null(scale) && scale > 0) paste("\nscale: ", format(scale), "\n")) class(aod) <- c("anova", "data.frame") attr(aod, "heading") <- head aod } gnm/R/hatvalues.gnm.R0000755000176000001440000000263312515147523014177 0ustar ripleyusers# Copyright (C) 2005, 2006, 2008, 2010, 2012 Heather Turner # # This program is free software; you can redistribute it and/or modify # it under the terms of the GNU General Public License as published by # the Free Software Foundation; either version 2 or 3 of the License # (at your option). # # This program is distributed in the hope that it will be useful, # but WITHOUT ANY WARRANTY; without even the implied warranty of # MERCHANTABILITY or FITNESS FOR A PARTICULAR PURPOSE. See the # GNU General Public License for more details. # # A copy of the GNU General Public License is available at # http://www.r-project.org/Licenses/ hatvalues.gnm <- function(model, ...) { X <- as(model.matrix(model), "sparseMatrix") var <- unclass(vcov(model, with.eliminate = TRUE)) eliminate <- model$eliminate scale <- model$weights/attr(var, "dispersion") hat <- rowSums((X %*% var) * X) * scale if (!is.null(eliminate)) { ## no covElim! if (length(model$constrain)) X <- X[, -model$constrain, drop = FALSE] hat <- hat + (2 * rowSums(X * attr(var, "covElim")[eliminate, , drop = FALSE]) + attr(var, "varElim")[eliminate]) * scale } hat <- naresid(model$na.action, hat) hat[is.na(hat)] <- 0 hat[hat > 1 - 100 * .Machine$double.eps] <- 1 if (!is.null(model$table.attr)) attributes(hat) <- model$table.attr hat } gnm/R/print.gnm.R0000755000176000001440000000313012515147523013330 0ustar ripleyusers# Copyright (C) 2005-2008, 2010 Heather Turner and David Firth # # This program is free software; you can redistribute it and/or modify # it under the terms of the GNU General Public License as published by # the Free Software Foundation; either version 2 or 3 of the License # (at your option). # # This program is distributed in the hope that it will be useful, # but WITHOUT ANY WARRANTY; without even the implied warranty of # MERCHANTABILITY or FITNESS FOR A PARTICULAR PURPOSE. See the # GNU General Public License for more details. # # A copy of the GNU General Public License is available at # http://www.r-project.org/Licenses/ print.gnm <- function (x, digits = max(3, getOption("digits") - 3), ...) { cat("\nCall:\n", deparse(x$call), "\n", sep = "", fill = TRUE) if (length(coef(x)) && (is.null(ofInterest(x)) || length(ofInterest(x)))) { cat("Coefficients", " of interest"[!is.null(ofInterest(x))], ":\n", sep = "") if (!is.null(ofInterest(x))) print.default(format(coef(x)[ofInterest(x)], digits = digits), print.gap = 2, quote = FALSE) else print.default(format(coef(x), digits = digits), print.gap = 2, quote = FALSE) } else cat("No coefficients", " of interest"[!is.null(ofInterest(x))], ". \n\n", sep = "") cat("\nDeviance: ", format(x$deviance, digits), "\nPearson chi-squared:", format(sum(na.omit(c(residuals(x, type = "pearson")))^2), digits), "\nResidual df: ", x$df.residual, "\n") invisible(x) } gnm/R/Diag.R0000644000176000001440000000242412515147523012262 0ustar ripleyusers# Copyright (C) 2005, 2008 David Firth and Heather Turner # # This program is free software; you can redistribute it and/or modify # it under the terms of the GNU General Public License as published by # the Free Software Foundation; either version 2 or 3 of the License # (at your option). # # This program is distributed in the hope that it will be useful, # but WITHOUT ANY WARRANTY; without even the implied warranty of # MERCHANTABILITY or FITNESS FOR A PARTICULAR PURPOSE. See the # GNU General Public License for more details. # # A copy of the GNU General Public License is available at # http://www.r-project.org/Licenses/ Diag <- function(..., binary = FALSE){ dots <- list(...) dots <- lapply(dots, as.factor) Levels <- sort(unique(unlist(lapply(dots, levels)))) facMatrix <- sapply(dots, as.character) f <- function(row){ if (all(is.na(row))) return(NA) if (all(!is.na(row)) && all(row == row[1])) return(row[1]) row <- na.omit(row) if (!all(row == row[1])) return(".") return(NA) } if (inherits(facMatrix, "matrix")) result <- factor(apply(facMatrix, 1, f), levels = c(".", Levels)) else result <- factor(f(facMatrix)) if (binary) result <- ifelse(result == ".", 0, 1) result } gnm/R/rstudent.gnm.R0000644000176000001440000000143412515147523014046 0ustar ripleyusers# Copyright (C) 2005 Heather Turner # # This program is free software; you can redistribute it and/or modify # it under the terms of the GNU General Public License as published by # the Free Software Foundation; either version 2 or 3 of the License # (at your option). # # This program is distributed in the hope that it will be useful, # but WITHOUT ANY WARRANTY; without even the implied warranty of # MERCHANTABILITY or FITNESS FOR A PARTICULAR PURPOSE. See the # GNU General Public License for more details. # # A copy of the GNU General Public License is available at # http://www.r-project.org/Licenses/ rstudent.gnm <- function (model, ...) { if (inherits(model, "gnm", TRUE) == 1) stop("rstudent is not implemented for gnm objects") else NextMethod } gnm/R/termPredictors.R0000644000176000001440000000127312515147523014425 0ustar ripleyusers# Copyright (C) 2005 Heather Turner # # This program is free software; you can redistribute it and/or modify # it under the terms of the GNU General Public License as published by # the Free Software Foundation; either version 2 or 3 of the License # (at your option). # # This program is distributed in the hope that it will be useful, # but WITHOUT ANY WARRANTY; without even the implied warranty of # MERCHANTABILITY or FITNESS FOR A PARTICULAR PURPOSE. See the # GNU General Public License for more details. # # A copy of the GNU General Public License is available at # http://www.r-project.org/Licenses/ termPredictors <- function(object, ...) { UseMethod("termPredictors") } gnm/R/cholInv.R0000644000176000001440000000602012515147523013014 0ustar ripleyusers# Copyright (C) 2006, 2010 David Firth and Heather Turner # # This program is free software; you can redistribute it and/or modify # it under the terms of the GNU General Public License as published by # the Free Software Foundation; either version 2 or 3 of the License # (at your option). # # This program is distributed in the hope that it will be useful, # but WITHOUT ANY WARRANTY; without even the implied warranty of # MERCHANTABILITY or FITNESS FOR A PARTICULAR PURPOSE. See the # GNU General Public License for more details. # # A copy of the GNU General Public License is available at # http://www.r-project.org/Licenses/ cholInv <- function (mat, eliminate = numeric(0), onlyFirstCol = FALSE, onlyNonElim = FALSE) { .Deprecated(msg = paste("'cholInv' is deprecated as it is no longer used ", "by gnm.")) m <- nrow(mat) n <- ncol(mat) if (length(eliminate) == 0) { ## the basic routine, no eliminated submatrix if (!is.matrix(mat)) stop("mat is not a matrix") Rownames <- rownames(mat) Colnames <- colnames(mat) result <- chol2inv(chol(mat)) if (!is.null(Rownames)) colnames(result) <- Rownames if (!is.null(Colnames)) rownames(result) <- Colnames if (onlyFirstCol) result <- result[, 1, drop = FALSE] return(result) } ## Now allow for the possibility of an eliminated submatrix if (m != n) stop("mat must be a symmetric matrix") n <- nrow(mat) elim <- 1:n %in% eliminate diag.indices <- (n * (0:(n - 1)) + 1:n) T <- mat[diag.indices[eliminate]] if (any(T == 0)) stop("an eliminated submatrix must have all diagonal entries non-zero.") W <- mat[!elim, !elim, drop = FALSE] U <- mat[elim, !elim, drop = FALSE] Ti <- 1/T k <- length(T) Ti.U <- Ti * U V.Ti <- t(Ti.U) Qmat <- W - crossprod(Ti.U, U) Qi <- cholInv(Qmat) result <- matrix(NA, if (onlyNonElim) n - k else n, if (onlyFirstCol) 1 else if (onlyNonElim) n - k else n) cols.notElim <- if (onlyFirstCol) 1 else if (onlyNonElim) 1:(n - k) else !elim rows.notElim <- if (onlyNonElim) 1:(n - k) else !elim if (onlyFirstCol) Qi <- Qi[, 1, drop = FALSE] result[rows.notElim, cols.notElim] <- Qi if (!onlyNonElim) { temp <- -crossprod(Qi, V.Ti) result[elim, cols.notElim] <- t(temp) } if (!onlyFirstCol && !onlyNonElim) { result[!elim, elim] <- temp temp <- crossprod(V.Ti, Qi) %*% V.Ti diag.indices <- k * (0:(k - 1)) + 1:k temp[diag.indices] <- Ti + temp[diag.indices] result[elim, elim] <- temp } theNames <- colnames(mat) rownames(result) <- if (onlyNonElim) theNames[!elim] else theNames colnames(result) <- if (onlyFirstCol) theNames[!elim][1] else if (onlyNonElim) theNames[!elim] else theNames result } gnm/R/meanResiduals.R0000644000176000001440000000654412515147523014221 0ustar ripleyusers# Copyright (C) 2010, 2012, 2013 Heather Turner # # This program is free software; you can redistribute it and/or modify # it under the terms of the GNU General Public License as published by # the Free Software Foundation; either version 2 or 3 of the License # (at your option). # # This program is distributed in the hope that it will be useful, # but WITHOUT ANY WARRANTY; without even the implied warranty of # MERCHANTABILITY or FITNESS FOR A PARTICULAR PURPOSE. See the # GNU General Public License for more details. # # A copy of the GNU General Public License is available at # http://www.r-project.org/Licenses/ meanResiduals <- function(object, by = NULL, standardized = TRUE, as.table = TRUE, ...){ if (is.null(by)) stop("`by' must be specified in order to compute grouped residuals") if (inherits(by, "formula")){ ## check single factor only if (ncol(attr(terms(by), "factors")) != 1) stop("`by' should only specify a single term") ## find factors as in mosaic.glm (own code) by <- do.call("model.frame", list(formula = by, data = object$data, subset = object$call$subset, na.action = na.pass, drop.unused.levels = TRUE)) ## following loop needed due to bug in model.frame.default (fixed for R 2.12) for(nm in names(by)) { f <- by[[nm]] if(is.factor(f) && length(unique(f[!is.na(f)])) < length(levels(f))) by[[nm]] <- by[[nm]][, drop = TRUE] } if (!is.null(object$na.action)) by <- by[-object$na.action,] } if (!all(sapply(by, is.factor))) warning("Coercing variables specified by `by' to factors") fac <- factor(interaction(by)) # drop unused levels if (length(fac) != length(object$y)) stop("Grouping factor of length", length(fac), "but model frame of length", length(object$y)) r <- object$residuals ## recompute weights for better accuracy w <- as.numeric(object$prior.weights * object$family$mu.eta(predict(object, type = "link"))^2/ object$family$variance(object$fitted)) agg.wts <- tapply(w, by, sum) #unlike rowsum, keeps all levels of interaction res <- tapply(r * w, by, sum)/agg.wts if (standardized) res <- res * sqrt(agg.wts) ## now compute degrees of freedom Xreduced <- rowsum(model.matrix(object), fac, na.rm = TRUE) ## suppressWarnings in rankMatrix re coercion to dense matrix if (as.table){ res <- structure(as.table(res), call = object$call, by = paste(names(by), collapse = ":"), df = nlevels(fac) - suppressWarnings(rankMatrix(Xreduced)), standardized = standardized, weights = as.table(agg.wts)) class(res) <- c("meanResiduals", "table") } else { res <- structure(c(res), call = object$call, by = paste(names(by), collapse = ":"), df = nlevels(fac) - suppressWarnings(rankMatrix(Xreduced)), standardized = standardized, weights = c(agg.wts)) class(res) <- c("meanResiduals", "numeric") } return(res) } gnm/R/plot.profile.gnm.R0000644000176000001440000000270612515147523014616 0ustar ripleyusers# Modification of plot.profile from the stats package for R. # # File MASS/profiles.q copyright (C) 1996 D. M. Bates and W. N. Venables. # port to R by B. D. Ripley copyright (C) 1998 # corrections copyright (C) 2000,3,6 B. D. Ripley # Copyright (C) 2006 Heather Turner # # This program is free software; you can redistribute it and/or modify # it under the terms of the GNU General Public License as published by # the Free Software Foundation; either version 2 or 3 of the License # (at your option). # # This program is distributed in the hope that it will be useful, # but WITHOUT ANY WARRANTY; without even the implied warranty of # MERCHANTABILITY or FITNESS FOR A PARTICULAR PURPOSE. See the # GNU General Public License for more details. # # A copy of the GNU General Public License is available at # http://www.r-project.org/Licenses/ plot.profile.gnm <- function (x, nseg, ...) { nulls <- sapply(x, is.null) if (all(nulls)) return(NULL) x <- x[!nulls] pnames <- names(x) pnames <- pnames[!is.na(x[pnames])] nr <- ceiling(sqrt(length(pnames))) oldpar <- par(mfrow = c(nr, nr)) on.exit(par(oldpar)) for (nm in pnames) { z <- x[[nm]][[1]] parval <- x[[nm]][[2]][, nm] plot(parval, z, xlab = nm, ylab = "z", type = "n") if (sum(z == 0) == 1) points(parval[z == 0], 0, pch = 3) splineVals <- spline(parval, z) lines(splineVals$x, splineVals$y) } } gnm/R/Mult.R0000755000176000001440000000216212515147523012341 0ustar ripleyusers# Copyright (C) 2005, 2006, 2008 Heather Turner # # This program is free software; you can redistribute it and/or modify # it under the terms of the GNU General Public License as published by # the Free Software Foundation; either version 2 or 3 of the License # (at your option). # # This program is distributed in the hope that it will be useful, # but WITHOUT ANY WARRANTY; without even the implied warranty of # MERCHANTABILITY or FITNESS FOR A PARTICULAR PURPOSE. See the # GNU General Public License for more details. # # A copy of the GNU General Public License is available at # http://www.r-project.org/Licenses/ Mult <- function(..., inst = NULL){ if ("multiplicity" %in% names(match.call()[-1])) stop("multiplicity argument of Mult has been replaced by", "\"inst\" argument.") dots <- match.call(expand.dots = FALSE)[["..."]] list(predictors = dots, term = function(predLabels, ...) { paste("(", paste(predLabels, collapse = ")*("), ")", sep = "") }, call = as.expression(match.call()), match = seq(dots)) } class(Mult) <- "nonlin" gnm/R/variable.names.gnm.R0000755000176000001440000000152512515147523015071 0ustar ripleyusers# Copyright (C) 2005, 2006 Heather Turner # # This program is free software; you can redistribute it and/or modify # it under the terms of the GNU General Public License as published by # the Free Software Foundation; either version 2 or 3 of the License # (at your option). # # This program is distributed in the hope that it will be useful, # but WITHOUT ANY WARRANTY; without even the implied warranty of # MERCHANTABILITY or FITNESS FOR A PARTICULAR PURPOSE. See the # GNU General Public License for more details. # # A copy of the GNU General Public License is available at # http://www.r-project.org/Licenses/ variable.names.gnm <- function(object, full = FALSE, ...) { if (full) names(coef(object)) else { setToZero <- object$constrain[object$constrainTo == 0] names(coef(object)[-setToZero]) } } gnm/R/DrefWeights.R0000644000176000001440000000424112515147523013630 0ustar ripleyusers# Copyright (C) 2007 Heather Turner # # This program is free software; you can redistribute it and/or modify # it under the terms of the GNU General Public License as published by # the Free Software Foundation; either version 2 or 3 of the License # (at your option). # # This program is distributed in the hope that it will be useful, # but WITHOUT ANY WARRANTY; without even the implied warranty of # MERCHANTABILITY or FITNESS FOR A PARTICULAR PURPOSE. See the # GNU General Public License for more details. # # A copy of the GNU General Public License is available at # http://www.r-project.org/Licenses/ DrefWeights <- function(model) { ind <- pickCoef(model, "delta[1-9]") if (any(!checkEstimable(model, diag(seq(along = coef(model)))[,ind]), na.rm = TRUE)){ message("Refitting with parameters of first Dref weight constrained ", "to zero") constrain <- pickCoef(model, "delta1") model <- update(model, constrain = constrain, start = coef(model), trace = FALSE, verbose = FALSE) } t <- terms(formula(model), specials = "Dref") DrefCall <- attr(t, "variables")[[attr(t, "specials")$Dref + 1]] preds <- match.call(Dref, DrefCall, expand.dots = FALSE)[["..."]] formula <- as.formula(DrefCall$delta) if (length(formula)) { dat <- model.frame(formula, data = model.frame(model)) X <- unique(model.matrix(formula, data = dat)) dat <- dat[rownames(X), , drop = FALSE] rownames(dat) <- rownames(X) <- NULL } else { dat <- numeric(0) X <- matrix(1) } nw <- length(preds) nmod <- nrow(X) delta <- matrix(parameters(model)[ind], nmod) ind <- c(t(matrix(ind, nmod, nw))) vcovDelta <- vcov(model)[ind, ind, drop = FALSE] wc <- 1/rowSums(exp(X %*% delta)) wu <- exp(X %*% delta)*wc XX <- matrix(apply(X, 2, rep, nw), nrow(X)) out <- list() for (i in 1:nw) { d <- -wu[,i] * wu d <- c(wu[,i] * (col(wu) == i) + d) * XX se <- sqrt(rowSums(d %*% vcovDelta * d)) out[[i]] <- drop(cbind(dat, weight = wu[,i], se = se)) } names(out) <- as.character(preds) out } gnm/R/print.profile.gnm.R0000644000176000001440000000213612515147523014771 0ustar ripleyusers# Copyright (C) 2006, 2008, 2009 Heather Turner # # This program is free software; you can redistribute it and/or modify # it under the terms of the GNU General Public License as published by # the Free Software Foundation; either version 2 or 3 of the License # (at your option). # # This program is distributed in the hope that it will be useful, # but WITHOUT ANY WARRANTY; without even the implied warranty of # MERCHANTABILITY or FITNESS FOR A PARTICULAR PURPOSE. See the # GNU General Public License for more details. # # A copy of the GNU General Public License is available at # http://www.r-project.org/Licenses/ print.profile.gnm <- function (x, digits = max(3, getOption("digits") - 3), ...) { #if (attr(x, "eliminate")) # coefs <- coefs[-seq(attr(x$cov.scaled, "eliminate")), ] if (length(x)) { if (any(sapply(x, function(x) isTRUE(is.na(x))))) cat("\nProfile is NA where coefficient has been constrained or", "is unidentified\n\n") print.default(x) } else cat("\nNo coefficients profiled.\n\n", sep = "") invisible(x) } gnm/R/rstandard.gnm.R0000755000176000001440000000173412515147523014166 0ustar ripleyusers# Copyright (C) 2005, 2006, 2008 Heather Turner # # This program is free software; you can redistribute it and/or modify # it under the terms of the GNU General Public License as published by # the Free Software Foundation; either version 2 or 3 of the License # (at your option). # # This program is distributed in the hope that it will be useful, # but WITHOUT ANY WARRANTY; without even the implied warranty of # MERCHANTABILITY or FITNESS FOR A PARTICULAR PURPOSE. See the # GNU General Public License for more details. # # A copy of the GNU General Public License is available at # http://www.r-project.org/Licenses/ rstandard.gnm <- function(model, ...) { so <- summary(model) res <- na.omit(so$deviance.resid[model$prior.weights != 0]) res <- naresid(model$na.action, res) res <- res/sqrt(so$dispersion * (1 - hatvalues(model))) res[is.infinite(res)] <- NaN if (!is.null(model$table.attr)) attributes(res) <- model$table.attr res } gnm/R/wedderburn.R0000644000176000001440000000437512515147523013566 0ustar ripleyusers# Modification of binomial from the stats package for R. # # Copyright (C) 1995-2005 The R Core Team # Copyright (C) 2005 David Firth # # This program is free software; you can redistribute it and/or modify # it under the terms of the GNU General Public License as published by # the Free Software Foundation; either version 2 or 3 of the License # (at your option). # # This program is distributed in the hope that it will be useful, # but WITHOUT ANY WARRANTY; without even the implied warranty of # MERCHANTABILITY or FITNESS FOR A PARTICULAR PURPOSE. See the # GNU General Public License for more details. # # A copy of the GNU General Public License is available at # http://www.r-project.org/Licenses/ "wedderburn" <- function (link = "logit") { linktemp <- substitute(link) if (!is.character(linktemp)) { linktemp <- deparse(linktemp) if (linktemp == "link") linktemp <- eval(link) } if (any(linktemp == c("logit", "probit", "cloglog"))) stats <- make.link(linktemp) else stop(paste(linktemp, "link not available for wedderburn quasi-family;", "available links are", "\"logit\", \"probit\" and \"cloglog\"")) variance <- function(mu) mu^2 * (1-mu)^2 validmu <- function(mu) { all(mu > 0) && all(mu < 1)} dev.resids <- function(y, mu, wt){ eps <- 0.0005 2 * wt * (y/mu + (1 - y)/(1 - mu) - 2 + (2 * y - 1) * log((y + eps)*(1 - mu)/((1- y + eps) * mu))) } aic <- function(y, n, mu, wt, dev) NA initialize <- expression({ if (any(y < 0 | y > 1)) stop(paste( "Values for the wedderburn family must be in [0,1]")) n <- rep.int(1, nobs) mustart <- (y + 0.1)/1.2 }) structure(list(family = "wedderburn", link = linktemp, linkfun = stats$linkfun, linkinv = stats$linkinv, variance = variance, dev.resids = dev.resids, aic = aic, mu.eta = stats$mu.eta, initialize = initialize, validmu = validmu, valideta = stats$valideta), class = "family") } gnm/R/MPinv.R0000644000176000001440000000521212515147523012445 0ustar ripleyusers# Copyright (C) 2005, 2006, 2010 David Firth and Heather Turner # # This program is free software; you can redistribute it and/or modify # it under the terms of the GNU General Public License as published by # the Free Software Foundation; either version 2 or 3 of the License # (at your option). # # This program is distributed in the hope that it will be useful, # but WITHOUT ANY WARRANTY; without even the implied warranty of # MERCHANTABILITY or FITNESS FOR A PARTICULAR PURPOSE. See the # GNU General Public License for more details. # # A copy of the GNU General Public License is available at # http://www.r-project.org/Licenses/ MPinv <- function (mat, tolerance = 100 * .Machine$double.eps, rank = NULL, method = "svd") { theRank <- rank if (!is.matrix(mat)) stop("mat must be a matrix") m <- nrow(mat) n <- ncol(mat) Rownames <- rownames(mat) Colnames <- colnames(mat) if (method == "svd") { Svd <- svd(mat) Positive <- rep(FALSE, length(Svd$d)) if (is.null(theRank)) { Positive <- Svd$d > max(tolerance * Svd$d[1], 0) } else Positive[1:theRank] <- TRUE result <- { if (all(Positive)) Svd$v %*% (1/Svd$d * t(Svd$u)) else if (!any(Positive)) array(0, dim(mat)[2:1]) else Svd$v[, Positive, drop = FALSE] %*% ((1/Svd$d[Positive]) * t(Svd$u[, Positive, drop = FALSE])) } attr(result, "rank") <- sum(Positive) } if (method == "chol") { ## Generalized inverse of a symmetric matrix using a ## streamlined version of the "fast" method of ## Courrieu, P. (2005). Fast computation of Moore-Penrose ## inverse matrices. Neural Information Processing 8, 25-29. ## ## No test for symmetry performed here! if (!(m == n)) stop("the matrix is not symmetric") S <- suppressWarnings(chol(mat, pivot = TRUE)) ## (non-full-rank case) if (is.null(theRank)) { theRank <- qr(S)$rank ## fails only on the bwt.po example ## theRank <- attr(S, "rank") ## seems less reliable in general } pivot <- attr(S, "pivot") oPivot <- order(pivot) Lt <- S[oPivot[oPivot %in% 1:theRank], oPivot, drop = FALSE] LLinv <- chol2inv(chol(tcrossprod(Lt))) result <- crossprod(Lt, crossprod(LLinv)) %*% Lt attr(result, "rank") <- theRank } if (!is.null(Rownames)) colnames(result) <- Rownames if (!is.null(Colnames)) rownames(result) <- Colnames return(result) } gnm/R/confint.profile.gnm.R0000644000176000001440000000335012515147523015274 0ustar ripleyusers# Modification of confint.profile.glm from the MASS package for R. # # Copyright (C) 1994-2006 W. N. Venables and B. D. Ripley # Copyright (C) 2006 Heather Turner # # This program is free software; you can redistribute it and/or modify # it under the terms of the GNU General Public License as published by # the Free Software Foundation; either version 2 or 3 of the License # (at your option). # # This program is distributed in the hope that it will be useful, # but WITHOUT ANY WARRANTY; without even the implied warranty of # MERCHANTABILITY or FITNESS FOR A PARTICULAR PURPOSE. See the # GNU General Public License for more details. # # A copy of the GNU General Public License is available at # http://www.r-project.org/Licenses/ confint.profile.gnm <- function (object, parm = names(object), level = 0.95, ...) { of <- attr(object, "original.fit") pnames <- names(coef(of)) if (is.numeric(parm)) parm <- pnames[parm] a <- (1 - level)/2 a <- c(a, 1 - a) pct <- paste(round(100 * a, 1), "%") ci <- array(NA, dim = c(length(parm), 2), dimnames = list(parm, pct)) cutoff <- qnorm(a) std.err <- attr(object, "summary")$coefficients[, "Std. Error"] parm <- parm[!is.na(std.err)[parm]] for (pm in parm) { pro <- object[[pm]] if (is.matrix(pro[, "par.vals"])) sp <- spline(x = pro[, "par.vals"][, pm], y = pro[, 1]) else sp <- spline(x = pro[, "par.vals"], y = pro[, 1]) print(pro[, "par.vals"][, pm]) print(pro[,1]) est <- approx(sp$y, sp$x, xout = cutoff)$y ci[pm, ] <- ifelse(is.na(est) & attr(pro, "asymptote"), c(-Inf, Inf), est) } drop(ci) } gnm/R/asGnm.R0000644000176000001440000000132612515147523012463 0ustar ripleyusers# Copyright (C) 2006 Heather Turner # # This program is free software; you can redistribute it and/or modify # it under the terms of the GNU General Public License as published by # the Free Software Foundation; either version 2 or 3 of the License # (at your option). # # This program is distributed in the hope that it will be useful, # but WITHOUT ANY WARRANTY; without even the implied warranty of # MERCHANTABILITY or FITNESS FOR A PARTICULAR PURPOSE. See the # GNU General Public License for more details. # # A copy of the GNU General Public License is available at # http://www.r-project.org/Licenses/ asGnm <- function(object, ...){ if (is.null(object)) return(NULL) UseMethod("asGnm") } gnm/R/print.vcov.gnm.R0000644000176000001440000000147312515147523014311 0ustar ripleyusers# Copyright (C) 2005, 2006, 2010 Heather Turner # # This program is free software; you can redistribute it and/or modify # it under the terms of the GNU General Public License as published by # the Free Software Foundation; either version 2 or 3 of the License # (at your option). # # This program is distributed in the hope that it will be useful, # but WITHOUT ANY WARRANTY; without even the implied warranty of # MERCHANTABILITY or FITNESS FOR A PARTICULAR PURPOSE. See the # GNU General Public License for more details. # # A copy of the GNU General Public License is available at # http://www.r-project.org/Licenses/ print.vcov.gnm <- function(x, ...) { if (!is.null(attr(x, "ofInterest"))){ print.default(x[attr(x, "ofInterest"), attr(x, "ofInterest")]) } else print.default(x) } gnm/R/predict.gnm.R0000644000176000001440000001716012515233503013625 0ustar ripleyusers# Copyright (C) 2005, 2008, 2010, 2012, 2014, 2015 Heather Turner # # This program is free software; you can redistribute it and/or modify # it under the terms of the GNU General Public License as published by # the Free Software Foundation; either version 2 or 3 of the License # (at your option). # # This program is distributed in the hope that it will be useful, # but WITHOUT ANY WARRANTY; without even the implied warranty of # MERCHANTABILITY or FITNESS FOR A PARTICULAR PURPOSE. See the # GNU General Public License for more details. # # A copy of the GNU General Public License is available at # http://www.r-project.org/Licenses/ predict.gnm <- function (object, newdata = NULL, type = c("link", "response", "terms"), se.fit = FALSE, dispersion = NULL, terms = NULL, na.action = na.exclude, ...) { type <- match.arg(type) if (type == "terms") { hasintercept <- attr(object$terms, "intercept") > 0L ## do not include eliminate term - cannot check estimability without ## creating full matrix, defeating point of eliminate if (is.null(terms)) { terms <- attr(object$terms, "term.labels") } else { terms <- setdiff(terms, "(eliminate)") } } if (missing(newdata)) { pred <- switch(type, link = object$predictors, response = object$fitted.values, terms = {pred <- termPredictors(object) ## see 6.3.6 white book & predict.lm if (hasintercept) { predc <- sweep(pred, 2, colMeans(pred)) const <- sum(pred[1,]) - sum(predc[1,]) structure(predc[, terms, drop = FALSE], constant = const) } else structure(pred[, terms, drop = FALSE], constant = 0)}) if (!is.null(na.act <- object$na.action)){ pred <- napredict(na.act, pred) } if (!inherits(pred, "matrix") && !is.null(object$table.attr)) attributes(pred) <- object$table.attr } else { modelTerms <- delete.response(terms(object)) ## evaluate eliminate in environment of formula if (is.null(object$eliminate)){ modelData <- model.frame(modelTerms, newdata, na.action = na.action, xlev = object$xlevels) } else { modelData <- model.frame(modelTerms, newdata, eliminate = eval(eliminate), na.action = na.action, xlev = object$xlevels) } ## use same contrasts as in original model contr <- lapply(model.frame(object)[names(modelData)], attr, "contrasts") for (i in which(!sapply(contr, is.null))){ modelData[[i]] <- C(modelData[[i]], contr[[i]]) } if (length(offID <- attr(modelTerms, "offset"))){ offset <- eval(attr(modelTerms, "variables")[[offID + 1]], newdata) } else offset <- eval(object$call$offset, newdata) modelTools <- gnmTools(modelTerms, modelData) varPredictors <- modelTools$varPredictors(parameters(object)) pred <- modelTools$predictor(varPredictors, term = type == "terms") if (type == "terms") { rownames(pred) <- rownames(modelData) } else names(pred) <- rownames(modelData) if (!is.null(offset)) pred <- offset + pred if (!is.null(object$eliminate)) { prede <- attr(coef(object), "eliminate") if (type != "terms") pred <- prede[modelData$`(eliminate)`] + pred } switch(type, response = {pred <- family(object)$linkinv(pred)}, terms = {if (hasintercept) { predc <- sweep(pred, 2, colMeans(termPredictors(object))) const <- sum(pred[1,]) - sum(predc[1,]) pred <- structure(predc[, terms, drop = FALSE], constant = const) } else structure(pred[, terms, drop = FALSE], constant = 0)}, link = ) if (!is.null(na.act <- attr(modelData, "na.action"))) pred <- napredict(na.act, pred) } if (se.fit) { V <- vcov(object, dispersion = dispersion, with.eliminate = TRUE) residual.scale <- as.vector(sqrt(attr(V, "dispersion"))) if (missing(newdata)) { X <- model.matrix(object) elim <- object$eliminate } else { X <- modelTools$localDesignFunction(parameters(object), varPredictors) elim <- modelData$`(eliminate)` } covElim <- attr(V, "covElim")[elim, , drop = FALSE] varElim <- attr(V, "varElim")[elim] switch(type, link = { if (is.null(elim)) se.fit <- sqrt(diag(X %*% tcrossprod(V, X))) else se.fit <- sqrt(diag(X %*% tcrossprod(V, X)) + 2 * rowSums(X * covElim) + varElim)}, response = { eta <- na.omit(c(family(object)$linkfun(pred))) d <- family(object)$mu.eta(eta) if (is.null(object$eliminate)) se.fit <- sqrt(diag(X %*% tcrossprod(V, X))) else se.fit <- sqrt(diag(X %*% tcrossprod(V, X)) + 2*rowSums(X * covElim) + varElim) se.fit <- d * se.fit}, terms = { if (missing(newdata)) { assign <- split(seq(ncol(X)), attr(X, "assign")) } else { M <- model.matrix(object) assign <- split(seq(ncol(X)), attr(M, "assign")) } if (hasintercept) { if (missing(newdata)) { X <- sweep(X, 2, colMeans(X)) } else X <- sweep(X, 2, colMeans(M)) } se.fit <- matrix(, nrow = nrow(X), ncol = length(terms)) s <- 0 adj <- hasintercept for (i in match(terms, colnames(pred))) { s <- s + 1 t <- assign[[i + adj]] se.fit[, s] <- sqrt(diag(X[, t] %*% tcrossprod(V[t, t], X[, t]))) ## check estimability of term Xt <- X Xt[, -t] <- 0 estimable <- checkEstimable(object, t(Xt)) is.na(se.fit)[estimable %in% c(FALSE, NA), s] <- TRUE } }) ## check estimability of predictions if (!missing(newdata) && type != "terms"){ estimable <- checkEstimable(object, t(X)) is.na(se.fit)[estimable %in% c(FALSE, NA)] <- TRUE } if (!is.null(na.act)) { se.fit <- napredict(na.act, se.fit) } if (inherits(pred, "table")) attributes(se.fit) <- object$table.attr else attributes(se.fit) <- attributes(pred) pred <- list(fit = pred, se.fit = se.fit, residual.scale = residual.scale) } pred } gnm/R/naToZero.R0000755000176000001440000000126312515147523013162 0ustar ripleyusers# Copyright (C) 2005 Heather Turner # # This program is free software; you can redistribute it and/or modify # it under the terms of the GNU General Public License as published by # the Free Software Foundation; either version 2 or 3 of the License # (at your option). # # This program is distributed in the hope that it will be useful, # but WITHOUT ANY WARRANTY; without even the implied warranty of # MERCHANTABILITY or FITNESS FOR A PARTICULAR PURPOSE. See the # GNU General Public License for more details. # # A copy of the GNU General Public License is available at # http://www.r-project.org/Licenses/ naToZero <- function(vec){ vec[is.na(vec)] <- 0 return(vec) } gnm/R/fitted.gnm.R0000644000176000001440000000143412515147523013455 0ustar ripleyusers# Copyright (C) 2008 Heather Turner # # This program is free software; you can redistribute it and/or modify # it under the terms of the GNU General Public License as published by # the Free Software Foundation; either version 2 or 3 of the License # (at your option). # # This program is distributed in the hope that it will be useful, # but WITHOUT ANY WARRANTY; without even the implied warranty of # MERCHANTABILITY or FITNESS FOR A PARTICULAR PURPOSE. See the # GNU General Public License for more details. # # A copy of the GNU General Public License is available at # http://www.r-project.org/Licenses/ fitted.gnm <- function(object, ...) { fitted <- NextMethod("fitted") if (!is.null(object$table.attr)) attributes(fitted) <- object$table.attr fitted } gnm/R/dropterm.gnm.R0000644000176000001440000000143612515147523014034 0ustar ripleyusers# Copyright (C) 2005 Heather Turner # # This program is free software; you can redistribute it and/or modify # it under the terms of the GNU General Public License as published by # the Free Software Foundation; either version 2 or 3 of the License # (at your option). # # This program is distributed in the hope that it will be useful, # but WITHOUT ANY WARRANTY; without even the implied warranty of # MERCHANTABILITY or FITNESS FOR A PARTICULAR PURPOSE. See the # GNU General Public License for more details. # # A copy of the GNU General Public License is available at # http://www.r-project.org/Licenses/ dropterm.gnm <- function (object, ...) { if (inherits(object, "gnm", TRUE) == 1) stop("dropterm is not implemented for gnm objects") else NextMethod } gnm/R/grp.sum.R0000644000176000001440000000131212515147523013004 0ustar ripleyusers# Copyright (C) 2010 Heather Turner # # This program is free software; you can redistribute it and/or modify # it under the terms of the GNU General Public License as published by # the Free Software Foundation; either version 2 or 3 of the License # (at your option). # # This program is distributed in the hope that it will be useful, # but WITHOUT ANY WARRANTY; without even the implied warranty of # MERCHANTABILITY or FITNESS FOR A PARTICULAR PURPOSE. See the # GNU General Public License for more details. # # A copy of the GNU General Public License is available at # http://www.r-project.org/Licenses/ grp.sum <- function(x, grp.end){ x <- cumsum(x)[grp.end] x - c(0, x[-length(x)]) } gnm/R/unlistOneLevel.R0000755000176000001440000000231612515147523014371 0ustar ripleyusers# Copyright (C) 2005 David Firth and Heather Turner # # This program is free software; you can redistribute it and/or modify # it under the terms of the GNU General Public License as published by # the Free Software Foundation; either version 2 or 3 of the License # (at your option). # # This program is distributed in the hope that it will be useful, # but WITHOUT ANY WARRANTY; without even the implied warranty of # MERCHANTABILITY or FITNESS FOR A PARTICULAR PURPOSE. See the # GNU General Public License for more details. # # A copy of the GNU General Public License is available at # http://www.r-project.org/Licenses/ unlistOneLevel <- function(theList){ result <- vector(length = sum(sapply(theList, function(x) if(is.list(x)) length(x) else 1)), mode = "list") count <- 0 for (i in seq(theList)){ theItem <- theList[[i]] if (is.list(theItem)){ for (j in seq(theItem)){ count <- count + 1 result[[count]] <- theItem[[j]] } } else { count <- count + 1 result[[count]] <- theItem } } return(result[1:count]) } gnm/R/confint.gnm.R0000644000176000001440000000231512515147523013635 0ustar ripleyusers# Modification of confint.glm from the MASS package for R. # # Copyright (C) 1994-2006 W. N. Venables and B. D. Ripley # Copyright (C) 2006 Heather Turner # # This program is free software; you can redistribute it and/or modify # it under the terms of the GNU General Public License as published by # the Free Software Foundation; either version 2 or 3 of the License # (at your option). # # This program is distributed in the hope that it will be useful, # but WITHOUT ANY WARRANTY; without even the implied warranty of # MERCHANTABILITY or FITNESS FOR A PARTICULAR PURPOSE. See the # GNU General Public License for more details. # # A copy of the GNU General Public License is available at # http://www.r-project.org/Licenses/ confint.gnm <- function (object, parm = ofInterest(object), level = 0.95, trace = FALSE, ...) { pnames <- names(coef(object)) if (is.null(parm)) parm <- seq(along = pnames) else if (is.character(parm)) parm <- match(parm, pnames, nomatch = 0) cat("Waiting for profiling to be done...\n") flush.console() object <- profile(object, which = parm, alpha = 1 - level, trace = trace) confint(object, level = level, ...) } gnm/R/model.frame.gnm.R0000755000176000001440000000233312515147523014371 0ustar ripleyusers# Modification of model.frame.glm from the stats package for R. # # Copyright (C) 1995-2005 The R Core Team # Copyright (C) 2005, 2006 Heather Turner # # This program is free software; you can redistribute it and/or modify # it under the terms of the GNU General Public License as published by # the Free Software Foundation; either version 2 or 3 of the License # (at your option). # # This program is distributed in the hope that it will be useful, # but WITHOUT ANY WARRANTY; without even the implied warranty of # MERCHANTABILITY or FITNESS FOR A PARTICULAR PURPOSE. See the # GNU General Public License for more details. # # A copy of the GNU General Public License is available at # http://www.r-project.org/Licenses/ model.frame.gnm <- function (formula, ...) { dots <- list(...) nargs <- dots[match(c("data", "na.action", "subset"), names(dots), 0)] if (length(nargs) || is.null(formula$model)) { fcall <- formula$call fcall$method <- "model.frame" fcall[[1]] <- as.name("gnm") fcall[names(nargs)] <- nargs env <- environment(formula$terms) if (is.null(env)) env <- parent.frame() eval(fcall, env) } else formula$model } gnm/R/gnmTerms.R0000755000176000001440000001722612515534061013220 0ustar ripleyusers# Copyright (C) 2005-2010, 2012 Heather Turner # # This program is free software; you can redistribute it and/or modify # it under the terms of the GNU General Public License as published by # the Free Software Foundation; either version 2 or 3 of the License # (at your option). # # This program is distributed in the hope that it will be useful, # but WITHOUT ANY WARRANTY; without even the implied warranty of # MERCHANTABILITY or FITNESS FOR A PARTICULAR PURPOSE. See the # GNU General Public License for more details. # # A copy of the GNU General Public License is available at # http://www.r-project.org/Licenses/ gnmTerms <- function(formula, eliminate = NULL, data = NULL) { if (!is.null(eliminate)){ formula <- as.formula(substitute(a ~ b - e - 1, list(a = formula[[2]], b = formula[[3]], e = eliminate))) } fullTerms <- terms(formula, specials = "instances", simplify = TRUE, keep.order = TRUE, data = data) if (is.empty.model(fullTerms)) return(fullTerms) inst <- attr(fullTerms, "specials")$instances if (length(inst)) { termLabels <- c("0"[!attr(fullTerms, "intercept")], attr(fullTerms, "term.labels")) instLabels <- as.list(attr(fullTerms, "variables"))[inst + 1] termLabels[termLabels %in% instLabels] <- sapply(instLabels, eval) variables <- as.character(attr(fullTerms, "variables"))[-1] offsetLabels <- variables[attr(fullTerms, "offset")] response <- variables[attr(fullTerms, "response")][1][[1]] fullTerms <- terms(reformulate(c(termLabels, offsetLabels), response), keep.order = TRUE, data = data) environment(fullTerms) <- environment(formula) } termLabels <- c("1"[attr(fullTerms, "intercept")], attr(fullTerms, "term.labels")) variables <- predvars <- as.list(attr(fullTerms, "variables"))[-1] specials <- which(sapply(variables, function(x) { length(x) > 1 && inherits(match.fun(x[[1]]), "nonlin") })) if (!length(specials)) { n <- length(termLabels) attributes(fullTerms) <- c(attributes(fullTerms), list(eliminate = !is.null(eliminate), unitLabels = termLabels, common = logical(n), block = numeric(n), match = !logical(n), assign = seq(length = n), type = rep.int("Linear", n), prefixLabels = character(n), varLabels = termLabels, predictor = lapply(termLabels, as.name), class = c("gnmTerms", "terms", "formula"))) return(fullTerms) } specialTerms <- rownames(attr(fullTerms, "factors"))[specials] specialTerms <- strsplit(specialTerms, ", inst = |,? ?\\)$", perl = TRUE) term <- sapply(specialTerms, "[", 1) inst <- as.numeric(sapply(specialTerms, "[", 2)) patch <- term %in% term[inst > 1] & is.na(inst) termLabels[termLabels %in% specials[patch]] <- paste(term[patch], ", inst = 1)") inst[patch] <- 1 nonsense <- tapply(inst, term, FUN = function(x) {!is.na(x) && !identical(as.integer(x), seq(x))}) if (any(nonsense)) stop("Specified instances of ", paste(names(nonsense)[nonsense], ")"), " are not in sequence") offsetVars <- variables[attr(fullTerms, "offset")] nonlinear <- termLabels %in% variables[specials] variables <- variables[-specials] predvars <- predvars[-specials] unitLabels <- varLabels <- as.list(termLabels) predictor <- lapply(termLabels, as.name) names(predictor) <- unitLabels n <- length(unitLabels) blockList <- as.list(numeric(n)) match <- as.list(!logical(n)) common <- as.list(logical(n)) class <- as.list(rep.int("Linear", n)) prefixLabels <- as.list(character(n)) start <- vector("list", n) adj <- 1 for (j in which(nonlinear)) { nonlinCall <- parse(text = unitLabels[[j]])[[1]] args <- eval(nonlinCall, as.data.frame(data), environment(formula)) args <- c(args, nonlin.function = deparse(nonlinCall[[1]]), list(data = data)) tmp <- do.call("nonlinTerms", args) unitLabels[[j]] <- tmp$unitLabels if (!identical(tmp$prefix, "#")) { bits <- hashSplit(tmp$prefix) if (length(bits) > 1) { n <- length(tmp$hashLabels) matched <- tmp$matchID > 0 & !duplicated(tmp$matchID) dot <- (tmp$hashLabels[matched])[order(tmp$matchID[matched])] prefix <- matrix(dot, max(tmp$matchID), n) prefix[cbind(tmp$matchID, seq(n))] <- "." prefix <- rbind(character(n), prefix) sep <- rep(".", n) sep[!tmp$matchID] <- "" prefixLabels[[j]] <- paste(apply(prefix, 2, paste, bits, sep = "", collapse = ""), sep, tmp$suffix, sep = "") for (i in unique(tmp$common[duplicated(tmp$common)])) { dotCommon <- dot commonID <- tmp$common == i dotCommon[tmp$matchID[commonID]] <- "." prefixLabels[[j]][commonID] <- paste(paste(c("", dotCommon), bits, sep = "", collapse = ""), tmp$suffix[commonID], sep[commonID], paste(tmp$unitLabels[commonID], collapse = "|"), sep = "") } } else prefixLabels[[j]] <- paste(tmp$prefix, tmp$suffix, sep = "") } else prefixLabels[[j]] <- tmp$varLabels varLabels[[j]] <- gsub("#", j, tmp$varLabels) predictor[[j]] <- parse(text = gsub("#", j, tmp$predictor))[[1]] blockList[[j]] <- tmp$block + adj match[[j]] <- as.logical(tmp$matchID) common[[j]] <- tmp$common %in% tmp$common[duplicated(tmp$common)] class[[j]] <- tmp$type start[j] <- list(tmp$start) adj <- max(c(0, blockList[[j]])) + 1 variables <- c(variables, tmp$variables) predvars <- c(predvars, tmp$predvars) } if (length(predvars) > 1) nObs <- call("length", predvars[[1]]) else if (!is.null(data)) nObs <- call("length", as.name(names(data)[1])) else nObs <- 1 attributes(fullTerms) <- c(attributes(fullTerms), list(eliminate = !is.null(eliminate), offset = which(unique(variables) %in% offsetVars), variables = as.call(c(quote(list), unique(variables))), predvars = {do.call("substitute", list(as.call(c(quote(list), unique(predvars))), list(nObs = nObs)))}, unitLabels = unlist(unitLabels), common = unlist(common), block = unlist(blockList), match = unlist(match), assign = rep(seq(class), sapply(class, length)), type = unlist(class), prefixLabels = unlist(prefixLabels), varLabels = unlist(varLabels), start = start, predictor = predictor, class = c("gnmTerms", "terms", "formula"))) environment(fullTerms) <- environment(formula) fullTerms } gnm/R/profile.gnm.R0000644000176000001440000001704512515147523013643 0ustar ripleyusers# Modification of profile.glm from the MASS package for R. # # File MASS/profiles.q copyright (C) 1996 D. M. Bates and W. N. Venables. # # port to R by B. D. Ripley copyright (C) 1998 # # corrections copyright (C) 2000,3,6,7 B. D. Ripley # Copyright (C) 2005, 2006, 2008, 2012 Heather Turner # # This program is free software; you can redistribute it and/or modify # it under the terms of the GNU General Public License as published by # the Free Software Foundation; either version 2 or 3 of the License # (at your option). # # This program is distributed in the hope that it will be useful, # but WITHOUT ANY WARRANTY; without even the implied warranty of # MERCHANTABILITY or FITNESS FOR A PARTICULAR PURPOSE. See the # GNU General Public License for more details. # # A copy of the GNU General Public License is available at # http://www.r-project.org/Licenses/ profile.gnm <- function (fitted, which = ofInterest(fitted), alpha = 0.05, maxsteps = 10, stepsize = NULL, trace = FALSE, ...) { fittedCoef <- parameters(fitted) coefNames <- names(fittedCoef) p <- length(coefNames) if (is.null(which)) which <- 1:p else if (is.numeric(which)) which <- which else if (is.character(which)) which <- match(which, coefNames) summ <- summary(fitted) sterr <- summ$coefficients[, "Std. Error"] fittedDev <- deviance(fitted) disp <- summ$dispersion ## use z cutoffs as in confint.profile.gnm zmax <- abs(qnorm(alpha/2)) fittedConstrain <- fitted$constrain fittedConstrainTo <- fitted$constrainTo auto <- is.null(stepsize) if (!auto) stepsize[1:2] <- stepsize prof <- as.list(rep(NA, length(which))) names(prof) <- coefNames[which] which <- which[!is.na(sterr)[which]] for (i in which) { par <- coefNames[i] prof[[par]] <- numeric(2 * maxsteps + 1) par.vals <- matrix(nrow = 2 * maxsteps + 1, ncol = p, dimnames = list(NULL, coefNames)) par.vals[maxsteps + 1,] <- fittedCoef asymptote <- c(FALSE, FALSE) if (auto) { ## set defaults sub <- 3 # no. of steps from MLE to zmax*se stepsize <- c(zmax/sub * sterr[i], zmax/sub * sterr[i]) ## estimate quadratic in the region MLE +/- zmax*se margin <- zmax * sterr[i] updatedDev <- numeric(2) for (sgn in c(-1, 1)) { val <- fittedCoef[i] + sgn * margin updated <- suppressWarnings(update(fitted, constrain = c(fittedConstrain, i), constrainTo = c(fittedConstrainTo, val), trace = FALSE, verbose = FALSE, start = fittedCoef)) if (is.null(updated)) break updatedDev[(sgn + 1)/2 + 1] <- deviance(updated) prof[[par]][maxsteps + 1 + sgn * sub] <- sgn * sqrt((deviance(updated) - fittedDev)/disp) par.vals[maxsteps + + 1 + sgn * sub,] <- parameters(updated) } if (all(updatedDev != 0)) { quad <- (sum(updatedDev) - 2 * fittedDev)/(2 * margin^2) lin <- (fittedDev - updatedDev[1])/margin + quad * (margin - 2 * fittedCoef[i]) int <- fittedDev - lin * fittedCoef[i] - quad * fittedCoef[i]^2 ## adjust so roots approx where deviance gives z = zmax int.adj <- int - zmax^2 * disp - fittedDev for (sgn in c(-1, 1)) { dir <- (sgn + 1)/2 + 1 root <- (-lin + sgn * sqrt(lin^2 - 4 * int.adj * quad))/ (2 * quad) firstApprox <- par.vals[maxsteps + 1 + sgn * sub, i] ## if likelihood approx quadratic use default stepsize, else if (sgn * (root - firstApprox) > 0) { ## not gone out far enough, check for asymptote val <- fittedCoef[i] + sgn * 10 * sterr[i] updated <- suppressWarnings(update(fitted, constrain = c(fittedConstrain, i), constrainTo = c(fittedConstrainTo, val), trace = FALSE, verbose = FALSE, start = fittedCoef)) if (!is.null(updated) && sqrt((deviance(updated) - fittedDev)/disp) < zmax) asymptote[dir] <- TRUE } ## if root more than one step away from firstApprox, i.e. ## less than two steps away from fittedCoef, halve stepsize if (abs(sgn * (firstApprox - root)) > stepsize[dir] && !asymptote[dir]) { prof[[par]][maxsteps + 1 + sgn * sub] <- 0 par.vals[maxsteps + 1 + sgn * sub, ] <- NA stepsize[dir] <- abs(root - fittedCoef[i])/(maxsteps/2) } } } } for (sgn in c(-1, 1)) { if (trace) prattle("\nParameter:", par, c("down", "up")[(sgn + 1)/2 + 1], "\n") step <- 0 init <- parameters(fitted) while ((step <- step + 1) <= maxsteps) { if (step > 2 && abs(prof[[par]][maxsteps + 1 + sgn * (step - 2)]) > zmax) break if (prof[[par]][maxsteps + 1 + sgn * step] != 0) next val <- fittedCoef[i] + sgn * step * stepsize[(sgn + 1)/2 + 1] updated <- suppressWarnings(update(fitted, constrain = c(fittedConstrain, i), constrainTo = c(fittedConstrainTo, val), trace = FALSE, verbose = FALSE, start = init)) if (is.null(updated)) { message("Could not complete profile for", par, "\n") break } init <- parameters(updated) zz <- (deviance(updated) - fittedDev)/disp if (zz > -0.001) zz <- max(zz, 0) else stop("profiling has found a better solution, ", "so original fit had not converged") prof[[par]][maxsteps + 1 + sgn * step] <- sgn * sqrt(zz) par.vals[maxsteps + 1 + sgn * step,] <- init #print(data.frame(step = step, val = bi, deviance = fm$deviance, #zstat = z)) } } prof[[par]] <- structure(data.frame(prof[[par]][!is.na(par.vals[,1])]), names = "z") prof[[par]]$par.vals <- par.vals[!is.na(par.vals[,1]), , drop = FALSE] attr(prof[[par]], "asymptote") <- asymptote } val <- structure(prof, original.fit = fitted, summary = summ) class(val) <- c("profile.gnm", "profile.glm", "profile") val } gnm/R/influence.gnm.R0000644000176000001440000000145612515147523014152 0ustar ripleyusers# Copyright (C) 2005 Heather Turner # # This program is free software; you can redistribute it and/or modify # it under the terms of the GNU General Public License as published by # the Free Software Foundation; either version 2 or 3 of the License # (at your option). # # This program is distributed in the hope that it will be useful, # but WITHOUT ANY WARRANTY; without even the implied warranty of # MERCHANTABILITY or FITNESS FOR A PARTICULAR PURPOSE. See the # GNU General Public License for more details. # # A copy of the GNU General Public License is available at # http://www.r-project.org/Licenses/ influence.gnm <- function (model, do.coef = TRUE, ...) { if (inherits(model, "gnm", TRUE) == 1) stop("influence is not implemented for gnm objects") else NextMethod } gnm/R/pprod.R0000755000176000001440000000210112515147523012535 0ustar ripleyusers# Copyright (C) 2005 Heather Turner # # This program is free software; you can redistribute it and/or modify # it under the terms of the GNU General Public License as published by # the Free Software Foundation; either version 2 or 3 of the License # (at your option). # # This program is distributed in the hope that it will be useful, # but WITHOUT ANY WARRANTY; without even the implied warranty of # MERCHANTABILITY or FITNESS FOR A PARTICULAR PURPOSE. See the # GNU General Public License for more details. # # A copy of the GNU General Public License is available at # http://www.r-project.org/Licenses/ pprod <- function(...) { factorList <- list(...) nFactors <- length(factorList) if (nFactors == 0) return(1) else if (nFactors == 1) return(factorList[[1]]) else { tryProduct <- try(factorList[[1]] * do.call("Recall", factorList[-1]), silent = TRUE) if (inherits(tryProduct, "try-error")) stop("multiplication not implemented for types of argument supplied") else tryProduct } } gnm/R/alias.gnm.R0000644000176000001440000000142712515147523013271 0ustar ripleyusers# Copyright (C) 2005 Heather Turner # # This program is free software; you can redistribute it and/or modify # it under the terms of the GNU General Public License as published by # the Free Software Foundation; either version 2 or 3 of the License # (at your option). # # This program is distributed in the hope that it will be useful, # but WITHOUT ANY WARRANTY; without even the implied warranty of # MERCHANTABILITY or FITNESS FOR A PARTICULAR PURPOSE. See the # GNU General Public License for more details. # # A copy of the GNU General Public License is available at # http://www.r-project.org/Licenses/ alias.gnm <- function (object, ...){ if (inherits(object, "gnm", TRUE) == 1) stop("alias is not implemented for gnm objects") else NextMethod } gnm/R/asGnm.lm.R0000644000176000001440000000361312515147523013073 0ustar ripleyusers# Copyright (C) 2006, 2008, 2010 Heather Turner # # This program is free software; you can redistribute it and/or modify # it under the terms of the GNU General Public License as published by # the Free Software Foundation; either version 2 or 3 of the License # (at your option). # # This program is distributed in the hope that it will be useful, # but WITHOUT ANY WARRANTY; without even the implied warranty of # MERCHANTABILITY or FITNESS FOR A PARTICULAR PURPOSE. See the # GNU General Public License for more details. # # A copy of the GNU General Public License is available at # http://www.r-project.org/Licenses/ asGnm.lm <- function(object, ...) { lmExtra <- match(c("effects", "assign", "qr", "contrasts"), names(object)) modelData <- model.frame(object) object[lmExtra] <- NULL object$call[[1]] <- as.name("gnm") constrain <- which(is.na(coef(object))) object <- c(list(formula = formula(object), eliminate = NULL, ofInterest = NULL, na.action = na.action(modelData), constrain = constrain, constrainTo = numeric(length(constrain)), family = gaussian(), predictors = fitted.values(object), deviance = deviance(object), y = model.response(modelData)), object) object$terms <- gnmTerms(object$formula, data = modelData) object$weights <- object$prior.weights <- rep.int(1, length(object$y)) object$aic <- 2 * object$rank + object$family$aic(object$y, object$weights, object$fitted.values, object$weights, object$deviance) if (is.null(object$offset)) object$offset <- rep.int(0, length(coef(object))) object$tolerance <- object$iterStart <- object$iterMax <- object$iter <- object$converged <- "Not available - model fitted by lm()" class(object) <- c("gnm", "glm", "lm") object } gnm/R/add1.gnm.R0000644000176000001440000001234712515147523013014 0ustar ripleyusers# Modification of add1.glm from the stats package for R. # # Copyright (C) 1994-8 W. N. Venables and B. D. Ripley # Copyright (C) 1998-2005 The R Core Team # Copyright (C) 2005, 2010 Heather Turner # # This program is free software; you can redistribute it and/or modify # it under the terms of the GNU General Public License as published by # the Free Software Foundation; either version 2 or 3 of the License # (at your option). # # This program is distributed in the hope that it will be useful, # but WITHOUT ANY WARRANTY; without even the implied warranty of # MERCHANTABILITY or FITNESS FOR A PARTICULAR PURPOSE. See the # GNU General Public License for more details. # # A copy of the GNU General Public License is available at # http://www.r-project.org/Licenses/ add1.gnm <- function (object, scope, scale = 0, test = c("none", "Chisq", "F"), x = NULL, k = 2, ...) { if (any(attr(terms(object), "type") != "Linear")) stop("add1 is not implemented for gnm objects with nonlinear terms.") Fstat <- function(table, rdf) { dev <- table$Deviance df <- table$Df diff <- pmax(0, (dev[1L] - dev)/df) Fs <- (diff/df)/(dev/(rdf - df)) Fs[df < .Machine$double.eps] <- NA P <- Fs nnas <- !is.na(Fs) P[nnas] <- pf(Fs[nnas], df[nnas], rdf - df[nnas], lower.tail = FALSE) list(Fs = Fs, P = P) } if (!is.character(scope)) scope <- add.scope(object, update.formula(object, scope)) if (!length(scope)) stop("no terms in scope for adding to object") oTerms <- attr(object$terms, "term.labels") int <- attr(object$terms, "intercept") ns <- length(scope) dfs <- dev <- numeric(ns + 1) names(dfs) <- names(dev) <- c("", scope) add.rhs <- paste(scope, collapse = "+") add.rhs <- eval(parse(text = paste("~ . +", add.rhs))) new.form <- update.formula(object, add.rhs) Terms <- terms(new.form) y <- object$y if (is.null(x)) { fc <- object$call fc$formula <- Terms fob <- list(call = fc, terms = Terms) class(fob) <- oldClass(object) m <- model.frame(fob, xlev = object$xlevels) offset <- model.offset(m) wt <- model.weights(m) x <- model.matrix(Terms, m, contrasts.arg = object$contrasts) oldn <- length(y) y <- model.response(m) if (!is.factor(y)) storage.mode(y) <- "double" if (NCOL(y) == 2) { n <- y[, 1] + y[, 2] y <- ifelse(n == 0, 0, y[, 1]/n) if (is.null(wt)) wt <- rep.int(1, length(y)) wt <- wt * n } newn <- length(y) if (newn < oldn) warning(gettextf("using the %d/%d rows from a combined fit", newn, oldn), domain = NA) } else { wt <- object$prior.weights offset <- object$offset } n <- nrow(x) if (is.null(wt)) wt <- rep.int(1, n) Terms <- attr(Terms, "term.labels") asgn <- attr(x, "assign") ousex <- match(asgn, match(oTerms, Terms), 0L) > 0L if (int) ousex[1L] <- TRUE X <- x[, ousex, drop = FALSE] z <- glm.fit.e(X, y, wt, offset = offset, family = object$family, eliminate = object$eliminate) dfs[1L] <- z$rank dev[1L] <- z$deviance sTerms <- sapply(strsplit(Terms, ":", fixed = TRUE), function(x) paste(sort(x), collapse = ":")) for (tt in scope) { stt <- paste(sort(strsplit(tt, ":")[[1L]]), collapse = ":") usex <- match(asgn, match(stt, sTerms), 0L) > 0L X <- x[, usex | ousex, drop = FALSE] z <- glm.fit.e(X, y, wt, offset = offset, family = object$family, eliminate = object$eliminate) dfs[tt] <- z$rank dev[tt] <- z$deviance } if (scale == 0) dispersion <- summary(object, dispersion = NULL)$dispersion else dispersion <- scale fam <- object$family$family if (fam == "gaussian") { if (scale > 0) loglik <- dev/scale - n else loglik <- n * log(dev/n) } else loglik <- dev/dispersion aic <- loglik + k * dfs aic <- aic + (extractAIC(object, k = k)[2L] - aic[1L]) dfs <- dfs - dfs[1L] dfs[1L] <- NA aod <- data.frame(Df = dfs, Deviance = dev, AIC = aic, row.names = names(dfs), check.names = FALSE) if (all(is.na(aic))) aod <- aod[, -3] test <- match.arg(test) if (test == "Chisq") { dev <- pmax(0, loglik[1L] - loglik) dev[1L] <- NA LRT <- if (dispersion == 1) "LRT" else "scaled dev." aod[, LRT] <- dev nas <- !is.na(dev) dev[nas] <- pchisq(dev[nas], aod$Df[nas], lower.tail = FALSE) aod[, "Pr(Chi)"] <- dev } else if (test == "F") { if (fam == "binomial" || fam == "poisson") warning(gettextf("F test assumes quasi%s family", fam), domain = NA) rdf <- object$df.residual aod[, c("F value", "Pr(F)")] <- Fstat(aod, rdf) } head <- c("Single term additions", "\nModel:", deparse(as.vector(formula(object))), if (scale > 0) paste("\nscale: ", format(scale), "\n")) class(aod) <- c("anova", "data.frame") attr(aod, "heading") <- head aod } gnm/R/print.coef.gnm.R0000644000176000001440000000205312515147523014243 0ustar ripleyusers# Copyright (C) 2005, 2006, 2010 Heather Turner # # This program is free software; you can redistribute it and/or modify # it under the terms of the GNU General Public License as published by # the Free Software Foundation; either version 2 or 3 of the License # (at your option). # # This program is distributed in the hope that it will be useful, # but WITHOUT ANY WARRANTY; without even the implied warranty of # MERCHANTABILITY or FITNESS FOR A PARTICULAR PURPOSE. See the # GNU General Public License for more details. # # A copy of the GNU General Public License is available at # http://www.r-project.org/Licenses/ print.coef.gnm <- function(x, ...) { if (!is.null(attr(x, "ofInterest"))) { if (length(attr(x, "ofInterest"))){ cat("Coefficients of interest:\n", sep = "") print.default(format(x[attr(x, "ofInterest")]), quote = FALSE) } else cat("No coefficients of interest\n") } else { cat("Coefficients:\n") print.default(format(x), quote = FALSE) } } gnm/R/getData.R0000644000176000001440000000143312515147523012766 0ustar ripleyusers# Copyright (C) 2006 Heather Turner # # This program is free software; you can redistribute it and/or modify # it under the terms of the GNU General Public License as published by # the Free Software Foundation; either version 2 or 3 of the License # (at your option). # # This program is distributed in the hope that it will be useful, # but WITHOUT ANY WARRANTY; without even the implied warranty of # MERCHANTABILITY or FITNESS FOR A PARTICULAR PURPOSE. See the # GNU General Public License for more details. # # A copy of the GNU General Public License is available at # http://www.r-project.org/Licenses/ getData <- function() { nFrame <- match(TRUE, sapply(sys.calls(), function(x) { identical(x[[1]], as.name("gnmTerms"))})) get("data", sys.frame(nFrame)) } gnm/R/update.gnm.R0000644000176000001440000000407312515147523013462 0ustar ripleyusers# Modification of update.default from the stats package for R. # # Copyright (C) 1995-2010 The R Core Team # Copyright (C) 2010, 2012 Heather Turner # # This program is free software; you can redistribute it and/or modify # it under the terms of the GNU General Public License as published by # the Free Software Foundation; either version 2 or 3 of the License # (at your option). # # This program is distributed in the hope that it will be useful, # but WITHOUT ANY WARRANTY; without even the implied warranty of # MERCHANTABILITY or FITNESS FOR A PARTICULAR PURPOSE. See the # GNU General Public License for more details. # # A copy of the GNU General Public License is available at # http://www.r-project.org/Licenses/ update.gnm <- function (object, formula., ..., evaluate = TRUE) { call <- object$call if (is.null(call)) stop("need an object with call component") extras <- match.call(expand.dots = FALSE)$... if (!missing(formula.)) { ## update.formula reorders nonlin terms as lin (main effects) ## therefore use substitute to keep order formula. <- as.formula(formula.) rhs <- formula.[[length(formula.)]] rhs <- do.call(substitute, list(rhs, env = list("." = object$formula[[3]]))) if (length(formula.) == 3) { lhs <- formula.[[2]] lhs <- do.call(substitute, list(lhs, env = list("." = object$formula[[2]]))) call$formula <- call("~", lhs, rhs) } else call$formula <- call("~", object$formula[[2]], rhs) call$formula <- formula(terms.formula(call$formula, simplify = TRUE, keep.order = TRUE)) } if (length(extras)) { existing <- !is.na(match(names(extras), names(call))) for (a in names(extras)[existing]) call[[a]] <- extras[[a]] if (any(!existing)) { call <- c(as.list(call), extras[!existing]) call <- as.call(call) } } if (evaluate) eval(call, parent.frame()) else call } gnm/R/addterm.gnm.R0000644000176000001440000000143412515147523013616 0ustar ripleyusers# Copyright (C) 2005 Heather Turner # # This program is free software; you can redistribute it and/or modify # it under the terms of the GNU General Public License as published by # the Free Software Foundation; either version 2 or 3 of the License # (at your option). # # This program is distributed in the hope that it will be useful, # but WITHOUT ANY WARRANTY; without even the implied warranty of # MERCHANTABILITY or FITNESS FOR A PARTICULAR PURPOSE. See the # GNU General Public License for more details. # # A copy of the GNU General Public License is available at # http://www.r-project.org/Licenses/ addterm.gnm <- function (object, ...) { if (inherits(object, "gnm", TRUE) == 1) stop("addterm is not implemented for gnm objects") else NextMethod } gnm/R/parameters.R0000644000176000001440000000131612515147523013560 0ustar ripleyusers# Copyright (C) 2006 Heather Turner # # This program is free software; you can redistribute it and/or modify # it under the terms of the GNU General Public License as published by # the Free Software Foundation; either version 2 or 3 of the License # (at your option). # # This program is distributed in the hope that it will be useful, # but WITHOUT ANY WARRANTY; without even the implied warranty of # MERCHANTABILITY or FITNESS FOR A PARTICULAR PURPOSE. See the # GNU General Public License for more details. # # A copy of the GNU General Public License is available at # http://www.r-project.org/Licenses/ parameters <- function(object) replace(coef(object), object$constrain, object$constrainTo) gnm/R/MultHomog.R0000755000176000001440000000175512515147523013342 0ustar ripleyusers# Copyright (C) 2005, 2006, 2008 Heather Turner # # This program is free software; you can redistribute it and/or modify # it under the terms of the GNU General Public License as published by # the Free Software Foundation; either version 2 or 3 of the License # (at your option). # # This program is distributed in the hope that it will be useful, # but WITHOUT ANY WARRANTY; without even the implied warranty of # MERCHANTABILITY or FITNESS FOR A PARTICULAR PURPOSE. See the # GNU General Public License for more details. # # A copy of the GNU General Public License is available at # http://www.r-project.org/Licenses/ MultHomog <- function(..., inst = NULL){ dots <- match.call(expand.dots = FALSE)[["..."]] list(predictors = dots, common = rep(1, length(dots)), term = function(predLabels, ...) { paste("(", paste(predLabels, collapse = ")*("), ")", sep = "") }, call = as.expression(match.call())) } class(MultHomog) <- "nonlin" gnm/R/checkCall.R0000644000176000001440000000160312515147523013265 0ustar ripleyusers# Copyright (C) 2006 Heather Turner # # This program is free software; you can redistribute it and/or modify # it under the terms of the GNU General Public License as published by # the Free Software Foundation; either version 2 or 3 of the License # (at your option). # # This program is distributed in the hope that it will be useful, # but WITHOUT ANY WARRANTY; without even the implied warranty of # MERCHANTABILITY or FITNESS FOR A PARTICULAR PURPOSE. See the # GNU General Public License for more details. # # A copy of the GNU General Public License is available at # http://www.r-project.org/Licenses/ checkCall <- function(){ badCall <- lapply(sys.calls(), "[[", 1) %in% c("model.frame.default", "model.matrix.default") if (any(badCall)) stop(paste(sys.call(-1)[[1]], "terms are only valid in gnm models.")) } gnm/R/zzz.R0000644000176000001440000000011412515147523012245 0ustar ripleyusers.onUnload <- function(libpath) { library.dynam.unload("gnm", libpath) } gnm/R/summary.meanResiduals.R0000644000176000001440000000330412515147523015704 0ustar ripleyusers# Copyright (C) 2010, 2012 Heather Turner # # This program is free software; you can redistribute it and/or modify # it under the terms of the GNU General Public License as published by # the Free Software Foundation; either version 2 or 3 of the License # (at your option). # # This program is distributed in the hope that it will be useful, # but WITHOUT ANY WARRANTY; without even the implied warranty of # MERCHANTABILITY or FITNESS FOR A PARTICULAR PURPOSE. See the # GNU General Public License for more details. # # A copy of the GNU General Public License is available at # http://www.r-project.org/Licenses/ # this should always be a summary based on single grouping factor summary.meanResiduals <- function (object, digits = max(3, getOption("digits") - 3), ...) { cat("\nModel call:\n", deparse(attr(object, "call"), width.cutoff = options()$width), "\n", sep = "", fill = TRUE) cat("Mean residuals by ", attr(object, "by"), ":\n\n", sep = "") q <- quantile(object, na.rm = TRUE) names(q) <- c("Min", "1Q", "Median", "3Q", "Max") print.default(q, digits = digits, na.print = "", print.gap = 2) if (attr(object, "standardized")) { cat("\nTest of Normality:\n") df <- attr(object, "df") if (df > 0) { chi.sq <- sum(as.vector(object)^2) p.value <- pchisq(chi.sq, df, lower.tail = FALSE) test <- c(chi.sq, df, p.value) cat("\nChi^2 =", format(chi.sq, digits = digits), "on", df, "df, p-value =", format(p.value, digits = digits), "\n") } else cat("\n(zero degrees of freedom)\n") } else cat("\nResiduals are not standardized\n") } gnm/R/ofInterest.R0000644000176000001440000000125012515147523013534 0ustar ripleyusers# Copyright (C) 2006 Heather Turner # # This program is free software; you can redistribute it and/or modify # it under the terms of the GNU General Public License as published by # the Free Software Foundation; either version 2 or 3 of the License # (at your option). # # This program is distributed in the hope that it will be useful, # but WITHOUT ANY WARRANTY; without even the implied warranty of # MERCHANTABILITY or FITNESS FOR A PARTICULAR PURPOSE. See the # GNU General Public License for more details. # # A copy of the GNU General Public License is available at # http://www.r-project.org/Licenses/ ofInterest <- function(object) { object$ofInterest } gnm/R/nonlinTerms.R0000644000176000001440000002345612515147523013736 0ustar ripleyusers# Copyright (C) 2006-2010, 2012 Heather Turner # # This program is free software; you can redistribute it and/or modify # it under the terms of the GNU General Public License as published by # the Free Software Foundation; either version 2 or 3 of the License # (at your option). # # This program is distributed in the hope that it will be useful, # but WITHOUT ANY WARRANTY; without even the implied warranty of # MERCHANTABILITY or FITNESS FOR A PARTICULAR PURPOSE. See the # GNU General Public License for more details. # # A copy of the GNU General Public License is available at # http://www.r-project.org/Licenses/ nonlinTerms <- function(predictors, variables = NULL, term = NULL, common = seq(predictors), call = NULL, match = numeric(length(predictors)), start = NULL, nonlin.function = NULL, data = NULL) { shadow <- predictor <- predvars <- vars <- unitLabels <- hashLabels <- offsetLabels <- varLabels <- blockList <- matchID <- suffix <-list() if (length(names(predictors))) { suffix <- as.list(names(predictors)) ID <- match(suffix, unique(suffix)) for (i in unique(ID[duplicated(suffix) & suffix != ""])) { dup <- ID == i suffix[dup] <- paste(suffix[dup], seq(sum(dup)), sep = "") } } else suffix <- as.list(rep("", length(predictors))) common <- as.list(common) adj <- 0 hash <- 0 dup <- duplicated(match) for (i in order(match)) { if (inherits(predictors[[i]], "formula")){ nonlinTerms <- terms(predictors[[i]], specials = "Const", keep.order = TRUE, data = data) twiddle <- "~ " } else { nonlinTerms <- terms(eval(substitute(~ -1 + p, list(p = predictors[[i]]))), specials = "Const", keep.order = TRUE, data = data) twiddle <- "" } if (attr(nonlinTerms, "intercept") & !match[i] & !nchar(suffix[[i]])) stop("\"nonlin\" function ", nonlin.function, " must either name ", "predictors that may include an intercept \n or match them ", "to a call") if (is.empty.model(nonlinTerms)) { predvars[[i]] <- vars[[i]] <- as.list(attr(nonlinTerms, "variables"))[-1] offsetLabels[[i]] <- vars[[i]][attr(nonlinTerms, "offset")] varLabels[[i]] <- predictor[[i]] <- unitLabels[[i]] <- NULL blockList[[i]] <- numeric(0) suffix[[i]] <- character(0) } else { unitLabels[[i]] <- as.list(c("1"[attr(nonlinTerms, "intercept")], attr(nonlinTerms, "term.labels"))) vars[[i]] <- predvars[[i]] <- as.list(attr(nonlinTerms, "variables"))[-1] specials <- sapply(vars[[i]], function(x) { length(x) > 1 && inherits(match.fun(x[[1]]), "nonlin")}) const <- attr(nonlinTerms, "specials")$Const if (length(const)) { unitLabels[[i]] <- unitLabels[[i]][!unitLabels[[i]] %in% vars[[i]][const]] predvars[[i]][const] <- lapply(vars[[i]][const], eval) } offsetLabels[[i]] <- vars[[i]][c(attr(nonlinTerms, "offset"), const)] varLabels[[i]] <- as.list(paste("#", adj, gsub("`", ".", unitLabels[[i]]), sep = "")) predictor[[i]] <- paste("`", varLabels[[i]], "`", sep = "") n <- length(unitLabels[[i]]) shadow[[i]] <- rep("#", n) hashLabels[[i]] <- unitLabels[[i]] matchID[[i]] <- as.list(numeric(n)) suffix[[i]] <- as.list(rep(suffix[[i]], n)) if (length(specials)) { nonlinear <- unitLabels[[i]] %in% vars[[i]][specials] vars[[i]] <- vars[[i]][!specials] predvars[[i]] <- predvars[[i]][!specials] } else nonlinear <- rep(FALSE, n) blockList[[i]] <- as.list(nonlinear - min(nonlinear)) if (dup[i]) hash <- last.hash else last.hash <- hash for (j in seq(n)) { if (nonlinear[j]) { tmp <- do.call("Recall", eval(parse(text = unitLabels[[i]][[j]]))) if (match[i]) { if (any(tmp$matchID > 0)) { shadow[[i]][[j]] <- tmp$prefix matchID[[i]][[j]] <- tmp$matchID matchID[[i]][[j]][tmp$matchID != 0] <- hash + matchID[[i]][[j]][tmp$matchID != 0] hashLabels[[i]][[j]] <- tmp$unitLabels } else { lbl <- ifelse(length(tmp$prefix), tmp$prefix, hashLabels[[i]][[j]]) nlbl <- length(tmp$matchID) tmp$suffix <- paste(lbl, tmp$suffix, sep = "") hashLabels[[i]][[j]] <- rep(lbl, nlbl) matchID[[i]][[j]] <- rep(hash + 1, nlbl) } } else { ## could paste call to suffix - but potentially v. long ## and would get cut off anyway: better to rely on ## make.unique for awkward cases ##if (any(tmp$matchID) | !length(tmp$prefix)) ## lbl <- hashLabels[[i]][[j]] ## else ## lbl <- tmp$prefix ## tmp$suffix <- paste(lbl, tmp$suffix, sep = "") if (any(tmp$matchID)) warning("Function using argument-matched ", "labelling (", parse(text = unitLabels[[i]][[j]])[[1]][1], ") used in unmatched predictor\n (see ", "?nonlin) - labels may be ill-defined.\n", call. = FALSE) nlbl <- length(tmp$matchID) hashLabels[[i]][[j]] <- rep(hashLabels[[i]][[j]], nlbl) matchID[[i]][[j]] <- rep(0, nlbl) } varLabels[[i]][[j]] <- gsub("#", paste("#", adj, sep = ""), tmp$varLabels) unitLabels[[i]][[j]] <- tmp$unitLabels blockList[[i]][[j]] <- blockList[[i]][[j]] + tmp$block suffix[[i]][[j]] <- paste(suffix[[i]][[j]], tmp$suffix, sep = "")[!is.null(tmp$suffix)] predictor[[i]][[j]] <- gsub("#", paste("#", adj, sep = ""), tmp$predictor) vars[[i]] <- c(vars[[i]], tmp$variables) predvars[[i]] <- c(predvars[[i]], tmp$predvars) common[[i]] <- common[[i]] * 10 + tmp$common } else { if (match[i]) matchID[[i]][[j]] <- hash + 1 common[[i]] <- common[[i]]*10 + seq(varLabels[[i]]) } hash <- max(c(hash, matchID[[i]][[j]])) } } blockList[[i]] <- unlist(blockList[[i]]) + adj adj <- max(c(-1, blockList[[i]])) + 1 shadow[[i]] <- paste(twiddle, paste(c(unlist(shadow[i]), offsetLabels[[i]]), collapse = " + "), sep = "") if (length(offsetLabels[[i]])) predictor[i] <- paste(c(unlist(predictor[i]), paste("`", offsetLabels[[i]], "`", sep = "")), collapse = " + ") else predictor[i] <- paste(unlist(predictor[i]), collapse = " + ") } common <- unlist(common) if (any(duplicated(common))) { common <- match(common, common) #common <- unlist(varLabels[common]) #common <- match(common, unique(common)) blockList <- unlist(blockList)[common] } else common <- seq(unlist(varLabels)) if (!is.null(call) && sum(match)) { fn <- call[[1]][[1]] call <- as.list(call[[1]][-1]) call[match] <- shadow[match > 0] if (is.null(names(predictors))) names(call)[match] <- "" else names(call)[match] <- names(predictors)[match > 0] sep <- character(length(call)) sep[names(call) != ""] <- " = " call <- paste(names(call), sep, call, sep = "") prefix <- paste(fn, "(", paste(call, collapse = ", "), ")", sep = "") } else prefix <- paste(c(call[[1]])) predictor <- term(unlist(predictor), sapply(variables, function(x) { paste("`", deparse(x), "`", sep = "")})) list(prefix = prefix, matchID = unlist(matchID), variables = c(unlist(vars), variables), predvars = c(unlist(predvars), variables), varLabels = unlist(varLabels), unitLabels = unlist(unitLabels), hashLabels = unlist(hashLabels), block = unlist(blockList), common = common, type = rep.int("Special", length(common)), predictor = predictor, suffix = unlist(suffix), start = start) } gnm/vignettes/0000755000176000001440000000000012515664670013107 5ustar ripleyusersgnm/vignettes/fig-profilePlot.pdf0000644000176000001440000002411512515147523016640 0ustar ripleyusers%PDF-1.4 %âãÏÓ\r 1 0 obj << /CreationDate (D:20060922120623) /ModDate (D:20060922120623) /Title (R Graphics Output) /Producer (R 2.3.1) /Creator (R) >> endobj 2 0 obj << /Type /Catalog /Pages 3 0 R >> endobj 5 0 obj << /Type /Font /Subtype /Type1 /Name /F1 /BaseFont /ZapfDingbats >> endobj 6 0 obj << /Type /Page /Parent 3 0 R /Contents 7 0 R /Resources 4 0 R >> endobj 7 0 obj << /Length 8 0 R >> stream q Q q 49.00 258.21 123.52 87.29 re W n Q q 0.00 0.00 395.25 394.50 re W n /GS1 gs 0.000 0.000 0.000 RG 0.75 w [] 0 d 1 J 1 j 10.00 M 70.08 258.21 m 156.96 258.21 l S 70.08 258.21 m 70.08 252.23 l S 91.80 258.21 m 91.80 252.23 l S 113.52 258.21 m 113.52 252.23 l S 135.24 258.21 m 135.24 252.23 l S 156.96 258.21 m 156.96 252.23 l S BT /GS257 gs 0.000 0.000 0.000 rg /F2 1 Tf 10.00 0.00 0.00 10.00 60.21 236.69 Tm (-0.6) Tj /F2 1 Tf 10.00 0.00 0.00 10.00 103.65 236.69 Tm (-0.2) Tj /F2 1 Tf 10.00 0.00 0.00 10.00 150.01 236.69 Tm (0.2) Tj ET 49.00 270.41 m 49.00 339.93 l S 49.00 270.41 m 43.03 270.41 l S 49.00 284.31 m 43.03 284.31 l S 49.00 298.22 m 43.03 298.22 l S 49.00 312.12 m 43.03 312.12 l S 49.00 326.03 m 43.03 326.03 l S 49.00 339.93 m 43.03 339.93 l S BT /F2 1 Tf 0.00 10.00 -10.00 0.00 34.66 264.71 Tm (-2) Tj /F2 1 Tf 0.00 10.00 -10.00 0.00 34.66 295.44 Tm (0) Tj /F2 1 Tf 0.00 10.00 -10.00 0.00 34.66 309.34 Tm (1) Tj /F2 1 Tf 0.00 10.00 -10.00 0.00 34.66 323.25 Tm (2) Tj /F2 1 Tf 0.00 10.00 -10.00 0.00 34.66 337.15 Tm (3) Tj ET 49.00 258.21 m 172.53 258.21 l 172.53 345.50 l 49.00 345.50 l 49.00 258.21 l S Q q 0.00 197.25 197.62 197.25 re W n BT /GS257 gs 0.000 0.000 0.000 rg /F2 1 Tf 10.00 0.00 0.00 10.00 47.61 212.79 Tm (Mult\(Exp\(.\), orig:dest\).educ2) Tj /F2 1 Tf 0.00 10.00 -10.00 0.00 10.76 299.35 Tm (z) Tj ET Q q 49.00 258.21 123.52 87.29 re W n /GS1 gs 0.000 0.000 0.000 RG 0.75 w [] 0 d 1 J 1 j 10.00 M 107.60 298.22 m 113.93 298.22 l S 110.76 295.05 m 110.76 301.39 l S 53.58 261.44 m 57.15 263.38 l 60.73 265.37 l 64.30 267.41 l 67.87 269.51 l 71.45 271.65 l 75.02 273.84 l 78.60 276.08 l 82.17 278.37 l 85.75 280.70 l 89.32 283.08 l 92.89 285.51 l 96.47 287.97 l 100.04 290.48 l 103.62 293.02 l 107.19 295.60 l 110.76 298.22 l 114.34 300.86 l 117.91 303.54 l 121.49 306.23 l 125.06 308.96 l 128.64 311.70 l 132.21 314.46 l 135.78 317.23 l 139.36 320.01 l 142.93 322.80 l 146.51 325.60 l 150.08 328.39 l 153.65 331.18 l 157.23 333.97 l 160.80 336.75 l 164.38 339.51 l 167.95 342.26 l S Q q 246.63 258.21 123.52 87.29 re W n Q q 0.00 0.00 395.25 394.50 re W n /GS1 gs 0.000 0.000 0.000 RG 0.75 w [] 0 d 1 J 1 j 10.00 M 260.61 258.21 m 364.24 258.21 l S 260.61 258.21 m 260.61 252.23 l S 295.15 258.21 m 295.15 252.23 l S 329.69 258.21 m 329.69 252.23 l S 364.24 258.21 m 364.24 252.23 l S BT /GS257 gs 0.000 0.000 0.000 rg /F2 1 Tf 10.00 0.00 0.00 10.00 250.74 236.69 Tm (-1.5) Tj /F2 1 Tf 10.00 0.00 0.00 10.00 285.28 236.69 Tm (-1.0) Tj /F2 1 Tf 10.00 0.00 0.00 10.00 319.82 236.69 Tm (-0.5) Tj /F2 1 Tf 10.00 0.00 0.00 10.00 357.29 236.69 Tm (0.0) Tj ET 246.63 265.03 m 246.63 341.28 l S 246.63 265.03 m 240.65 265.03 l S 246.63 280.28 m 240.65 280.28 l S 246.63 295.53 m 240.65 295.53 l S 246.63 310.78 m 240.65 310.78 l S 246.63 326.03 m 240.65 326.03 l S 246.63 341.28 m 240.65 341.28 l S BT /F2 1 Tf 0.00 10.00 -10.00 0.00 232.29 259.33 Tm (-2) Tj /F2 1 Tf 0.00 10.00 -10.00 0.00 232.29 292.75 Tm (0) Tj /F2 1 Tf 0.00 10.00 -10.00 0.00 232.29 308.00 Tm (1) Tj /F2 1 Tf 0.00 10.00 -10.00 0.00 232.29 323.25 Tm (2) Tj /F2 1 Tf 0.00 10.00 -10.00 0.00 232.29 338.50 Tm (3) Tj ET 246.63 258.21 m 370.15 258.21 l 370.15 345.50 l 246.63 345.50 l 246.63 258.21 l S Q q 197.62 197.25 197.62 197.25 re W n BT /GS257 gs 0.000 0.000 0.000 rg /F2 1 Tf 10.00 0.00 0.00 10.00 245.24 212.79 Tm (Mult\(Exp\(.\), orig:dest\).educ3) Tj /F2 1 Tf 0.00 10.00 -10.00 0.00 208.38 299.35 Tm (z) Tj ET Q q 246.63 258.21 123.52 87.29 re W n /GS1 gs 0.000 0.000 0.000 RG 0.75 w [] 0 d 1 J 1 j 10.00 M 309.71 295.53 m 316.05 295.53 l S 312.88 292.36 m 312.88 298.70 l S 251.20 261.44 m 254.21 262.58 l 257.22 263.77 l 260.23 265.00 l 263.24 266.29 l 266.25 267.62 l 269.26 269.01 l 272.27 270.45 l 275.28 271.94 l 278.29 273.49 l 281.30 275.10 l 284.31 276.76 l 287.32 278.48 l 290.33 280.26 l 293.34 282.10 l 296.35 284.00 l 299.36 285.97 l 302.37 287.99 l 305.38 290.08 l 308.39 292.22 l 311.40 294.43 l 314.41 296.69 l 317.42 299.01 l 320.43 301.40 l 323.44 303.83 l 326.45 306.32 l 329.46 308.86 l 332.47 311.45 l 335.48 314.09 l 338.49 316.77 l 341.50 319.49 l 344.51 322.25 l 347.52 325.04 l 350.53 327.86 l 353.54 330.71 l 356.55 333.58 l 359.56 336.46 l 362.57 339.36 l 365.58 342.26 l S Q q 49.00 60.96 123.52 87.29 re W n Q q 0.00 0.00 395.25 394.50 re W n /GS1 gs 0.000 0.000 0.000 RG 0.75 w [] 0 d 1 J 1 j 10.00 M 67.42 60.96 m 164.72 60.96 l S 67.42 60.96 m 67.42 54.98 l S 86.88 60.96 m 86.88 54.98 l S 106.34 60.96 m 106.34 54.98 l S 125.80 60.96 m 125.80 54.98 l S 145.26 60.96 m 145.26 54.98 l S 164.72 60.96 m 164.72 54.98 l S BT /GS257 gs 0.000 0.000 0.000 rg /F2 1 Tf 10.00 0.00 0.00 10.00 57.55 39.44 Tm (-2.5) Tj /F2 1 Tf 10.00 0.00 0.00 10.00 96.47 39.44 Tm (-1.5) Tj /F2 1 Tf 10.00 0.00 0.00 10.00 135.39 39.44 Tm (-0.5) Tj ET 49.00 65.60 m 49.00 144.92 l S 49.00 65.60 m 43.03 65.60 l S 49.00 81.47 m 43.03 81.47 l S 49.00 97.33 m 43.03 97.33 l S 49.00 113.19 m 43.03 113.19 l S 49.00 129.06 m 43.03 129.06 l S 49.00 144.92 m 43.03 144.92 l S BT /F2 1 Tf 0.00 10.00 -10.00 0.00 34.66 59.90 Tm (-2) Tj /F2 1 Tf 0.00 10.00 -10.00 0.00 34.66 94.55 Tm (0) Tj /F2 1 Tf 0.00 10.00 -10.00 0.00 34.66 110.41 Tm (1) Tj /F2 1 Tf 0.00 10.00 -10.00 0.00 34.66 126.28 Tm (2) Tj /F2 1 Tf 0.00 10.00 -10.00 0.00 34.66 142.14 Tm (3) Tj ET 49.00 60.96 m 172.53 60.96 l 172.53 148.25 l 49.00 148.25 l 49.00 60.96 l S Q q 0.00 0.00 197.62 197.25 re W n BT /GS257 gs 0.000 0.000 0.000 rg /F2 1 Tf 10.00 0.00 0.00 10.00 47.61 15.54 Tm (Mult\(Exp\(.\), orig:dest\).educ4) Tj /F2 1 Tf 0.00 10.00 -10.00 0.00 10.76 102.10 Tm (z) Tj ET Q q 49.00 60.96 123.52 87.29 re W n /GS1 gs 0.000 0.000 0.000 RG 0.75 w [] 0 d 1 J 1 j 10.00 M 121.12 97.33 m 127.45 97.33 l S 124.28 94.16 m 124.28 100.50 l S 53.58 64.19 m 56.18 64.63 l 58.78 65.10 l 61.38 65.61 l 63.98 66.16 l 66.58 66.75 l 69.17 67.38 l 71.77 68.05 l 74.37 68.78 l 76.97 69.56 l 79.57 70.39 l 82.17 71.29 l 84.77 72.25 l 87.37 73.28 l 89.97 74.38 l 92.57 75.56 l 95.17 76.82 l 97.77 78.17 l 100.37 79.60 l 102.97 81.12 l 105.57 82.74 l 108.17 84.46 l 110.76 86.27 l 113.36 88.19 l 115.96 90.21 l 118.56 92.32 l 121.16 94.54 l 123.76 96.85 l 126.36 99.26 l 128.96 101.76 l 131.56 104.35 l 134.16 107.02 l 136.76 109.76 l 139.36 112.57 l 141.96 115.44 l 144.56 118.37 l 147.16 121.34 l 149.76 124.35 l 152.35 127.39 l 154.95 130.45 l 157.55 133.49 l 160.15 136.50 l 162.75 139.43 l 165.35 142.28 l 167.95 145.01 l S Q q 246.63 60.96 123.52 87.29 re W n Q q 0.00 0.00 395.25 394.50 re W n /GS1 gs 0.000 0.000 0.000 RG 0.75 w [] 0 d 1 J 1 j 10.00 M 256.84 60.96 m 362.65 60.96 l S 256.84 60.96 m 256.84 54.98 l S 283.29 60.96 m 283.29 54.98 l S 309.74 60.96 m 309.74 54.98 l S 336.20 60.96 m 336.20 54.98 l S 362.65 60.96 m 362.65 54.98 l S BT /GS257 gs 0.000 0.000 0.000 rg /F2 1 Tf 10.00 0.00 0.00 10.00 251.14 39.44 Tm (-8) Tj /F2 1 Tf 10.00 0.00 0.00 10.00 277.59 39.44 Tm (-6) Tj /F2 1 Tf 10.00 0.00 0.00 10.00 304.04 39.44 Tm (-4) Tj /F2 1 Tf 10.00 0.00 0.00 10.00 330.50 39.44 Tm (-2) Tj /F2 1 Tf 10.00 0.00 0.00 10.00 359.87 39.44 Tm (0) Tj ET 246.63 62.64 m 246.63 128.34 l S 246.63 62.64 m 240.65 62.64 l S 246.63 84.54 m 240.65 84.54 l S 246.63 106.44 m 240.65 106.44 l S 246.63 128.34 m 240.65 128.34 l S BT /F2 1 Tf 0.00 10.00 -10.00 0.00 232.29 56.94 Tm (-1) Tj /F2 1 Tf 0.00 10.00 -10.00 0.00 232.29 81.76 Tm (0) Tj /F2 1 Tf 0.00 10.00 -10.00 0.00 232.29 103.66 Tm (1) Tj /F2 1 Tf 0.00 10.00 -10.00 0.00 232.29 125.56 Tm (2) Tj ET 246.63 60.96 m 370.15 60.96 l 370.15 148.25 l 246.63 148.25 l 246.63 60.96 l S Q q 197.62 0.00 197.62 197.25 re W n BT /GS257 gs 0.000 0.000 0.000 rg /F2 1 Tf 10.00 0.00 0.00 10.00 245.24 15.54 Tm (Mult\(Exp\(.\), orig:dest\).educ5) Tj /F2 1 Tf 0.00 10.00 -10.00 0.00 208.38 102.10 Tm (z) Tj ET Q q 246.63 60.96 123.52 87.29 re W n /GS1 gs 0.000 0.000 0.000 RG 0.75 w [] 0 d 1 J 1 j 10.00 M 329.73 84.54 m 336.07 84.54 l S 332.90 81.37 m 332.90 87.71 l S 251.20 64.19 m 253.80 64.20 l 256.40 64.21 l 259.00 64.22 l 261.60 64.24 l 264.20 64.26 l 266.80 64.29 l 269.40 64.32 l 272.00 64.36 l 274.60 64.41 l 277.20 64.46 l 279.80 64.53 l 282.40 64.62 l 285.00 64.72 l 287.59 64.84 l 290.19 64.99 l 292.79 65.18 l 295.39 65.40 l 297.99 65.68 l 300.59 66.01 l 303.19 66.41 l 305.79 66.90 l 308.39 67.50 l 310.99 68.22 l 313.59 69.10 l 316.19 70.16 l 318.79 71.45 l 321.39 73.00 l 323.99 74.88 l 326.59 77.13 l 329.18 79.82 l 331.78 83.00 l 334.38 86.76 l 336.98 91.11 l 339.58 96.06 l 342.18 101.63 l 344.78 107.67 l 347.38 113.90 l 349.98 120.03 l 352.58 125.82 l 355.18 131.14 l 357.78 135.86 l 360.38 139.85 l 362.98 142.95 l 365.58 145.01 l S Q q 0.00 0.00 395.25 394.50 re W n BT /GS257 gs 0.000 0.000 0.000 rg /F3 1 Tf 11.00 0.00 0.00 11.00 56.26 361.03 Tm (Profile traces for the multipliers of the orig:dest association) Tj ET Q endstream endobj 8 0 obj 8744 endobj 3 0 obj << /Type /Pages /Kids [ 6 0 R ] /Count 1 /MediaBox [0 0 395 394] >> endobj 4 0 obj << /ProcSet [/PDF /Text] /Font << /F1 5 0 R /F2 10 0 R /F3 11 0 R >> /ExtGState << /GS1 12 0 R /GS257 13 0 R >> >> endobj 9 0 obj << /Type /Encoding /BaseEncoding /WinAnsiEncoding /Differences [ 45/minus 96/quoteleft 144/dotlessi /grave /acute /circumflex /tilde /macron /breve /dotaccent /dieresis /.notdef /ring /cedilla /.notdef /hungarumlaut /ogonek /caron /space] >> endobj 10 0 obj << /Type /Font /Subtype /Type1 /Name /F2 /BaseFont /Helvetica /Encoding 9 0 R >> endobj 11 0 obj << /Type /Font /Subtype /Type1 /Name /F3 /BaseFont /Helvetica-Bold /Encoding 9 0 R >> endobj 12 0 obj << /Type /ExtGState /CA 1.000 >> endobj 13 0 obj << /Type /ExtGState /ca 1.000 >> endobj xref 0 14 0000000000 65535 f 0000000021 00000 n 0000000163 00000 n 0000009192 00000 n 0000009275 00000 n 0000000212 00000 n 0000000295 00000 n 0000000375 00000 n 0000009172 00000 n 0000009405 00000 n 0000009662 00000 n 0000009759 00000 n 0000009861 00000 n 0000009910 00000 n trailer << /Size 14 /Info 1 0 R /Root 2 0 R >> startxref 9959 %%EOF gnm/vignettes/screenshot1.png0000644000176000001440000014216412515147523016054 0ustar ripleyusers‰PNG  IHDR¤3YGœ‰zTXtRaw profile type exifxÚUŽË Å ï®"%øÇ—E‰”^ùψD(s°W+ ów_´ „•¼EG\xzê^¡óĘEYÆ®9y¶I%]5™Î€ìÁ¾úÓ¿4Cdž‡–]OÓšõ +oä’ìÎScß¾.|ôôH†,LR?k ìiTXtXML:com.adobe.xmp ‡EõsBITÛáOà IDATxÚìÝy\Tç½0ðßs–ÙeA\P!¢“ ¢à¾EcLmbÚÜ.IÚ{û¦í›n÷öMoÛtoÚÞô6·Is“6iÜ5&Ä 5¨`¢1 .€Ê¦²Ï 0ë™sÎó¼ «lƒQ£øû~È'0ç<ËYœßyžóœçôÓ(e”QJ©¢(Öp „Bè®e³9!ÇŽpÇN aÀXhˆbb¢úMI`€Ið+w- A­ÜÏ22œ•»> ÃYy€HpXB‚Ûº>›ÜʃîúÏX#ôþlå[ãŽUƒ^Hÿ‚^y8' j‡áœÐäÎàÖ'ÃßÝÝXÿ‚Z¹ëÿ,ÈõY¿Yµò ©Ø5`ÃY¹×ªìÖí?Ħõ\À‚Ú!l¨½ÛÏZ,èÝôúÝ«06Œ•opkÍiè}̆³òΟåü¾ 8­„qŒ&0”2³Éd±„€×ëÃk"„Bèîèx„`¥Ôd2†……2Æp!„B#CXXhGû@ ”âA!„F$UU !‚Ùl ²YÏ£Œ2 X0IB °&c €B€Çó\`øcŒ‚Ç!„º¹BCCÀfsa8¸0¯RY–UUUÒiõø˜Y±á aæH“¾Ñå­nuž¼fïRN§¡g „1¦ÈŠä÷'$$ÄÆÆX­K˜ÃÑj³µ\»VW]]­Õh@#ò<&ÿ…QEe„çù¨Uœ­NLaf£€gÚ½„Q¯Ëíö)*às˜É 0¿ÛëôøÀ !ïiwø˜ÖdÕprÝ”ó糜À¡õµÄȜ٘ͦ!"=¥~¿·WVÇëôzSGO©êuz½~ª2„çµz½Iøeª$¹<’_ „ãE£™BøÎ"=­íA¦åõ¹}^¿¢Ð@&šÐ0s»’Úñ(Œz Õ9òþçº9¡µOIz ï°èë7oø•ïJ¥(ŒŽ×†™y?¥»“«^§×+S•2BA¯%ŠOö« ¯ÑëB ]adXùWgFý^ŸÛHK8žuz³ž'ƒíà+Ü;Nû|n¯,« ôÆ0O<ìsP},vÞ¬PùôËßûõñvµZš57‘Ômá{;¯ú ¯Ókê·Žg<¿ãëØÓוn0ø{ž?2à!î?Âös÷9ÕM!z-‡ß„øhC´ìE‘—EG3ÜÚ M‹Wžb5V¨pxL"cÒª^g”)lY´éÍÏó‚¬*±±1:Îãñºô)¥ŒnåCÇ4~„Ȳ¢×bbcš›qÈ+&¹µ3¿óô¼hÌÿû?ß+ªu(ºQ “’¸r‡FÔ1ÅíÔLXñõ KgŽ ådûå>ØúvÞUÁÄK¾ðß|jÑäبP-íUçüóùMG%Eœºú‡OdM‹Ò•Zk?yý÷ÿ<¥vôY¬éõõP»ùß~|Ä¥±ÎúâW—¦OHeä¼W?üÕ »Ä/þÇ¿e&X €¯©ìÄοï8ÑÊd–ôÈ“KfíÜé¢7¯ŒZÖѽÃS=®áVž'Dõ¸4“V>óä²cMœßv±àÝÿ<Þ¬Õ AenäeÉ<óºúɶzN @½Rø‚g¿´`J\t˜Žê¬9uèDkÂÜŒé1ð5žÞ»ñ•œKœQ䆗ÿ–*Å×åt¡z§ûåmßpô·sC 1½¸é1¨úë¿þÑ•••6ÆTíÝSÝ¢ˆ·Oé¿|WÛyþœz«AñKýbÎ'õbˆšîN·'ðQŸ†Ó^ªo¼ ê4ïᱞqÐ1`¾ÿuÞ(½ân§ŠL¹Ñ§þ¾èÊÙ¦6ªÈe-í¿-¬m”ÔÀ"ÅÝ>”^VTÆ@UhtL´×ë•eY–Y–EQEUÿWº>÷z½11ÑŠªRŽU–ÌiO×@õ?ÿô?‡/·+ ¯írÑÇ}Zü2þ©ÿ÷ÂúYãB9` †'Íê?~¶&Rr)~š2s\t¨–P¿‚ž8ï_žÿÊxæöËnýýÏ}gÙ´(ÚÖTo“áZ¯·ûêGv¶Ú­ -ʃŸE<´0-i”‘YR@ϵ¶+ªjˆ°ˆ’³Í-ƒntòüoþàá(ŸÜsî(ZfÝ›FÙT^¦~Œ{òÇ?~læXuµKë„OÿûséZ—DƒÊÜ­øún”ÊAGBꧡ)³Ædžé8*ËÀ™ÇÎ|dýÂé1UÐEN_û­o¦‰n•†•¿¢ÿ¥!ëL%¯0åk/üûc÷'…ñm6E•8ë¾QLVå¦T…%F»Ï+&{hì£ÿñ˯dN¥*«bX¨ßãWø¶HpŸ3Ö9[-uµÙ­-õ­þ®óÀ×nw´Úëí^ÚëjºßÒ•RïÅùÔNŒžé»O`Žy8Õ‡þ§‡?øƒ?wíO€Ðõ@€PU£#Ôía”@Êè°ï¯š÷ûŽ<š¶«ºíÛË禒vêt¥jœQOU•¨j2ý~麞ƒ@7~ç'*!DU£ÁÀ(z*gUå£G4–:D½†ç€¢ •}æ9„x ÿú“?ä·E-ûþK_ž0fåÚÔ}.Ty´üé÷_«´¬øùo¾2ΘžO.\æ"ãFqPõ÷Ÿüj“Ê Z³±«»ñÝŸ½°ãšÊ‰š£Rg&?ùþß*dN§1ñ­¯÷‹¯‰ÆP³1üþïþâ IÖ©S-Û«ÛûV¿OÑ嵬kŒ»¡Ê—]ö†f~i‘àÚ–þ|g:záþûëãg¬L=qLÖv¶_ƒÌ¼k£ "OXßÊ¿Þ˜úƒ?k¦.½öÿ#Ÿ_üóß=I3åÁ8®è¢/l^°ùk5|èœç†¬3•%Ëœ'憸?þã_>éóø/_Z1Ä.-«î¼15h…OUFèº1ä5=ø/+£ä³oýâ7¹µ>Ð… ›;À½\¤?À>Ïwˆ¥ø»üí¢B8^#t¼l¢ìo?~ᄇNK4÷uŸÓŠ×œq}é&BI÷û¨X؃C×g?Tvî‡'°¸Ð>§ºI+œ:¡{¾8Â309¶€¶6O2Ëëf¥þãXñ“÷M¥v¥ÝÞyi@tTܘ§ŒEQ€òñÇŸôÍ;#ãÀÊZßÑÅ?TGD«¸n^¦*BBJ<øÎîý¤U"Ô”wéË&‰ñ)£¹“uÝI5ÄQVî€q£ÕÈ3¢ÔÚ-±&~ýå—žÌËÙ¹ÿD‹ÌÄîõufƒž`jÏ« ­Ùh˜äÕ¥løÎ7–Ž7w/Ò5Ü[Ò»hJã>SåUJbSâb¿øÛ׾صVX¬EPëiG¯xð™kÍFCÏ3¢×!œ»¦´fŽ­IÓ֋H²èC <ë¬Fùë™O ²ÎQ“cÀwþ`©[§ã!¸]ªUáîS‡)Š0-Hgv½&˜MᣊOx‹JLÕŸÚzž!&ƒ§Þ£\ƒÞ,‚ßÉzÞªb¯/1µçÝ 16¨âzï蹺Oྦྷº^O!4 0è‰8F!N¯¤U)cýeµuï–µ=ñÐ}ïžO˜:9TÓùåʹ¼>Bá!Äëñ1¦!Üüù™[õ]í{ÆÀëõRF €×ã#„ =µáû;€FOŸ¶ýP»¨ç9”ªŠÊ‰\w#Éúz]ôȇ¨~5!!œŽ«~ëÇ¿ªYóðà Sf­|nÖ̱?þÙ&[Gu8Bc”‘ëßoÄ€úýÜ}O{éxZûá–/HqeQüࣞ{ÝcÏ3vc•gŒ†5Ú>Ùÿicç%“žpØ0ò‡ ëLXG9 6§s¯d…{ìù®ú0F;&|FÉ [4`ýI¯!¦;³÷þd¬ÏyÔ_é½WÜu'>'0ðúþNõmMÄ€Oæ!4Òq_êüÆqüE§¥ªLU¹¢ÕûÇ ­ÿgÁý«¢Äï/}è¿ÊZÏ;.Fëw¹ÛÝJŸŽFU;ZµG¶ïÉ?š_|MfÇJçáøª<§6”ט 'vnÚöÏw¶mÎÉÛè‚Kà¸ɼïnê, >ÂñÁÔ8NmºØºÔeD0bˆ×t 2ÈåÉ{O+ê@—ºòÁQR›ËÑæWAp‹TeÀú“þJô“þJ— ôŒâA×»¤¾'0eJ§:È” ùOðîâŸî–ýÀ£ñ)ÕˆbÍ=u0EƒÞðÜÂäÉþ&¹Ù–dûÞ’µMULñðšÃ-ŠF4pÀ ‚ÐÒÒl4êUU%„r L†¸aÏ LžK)¡©©Iä… fí%­ýÐ_w>ô‹G'± žþႧ;—øŠ~õ­W/´}²)oå Ùa3ÿõ÷›¿Ñq%S¿o׉~€ˆÀdyôú?½˜eloiñ££@mªsßwò2›>QLÿî½åQ4Þüÿ÷ü–Jîº&u\ªrA´)ñË/½4÷šG?VD”é™E÷&óÚöáWžp¼íØÖ—~/ÔúÔOßzJ–©(rþ“¿þΟ.±3±Ÿ9»®e\ðä4ÁçOx!˜:s¼ÐüÑŽÓËŸ›n˜õÍ_¿óÍku#î±ÏÙ>Ú~rÙwgÒŸýõ¦§e•¹Š×Ÿþù[äßõ¿Ø÷ð²~8#=`ÿ¥ÿ®{6 Ödq½þ¥òÕ>'ð¦+cžì{ª3øÆK„F¶gŸýæ£ñ•‰ñp;_+!S•:ÉS§Ø™ª¨­-IÎÚ8 eªB8®ÈÅŽºxA€ hìv‡Ëí¡”Ê²ì—ý~¿ì÷Ë~ç/²_–eƘÝÑÚêp¢Ô¨BNšrö½—7TÔ»}»ÔÛR]tô¼[xQ-ëÅ_ï<}ÅEµ­æ£M¿{aûU^Ók’²Þ_lœÈ¹¯6yùˆè¨0h¿úéÎ?ÿ¥Ð«Ñ·üå¯{/Øe zïm§}7>pR±å7?©jSM1‰“ÆGé¤öº²sµ>\¬ï±i7Ry¢V:ýÚÏ~½£ð²CEŽº«¨ 0 7ùÃ8‡4œüIPufDÔz>ùóO_Úññ™‹5—ËJO_‘€*2»îÖÀ œø½¶‘h´ÞÂÿyñ¥÷‹kÛUàDžyël¼NÊúß"^3pý»'¸Æ¾þJ¥ÿÒµ¤ç \q½ÿt}O`"ô{ª‹pÄ2þàÏHOæÌ~@£ûçªÒ0Ùù³Hw“©"÷Ó²Äv"þ°ÁìÓEQJ™ì÷«”&%%Zó]sétÏ®ÃqTe—/_âç¹Îùõ‚@©$ɲBUÖÑ:â^§·Áe¿ìó¦TáD­¨ÄFÝ.IaDcÐéP%É%Qà5!FTÙãUU˜—„4¢^ËqŒ©>¯ìWÓ•ƒ(@¯Lߪªßï•U•vÜ¡æxA«5=×ä,ºï»ƒ†[yÒ™Jê¬á8N«àÆ2ïÛ·Óc)ó»}^xΤ!ýV#Øüƒ¨3£L0›åæ§Âˆ&våþü©$ζïçßÞ\§Õ€;$ø ÷ÞNY’%¿ªNQ0ên -¤þ}7–QÉ%ùð©£ãÉnŸGN£5ë¸þK×qÞžùWÜuØ÷æiÿ§:BhóûåU«!sf?σ5Y–´~ßÓ!íéZ ¨Úñ„xàmu_$i_o‘4z­VÛu_€R&I’ª¨11Qƒ¡g§f ê{<žúúžçQ£**Î(‹zÝtö3_|é™Dês{ÁhÔÔnþÞ¯ö8ˆ†Ãs!„‚"ËòªUõŠ[¿$iµZ™ðhåg‰®yzo’ ›9ê¤ÜeE<êÕŸ”M½N+ŠÎD„1Æ¢ÕjeÎõê5“Éd2u:!cÔçó¹œn·Û¥Õèx—ÀopÔÝ{ÑÕUµŒa4H¶K§÷mÙº·´"ÞcF¡á!sf? |0« ¢ÀqB`ê;JUJÇŽãEQE‘ô÷>{€Î‰ØY••ªþ|žãyQ‘1ªª*cßgúiÜûY¥9õ O4QÃãnA¡aPµ«e\W€_!DåyA¯×s²!¸=Ïq„NËiYÇ„ÝŒ0Êc²ì‡Žg˜ 6ÕPß3HÔ^‡ Ï„ºAuã÷l¬Sê—å›Ý‚Ãã€BÝÒ`Ñ!„ñ-{Ü!„ÐöíB¡ìßxómÜ!„ÐÈóµ¯~pú,„Bh„Ã`Ba°G!„{„BÝ›ÁÞ[üÓ¬ì_”ú\~ægYY/–ønn©RÕ–ç§MNN{ì­w×ï¯~ð eÝšš „Bw|°÷—ÿiÁ¤I“Vü­ªsÎ<¥ö«'Mš4ï×çîظè-yí/¥sÞ(¼P¼ý©Ñç»~ú¡UÏ}{iÌï÷ÓÄ.dé€E~!/BÝý-{cœýýœ*?È5{ÞmŒ ¹£_0ê·_ñX§ÄéIïßùˆY¬¹?b€÷¨ ¾!„ÑÁ^7ñÑ¥lﮋø/¿Ÿ#-zl²>왧lËypʤISXóƒÍÜ ¨³ä]4mÒä´¬ùó »Ú„ëþî_¤Mš4iÚ‚¯ÿùc›:`‘Ò•Üß|eáôI“&M¹åóï^U*«OžŠmÿ¿®ú^¡»ô§ËÈX¼vñ’Žßg?þ·³ŸvuÔ÷“nü~êé-üÑܬç_þéW×,[0ûÅßz»ÜË\Ç~ô­muï|yîý÷Ï{zWƒª6úÕ†Ìû’'%O™6{ýK¥^¹æƒ—ÿ°ý²¸Rêþ!„º“ZöÚq¯ÒÜ~Þ ¾²ûùk'è ÜE¿}æ7³þûعsÇ^Ypù·ÏüöSsþö_ÿbÿÂ;…g ·}]yï`£XÙwþ/_ûAÁ”ÿÜWr¾øƒï~ï»ÿ¾¿¹ÿpï-}ùk?8’øƒ÷Š/\8µï¥Ç“d€²úæ™K½’ó»tcêÏö}òñws÷wüþÑæg’4ƒåße zJuùi?߶ïüwÖ]ùóO÷4ª`šó›¿|!2æÉ· >ýôèëDùK^yñÀØŸ~x¶¼¬ôØæ/‰ä†Ãÿ|ûÀU ÷ï!„Ðí# ½Š&aùÚЯl+m^›“g\óF‚öl h^Øyпà_»?\˜ùÕ,Úôç¿ý…]‡é¢?mH5ó`žýoÍÜö»Ž•·î²Ïûõ¿fFkâë™×Wï,veî'Ö_Øú~ëüß~wñX=D$ÏŒ¨¬ïˆ{úÉsùýƒoM¿ù÷\Ú_=- F,úê’8 ÄÏ™þΩkþuQúÞ9^˹ËË.ÔLg‰¿/@xà…g:–züŽÐdëÖ­¸aÖ¯_?Ì`BÜâÇFÿãooØ.YýƱ2ð©Ò^ç ¹?¢£Á¬‰H õo“Úë<¡EºCÌØÐÀmpÅQck-|~é\ `²“Ll÷Ò~‚½ÒzÕ–©ëýaeùúÏsˆé7ÿî¥ýæ©Z€3Z ^"j8Õ§ô}€.åÿüéÛ¿ÿã/¿øçjŸý¥üìß2Gã twxá…p' 4’TTT 7Ø™½>îç?:6í—?æ¡#Ø !1†öj›&ˆào©jÓG‡jC¢ ž†v@¹µÉEã„°XëèØÿ:ðËôžÍao? ]!,ÎÔv¹ÑÈ)K×ož¬uˆ+—þòï±´ßz˜éñÎ@.4íÉ_¼ýä/d[ɦç¿òýß=xø¥Y<ãÐÝùÕ€ºKmݺµïå{pÏÙó£–þfã?6þnEdwSU?ù‘â¡?¼uªUU[O¿õ‡ƒ|öº)–äµóÙ¡kQüÕ9o~ì¦+?¾Ú˜ûË¿¹ê¥ z[*ŽåÛû½g¯Ÿ¼þáÐ#¿þsÞ/¥RKyÑ…Vu€²‚ÎsÈü{,Nž¼i”ÎYsÕØF©úÀÞO¯8&˜­!'è!„º9©ÑÇMhú}ϱlÆûôÚ÷Çæ~#cÊ”‡žÍÿÞkÿ~¿‰3ÍüáËOy^zdñªµ~v!%££Gô)ß~óws*~½fÆäIS\ýß*j †êS¿ý¿¿Ê({qeÚäÉ3–?¿¹ÌÅú/‹Ÿçùw/Nžš„GžÉ¼úÿæLž4ý‰­u²óÂö¯}`jrrÚÊ?¹ûÕ÷fèq€B¡;™3û|Å-B÷²@§vã#4"ÿEí«_^µêœ!„á0Ø#„BìB!„Á!„Bw,wBwµ ç¿»n:-„{€çŸ÷Bw¸?üá0}úôžž>}úºOV®\‰ƒíÂ`ß¿ôôtÜAݱæÏŸß÷Õ+Wž>}wB¨'¼gÐH³~ýzŒ÷! öL+W®ìŠ÷¸7B왺îÍcã!tcÁ^ŒÊx$+VĽ6Ðþ‰ÎúBVœwú|›õظG]gÞñ–´E“š÷Òý&Î4aÁòY£ENkh½j“˜tíØî#5¾ òÓOZ½@Þ¿»ònz Üàu–›J¯-Y1#üvO-ôy6ë»÷× ËG݃´:]ðÁ^Œ~`&œz§¥G£®‹·]c®[¼ãã–a…8"è´ô.ÛcCÔYµ)R¿üPì©=µ~<½ÐçجïjÜãw¡á´ì513&xÎëõ>ØAûÂÓV>šm)έ“-'ðžK‡÷R'ðÖ˧Æbë7ÌRTGÉþÜÒ¯“ÕX§ÌÍN-0 Äw¥ðÃcå­JgÞä‘Ò¡×r¤´E³Ç™u&MSÁŽ}çÛ‘X Ož“•)Qgýu•­Fÿ§»>nVÍ™·ÎX±üqò‘mçÆÎMµpŒpœÚ\’w¤´Ù\˜«²Ø™•£©­Æh>ßf=6îBÃöBXb¬Z{Âl¬Õ^œ{4å¹Ec¿½=ß©gJ^µ.3fãÁ:@µú`sÙÔõËüïí¼ØO—8’ºjuBÙ®m:ÃS—®^ʶï©ð°@úԛ˦nxêÁuµŸÝ»sc³ŸÓš 4é9ÓäeËÇ”ål?Ü®§Ÿ÷ä Ã)2TΪíTÎNÿSßZ¶ÌöÁ¶­*2iÑ#+Ù»ïuÒ¡êÜÁ×P㟛&T7)xv¡Ï³Y{„аƒ=gŒÐK—¼Ã¼-Õä.wªÔU]æN³ˆP']©ÈY3|y[Î9:â¥l/=x,qÝŒˆËÇš{Ô€(õùÛ>¬ñ2 ’ÓÕ‘xTútïÑíkRoíG;Þ7¹•àrfÎÒ}G.ê j{yÞá„ÇfŽ.;ÜdìV½IŸlâ  O.t½øâ‹7v5€Â`ßSá`ÃÊŸu¯Î( 2-o ‹›»öñ‡zv^'¶èy€Á^¶×4xûdɇD¨uŸöêD÷;êìÁæ,Ûkí=/Gü-W•ûÍ<ì áS)žYè6Ú½{7öÕ#„nB°WÛ=º#í·!Ž)öšúš†÷÷×ÜÈ@}ÕÕ*L לóøo(g1|l¸XÞÝý ˆlÅÁ÷hð¦Ò x IDAT§Ñ‰£ñÑm„ w„Ðà‚}Î^m¯ªbñcL7qÕã3ÅX:žÚ'š°øñ±Æ@öª½øc߃‹¦Ö’Îkó˜ûfŒ ªt¥±ðlØ‚¬I¡BgΣR–|quJTÎL›´pþÄ>°0lÊ‚lÃé¢feÈ:wîOcìXR]Õ†ÁÝ^¼‡!„-ûž1´¹ø·*9¬¨Ç3佞³<±ÇsöÄ0.ëáì¤X³ÀŸ“Wéá,Ó–ÌŸÈ-ÙShS¤š¼ãWåÑëòúU&·œ?ÖÐÙm ]ÍÏaÓ³V1”'@eY‘Û«‹Žy‹ykúŠÅ“-aqã{‹§m%ìcó—}ñ~€pœÚZYxp_EG—Ä 90ÿ•#ùuñ ®ESOïÞsÞÙÕ›1X¸I÷ çöàà<„BwŠ=ÿ¼Àó|°Ñ¾ð#çã³bNí¿ÒÙAÞùœ}?˜§2oË«yÝÍxÇ™œ×ÏäôZ‡:˾S>@qÔ{­hﶢ~»lE9›‹©*“šJr·” ?g&ÛÏz÷Ü@‰«3ˆÑ÷§º>ÚÚ‚±!„Ðçù½ðt°æ*;ø)‹·ât¹v“„Ç*Ÿ,sáð<„BwL°ÿÕO|ËTGqîG#w‡„O_¶ì„XúØãéŽ ÷{Î[¥ñäáF<³BÝa-{AF°ÙTûéÝOïÆBhÄx>Üjåx ö!„ÐÈmÙÇ î „Bhd{AlÙ#„B#½e÷ìB¡‘JèèÆÇ`Bì–={„Bhd{á–{1fé³+Z7¾qÂAo[Z!×攋¬Å»6Uº)púøÙëÖ&´å½ò~‰{È´ª½8÷hÊs‹Æ{{¾SΔ¼j]fÌƃýôÔ ÖËÒ[ vÙñýò!„îÍ`ßÿö ÏÙãűçö¿ÒñºwÕUU°¿fâÜ]ïÕ”úüm9§›ý@%g0÷Ø…Ñé©m‡ó+Ý€zk(icè½T“w¸Ü©® ªËÜVKß—ññaS—,Ž(ùàØU 7B¡{‘Àó”ôÑ;Þdå<=»ÀUgC®çz®&Ûk¼lX…ó¦pµÎÞãª@i»j÷+Ö½6£”õmÓ§­\_¹çÃKnìÂG!tï¶ìcf/4Ø«.Œ6r½¢t´Erx?küT6>Öªé›CãÂ5äfm›&rÖê%gß˽Ўý÷!„îé`ßznÇ {æ®(¨›º<#Þˆ÷¼9iÞò„‹U¾ÏZ¸Ò|ªÄ Îs{ˆ W=‘Á+~Ég¿pôhíÜ®Û%†qYg'%Àšþœ¼Jg™¶dþøDné̆œB›Ú±óŠëç®}¢#øó!£äCgO7á¼:¡{u_-«! ɱ.¨ÏÑ]Oày/5 y éºó>XZæ©ÌÛòj^w-gr^?“Óõ·¯üß—ãáEÝ£üÍåç¯I¡‰SCq2tÀóÂcK#î¹ñKòTv©GÍ#„ÐgÁ¤V›OЗͩ0Ü€çù„…O Ÿc°çL³—¥§ÖŸÜ—ïÀÑt!ôYb½Ïnó£U-íJh¸H€©žæÚš†6?hLVŽAÇàè>G#-ØÃí˜`ÔUqh;Î+‰B7ë[Õk³Ë†è£Á/6·´ÉáÕ]WÙ ŒŸ:JÏ\×.VÊ,þ,ø>{„Y±ÞÓÒªÂCNk ½-­ ^{;„FGxBsTTçÓÔ}ŽFdË~ÐGïBÝ5TwK›já ­%L#ÙìFý*¯;æ1á4ÚŽïü>G#-Ø ·ç}ö!„nG¬wµ´«ŒÖž/½æŒ9l¾ÈhNÃQ¿Â@¸/«  |ŽF^Ë^òÑ;„Bw>¦´7;!dlò³@¨ß^UÑdóFG‡‡@u‹3Ú*€ährÒ@PçôýŽF\°Tõs ‡Bè&Åz¹­ÅÍ…Õ‰÷ÞyK¤¹áj‹;:>6atmmù…zQuF³àïXnìÿs4²<ïW)vã#„ÐÝh¬¦Y{$†%ÝÆщ“Gw-ˆí\ïÿs4ÂZö<ÀmöbÌÒgW´n|㄃ޮ´œaLzÖìäp1ŽóÕ|”[pÙ‰¯¾C!t{ùÖ{êkªª æ÷7-­`™8ŽžyãU/ÐDe<º,ͶãT+†{„B÷d°A¸ÕÞ©öS{ÜÖ´Š­ø¨­óÓù‹ÜÜ0áT+ÞŽB!t¯‚{ôŽÓÇLÏÊœF¢ÚÎ9ZÚ$u̸Ì[g,_œ)z-GJ[4{œYgÒ4ìØw¾B×»ë´z«®lÓ›ÇZ®› W™–•Î)*¥íåÇ«&¬¯­>¹wO‘M4-ž¶òÑlKqnå˜Ôh8÷\:¼ÿ“þ'ÖçÍI³&¹ÊÏËxÀBݳ-ûÁ½#3Öf²üœ-W<€7%Î]½&}ߎB»  ÚN}°¹lꆧ\÷XûÙ½;76û9­Ù@;:Ì;Þ]'Æ.\;¶Ï5„9eÅ"kñ®M•n œ>~öºµ my¯¼_â2­j/Î=šòÜ¢±ÇßÞžïT3%¯Z—³ñ`]¯ˆÎ‡OxÝ‚”Ç‘¿ÿ³Âƒ/„@!tO{‘'Î=·÷øO |«®ª‚ý5ç&èz¯¦ÔçoË9Ýì*9ƒ¹Ç.ŒNOm;œ_é¦Ô[{ü@IÛp²T“w¸Ü©® ªËÜV‹Øçšàô®¿ýñ¥¿Ÿ ÉZ<5çD!tÏ4ò&+ßØàé9²MuÖ9táú^Éd{MƒwxMgÞ®ÖÙ{\(mWíþaåÁº×f””zëOç~Ø~߬(5B¡{Ô Óç©.Œ6r m´;JG[$›÷³ŽlW6>Õª¹pµsÔœ®i¹%›H%èµ5Bèvغu+îtW{æ®(¨[¿<Ãöþǵ À›“æ-O¸¸¯Ð÷Y‹UšO•˜Wg&½wø²KΑ=AÇšoÊ&icgÍ´ÔoôQ¢œ±prÃñ­8!tëUTT¬_¿÷º«‚=€ÜðÑÎü V?‘É1FˆÒ\zà½[ÇÐxÞš¾bñdKX\[¿a–  :Jöç–¶wŒ§ÓGÆâO”H× v©õužÛs@\¸ê‰ ^ñK>û…£Gkçvuµ––Æe=œ”køsò*=œeÚ’ùã¹¥3r m €ÔXÑ4vΚ á"S)€§êø{ù×0Ö#„Â`?ê½V¸{sa‹T[QÎæ¢ÆÓÌßT¼wsqç_úä)¤ëÎû`i™§2oË«yݵpœÉyýLNÏnƒÖ‹ï¾ø1\„B`ˆz·ñšÃ’<•]êÿQy„BÝÒ–ý­¼Ð0MÌ^–>J œZr_¾CÅ‚B¤`O]‡¶Wà1@!„nmëwB!„Á!„BìB!tGúôÓ1Ø#„B#¼aÁ!„éÑwB!4²ÝúGïĘ¥Ï®hÝøÆ ½­i8Ó䇟X˜ä=ôÊ?/øðP#„n½‰'âN¸aø0ö-Bo}°§¾¦ªš`Þq“Ó‚nlFJÓ¾5 3B#ÖnëÖ­ø¡[†Üú`¯ÚOí9pûÓjb˜)ü~³1oU „º—±¡ƒ=§™ž•9)Œ0Dµ=r´´Iêh+óÖ˧ÆEJ‡^Ë‘ÒÍgÖ™4M;öo§Ðõî:­Þª+Ûôæ±–ë&ÃÕF¦eg¥„sŠJi{ùñª kçk«OîÝSdSMˇ§­|4ÛRœ[9&5ZNà=—ïÿ¤×Äú¨º¢=×ü€]j!„îi”±!‚½™±6“åçl¹â¡¼)qîê5éûvÚUÕvêƒÍeS7<õàºÇÚÏîݹ±ÙÏiÍÚqƒ½ãÝubìµcû\C˜SV,²ïÚTé¦Àéãg¯[›Ð–÷Êû%î!ÓªöâÜ£)Ï-{üííùN8Sòªu™1Öuö÷sa)³#Î:éÐáQF!tOSU:h71Nœ{nïñ+ž@øV]Uûk&ÎMèB‰RŸ¿-çt³¨ä æ»0:=µíp~¥›Poíñ%mù·.Õä.wªë‚ê2·Õ"vUÇ4qvBõ±rÅ#ŒBèž—’òA[ö¼ÉÊ7VxzMÕYçÐEê¹Ë¾îOe{Mƒwx£àxS¸ZgïqU ´]µûµÃÉ‚u—È(íQ¼hMŠ [¶~2oŽeK¦7í>mÇWê!„ºýãïo ìU—FF¹†6Ú¥£-’ÍûYͪÓƧZ5®ú;›ú¡qáš–›²Uþš=oþ£óÝÄÇVs¹éBÝ«4Í ÝøÌ]QP7uyF¼!°oNš·<ábAÕg~j]i>UbžŸ™dâ8CBFöÃgäB¡›LQÕ!èÉ íÌŸ±`õ™c„(Í¥Þ+±u4“ykúŠÅ“-aqQlý†Y*€ê(ÙŸ[ÚXÜ1ž^NkˆzG½× wo.ìo‘j+ÊÙ\4`ÂÀxúù›Š÷n.îüKŸ<…tÝy,-óTæmy5¯»Ž39¯ŸÉé§zöâÝÝù#„ÐÝÇNoá©ßÙTUUsÕ0e\˜€{ݤ`›ªaIžÊ.åIx@Bh€„ט­mc‹Oa `r[}íµ§Ÿró¨1c£ÌsÕ^¨…Qaþæf—Ì"âF³Æê«vÂ#ÇE™xÔÛrõJƒÃ£¯‹‰cÕùÊ/º£’Ç…‰¨«öB5;eŒ‰ô-û0Ø3MÌ^–>J œZr_¾{ÚBhPŒÊn»Câ‘POÃ¥šöЄäD³ ·]½\]«›œ @ý­íBRò}zÕ^U^}Ñ•0yŒFjºxéj³eR”ŽPW]e7|ÜÔ Fp7VVV^Ó&ÇYÂÅæ—jP­Â õÔ÷-‚áþ.;m>Ï`O]‡¶ã$Ò!Ô÷µ»ælq ÑG á¨×nWBÆD†h8mhTdc¹ÝMCy BhT„‘'À‡Zu×ꌣ-:À®il÷Sн•…Ž4 À4:&´¥ÚæµjmíJX¸HÝ6'6NÏõ_„%”Çr7;€w|Bè®@Œc§Nãdwsues‹;Â"0U’UWÍ…s`LB„;¢1!áEž`„ãcÀ¨Ÿr&±£yN-O]*#¦P«¦ÁÖ®„‡y[œ\Ø8<6íï*ìBè.ŠøœhÕZv­y´)ZË‹¢hI˜oìù5u‘†£’Â0ER9‘'bH„®ÞÞîáÛ\¼%JÏÐ~‹@wµ]kh÷SŒ*¾öv·ÌŒ%œ!<”´Õ5»UÕÝ\ß!V=@Ä«ÎÛP×à,îF‹@wɯ`Ë!„î¶h/˜£¬|Eƒ]4¶áêµòRIeœ¨3YbâÍA5ôL±‰1W®Tž­WׇŌ‹ ÌqF³Õp¥Ö¥¶hIÇuÁ îS§LÂ`Bw~x×GO™Ö3Vëc:ÿÖ„D ‰îµ¶)~êäÎßùФÔЮ.ëÄû¬]ûˆ±“"ú¼’”ˆá㧅_דзtWa»ñB¡‘ÌQ}éÖ·ìŘ¥Ï®hÝøÆ ½]iµãÖ=»ØÐÒ&3æ¾xhoa‹‚Ç!„Ð=‰Üú`O}MU5Á¼ãþf¦•?~w[‰ ‘ „B·áÑ;Õ~jÏÏ!-B!„‚ öœ>fzVæ¤0ÂÕvöÈÑÒ&©£ÅÌ[g,_œ)z-GJ[4{œYgÒ4ìØw¾B×»ë´z«®lÓ›ÇZ®› W™–•Î)*¥íåÇ«&¬¯­>¹wO‘M4-ž¶òÑlKqnå˜Ôh8÷\:¼ÿ“ºžë £f=²aà8µùtî¡R›ŒÇ!„ûþˆ‘k3Y~Ζ+ À›ç®^“¾oG¡]Pm§>Ø\6uÃS®{¬ýìÞ›ýœÖl 7Ø;Þ]'Æ.\ÛgÀ'gNY±ÈZ¼kS¥›§Ÿ½nmB[Þ+ï—¸‡L«Ú‹s¦<·hìñ··ç;UàLÉ«ÖeÆl[Îéf?PÉÌ=vatzjÛáüJ7 ÞÚãJÚ†s‡]ªÉ;\îT×Õen«Eì^H=õÕ î@g€ÚváØå¨iQ"k„B÷¢yðz¼ÉÊ7VxzŽ„Wu]¤ž»ìëþT¶×4x‡7Ž7…«uöWJÛU»_;œ,Xw‰ŒÒAŠW}NfÐâ3†¡ÛbëÖ­¸Ð…'ƒwã«.Œ6r m´;JG[$›—~Æ’U§Oµj.\õw6õCãÂ5-·d+µ£ÆÍEøäBèÖ«¨¨X¿~=ît§´ÁËÜuS—gÄ«ñæ¤yË.Tù>k±Jó©óṳ̈ÀüŒœ!!#{‚ŽÝœ'å V>±ü¾M`ûŒ‰sæÎ_?:!„ºg 1@OnøhgþŒ«ŸÈä#Di.=ð^‰­#pòÖô‹'[Ââ¢Øú ³TÕQ²?·´5°¸c<½œ>2Öÿx¢Ä@ºV°ûH­¨óÜžâÂUOdðŠ_òÙ/=Z;·ë¾ú`i‰a\ÖÃÙI °f?'¯ÒÃY¦-™?>‘[:³!§Ð¦x.æîçš¿þ_Ì RBÛÎÚsÎIñP#„º7ñ>{ê½V¸{sa‹T[QÎæ¢ÆÓÌßT¼wsqç_úä)¤ëÎû`i™§2oË«yݵpœÉyýLN¯K{yþ»åùx|B!v‡Œ[,ÉSÙ¥^Ê#„BèæDÙϱlÎ41{Yú(€pjýÉ}ù¼±ŽB¨`O]‡¶Wà1@!„nmëwB!4‚I×ÜìB!lÙ#„Bƒ=B!„0Ø#„Bèó!à.@õëÈ‘#¸ºÃÍŸ?ÿÎöbÌÒgW´n|㄃ÞÖ´À›fe=”hÂ˵G>(¸†Sö 4LÏ<ó î„îXõõõN§óÎöÔ×TUÌ;îojZMtƺ¥ç÷¾»¹QbxB „º—Ýú`¯ÚOí9p›ÓŠ1sG—îÜUÚŽ/ÀA!„Á~¨8}Ìô¬ÌIa„1 ªíì‘£¥MmeÞ:cùâÔ¸HéÐk9RÚ¢ÙãÌ:“¦©`ǾóíºÞ]§Õ[ue›Þ[Îéf?PÉÌ=vatzjÛáüJ7 ÞÚãJÚ†Ž¥š¼ÃåN5p]P]æ¶ZÄîúˆZÞœ¾ì>ÿ';ÞÙ¸uOYØ‚•)!øŒ!B! öýàMV¾±ÁÓ³7\uÖ9táú^Éd{Mƒwx gÞ®ÖÙ{\(mWíþaåÁº×f”öLÊ|ý(÷±Ü¢‰õT|BMˆÑ!„û~¨.Œ6r½¢t´Erx?ëÝpÕiãc­šMýиp ¹9ÛD=×.µx¤®ÛüLU±!„ûþÏº©Ë3â ÕxsÒ¼å  ª|ŸµX¥ùT‰y~f’‰à ÙtìfÝVWšK*"fO%Ö”™¦Ë—Úp´B¡{Ôôä†væÏX°ú‰LŽ1B”æÒï•Ø:Ú̼5}ÅâÉ–°¸(¶~Ã,@u”ìÏ-m ,îO/§Œ5Ä?ž(1®ì>Rëê<·ç€¸pÕ¼â—|ö GÖÎíºí>XZb—õpvR¬YàÏÉ«ôp–iKæOä–ÎlÈ)´)¶}[ɾ³–<²Þ ˆ‚RÿéÞ#65B! öý£Þk…»7ö·Hµål.0a`<ýÀüMÅ{7wþ¥OžBºî¼––y*󶼚×] Ç™œ×Ïä\¿šT2gÛI<À!„Ðr+[°$Oe—êpB[„Bèö·ìoå…†ibö²ôQáÔú“ûòØ׎B¨`O]‡¶Wà1@!„nmëwB!„Á!„BìB!„Á!„BìB!„Á!„BÃtë½c–>»¢uã'ôö¤ÕM|âY\c{çCûBH¤’÷×ÍqÆ„Bìo êkªª æ÷7+­Òz¡àý³E É„èìÇ’8]B! ö·Œj?µçÀíL«4î1a¿>>Òv±@ÁCBƒý8}Ìô¬ÌIa„1 ªíì‘£¥MRÇ kxëŒå‹Sã"¥C¯åHi‹f3ëLš¦‚ûηSèzwVoÕ•mzóXËu­kmdZvVJ8§¨”¶—¯š°v¾¶úäÞ=E6eд|xÚÊG³-Ź•cR£EàÞséðþOúŸXŸMŠk+ÿDÆ#Bƒ}ÿÄÈŒµ™,?gËàM‰sW¯Iß·£Ð®¨¶Sl.›ºá©×=Ö~vïÎÍ~Nk6ÐŽìï®c®ÛçÂœ²b‘µxצJ7N?{ÝÚ„¶¼WÞ/q™VµçMynÑØãooÏwªÀ™’W­ËŒÙx°®oHç̉c]e§üx BÝ»OŒçÆžÛ{üŠ'¾UWUÁþš‰st½WSêó·åœnö•œÁÜcF§§¶ίtSê­=~ ¤ £ÞRMÞáòÀxêª.s[-b[gJ'•×ãÐ<„BìÀ›¬|cƒ§çHxÕYçÐ…ë{%“í5 ^6¬byS¸ZgïqU ´]µû‡•ë^›QÚoRbˆ¯–_óâaF!„Á~ªËF#£\¯(m‘^ú‹U6>ÖªéÑÔ ×›¼qºØ‰äâe„Bìl<»+ ê¦.ψ7VãÍIó–'\,¨ò}Öb•æS%æù™I&€3$ddOÐ1vs·M;I¼\ãfx”BÝÓ† '7|´3Æ‚ÕOdrŒ¢4—x¯ÄÖ14ž·¦¯X<ÙÅÖo˜¥¨Ž’ý¹¥­ÅãéEàô‘±†øÇ%Òµ‚ÝGj}@çö®z"ƒWü’Ï~áèÑÚ¹]·ÝKK ã²ÎNJ€5 ü9y•Î2mÉüñ‰ÜÒ™ 9…¶Ø –hZQê¤xBa°õ^+ܽ¹°¿Eª­(gsÑ€ ãéæo*Þ»¹¸ó/}òÒuç}°´ÌS™·åÕ¼îZ8Îä¼~&§ŸÞƒú‚Ýõx„BÝóî¹ñKòTv©‡Í#„B·¿e+/4L³—¥§ÖŸÜ—ïÀImB¡ì©«âÐö <!„Эm]ã.@!„0Ø#„Bƒ=B!„0Ø#„Bƒ=B!„0Ø#„Bh˜ný£wbÌÒgW´n|㄃޶´¼yÂœ%ŒÑ1FÀ}ùØ5œ5!^œ† IDAT„û[…úšªj‚yÇýMKË[ï_™æ:¸õ@3çÑe)Ž]%íîBÝ›n}7¾j?µç@…‹Þ¾´šQcáÂ…–À5‚Ôpî’f¬UÀCB[ö] èc¦geN #ŒQmg-m’XgzÆòÅ©q‘Ò¡×r¤´E³Ç™u&MSÁŽ}çÛ)t½»N«·êÊ6½y¬åºÉpµ‘iÙY)ᜢRÚ^~¼jÂÚùÚê“{÷Ù”AÓòái+ͶçVŽIxÏ¥Ãû?éžX_ª;ë\8=¾òx­‡ò!ã¦ÇÛ/œ•ñP#„Â`ß/12cm&ËÏÙrÅCxSâÜÕkÒ÷í(´«ªíԛ˦nxêÁuµŸÝ»sc³ŸÓš ´£%Þñî:1váÚ±}®!Ì)+Y‹wmªtSàôñ³×­MhË{åý÷iU{qîÑ”ç=þöö|§ œ)yպ̘ë:":m?—ûÑœ§¾õ®ÅbuycÓ%¾Ô!„Ð=kÐn|bœ87öÜÞãW:†·©®ª‚ý5ç&èz¯¦ÔçoË9Ýì*9ƒ¹Ç.ŒNOm;œ_é¦Ô[{ü@IÛp²T“w¸Ü©® ªËÜV‹Ø½M¦äÅóµGþúòßÞüëýoQÄâãô5B! öýàMV¾±¡×@vÕYçÐ…ë{%“í5 Þá5yS¸ZgïqU ´]µû‡•ë^›QÚ#):y=QpÉ©€â8ô”!}¼£=B! öýP]6mäzEéh‹äð~Ö‘íªÓÆÇZ5=šú¡qáš› ¯!ªÚý©¢ò"Á!„ûþÏº©Ë3â ÕxsÒ¼å  ª|ŸµX¥ùT‰y~f’‰à ÙtìæÜXWeUáÓ#5D÷Àƒš —øàB¡{Õôä†væÏX°ú‰LŽ1B”æÒï•Ø:†ÆóÖô‹'[Ââ¢Øú ³TÕQ²?·´5°¸c<½œ>2Öÿx¢Ä@ºV°ûH­¨óÜžâÂUOdðŠ_òÙ/=Z;·ë¶û`i‰a\ÖÃÙI °f?'¯ÒÃY¦-™?>‘[:³!§Ð¦€jûtwáìÅ_xÒÀ€€óbþ¾Ò6Œõ!„0Ø„z¯îÞ\ØoÚV”³¹hÀ„ñôó7ïÝ\Üù—>y éºó>XZæ©ÌÛòj^†ü™œ×Ïä\WveÁ»•xxB¡;fn|Á’<•]ê~T!„B·­e+/4L³—¥§ÖŸÜ—ïPñ€ „B#)ØSWÅ¡íx B¡[ÛºÆ]€Ba°G!„{„Ba°G!„{„Ba°G!„Ð0 çÑ;>tê¢eÉÚ¯¿Uê¾-µc–>»¢uã'·wºÛ[S®ž²paªEàxÎ_÷éÁüòVO@„BwVË^m;·×áZßÍy_ è'­^9N;è*Ô×TUÓóM¸wo¹œiòŠe‘å»·nÞ´ñ-‡Z’W̉ñüC!t‡û›Œ:-?Äå…ýÔž.:ÊåtBÕ¡‚*Pç>®Žš:£=B¡Û`Èn|1|ʼ…Ó#U–üÞæÒ½–j¬Sæf§áâ»Røᱞ}Óœyü¼åó&˜O¥>‡Gã=¹#¯Nà­3–/NÛùº¼ë^˜]ï½Óê­º²Moké1‘.ž¶òÑlKqnå˜Ôh8÷\:¼ÿ“î‰õùÐ)YY“Ã4‚ h4ÄyùØã5 û™Êàô1Ó³2'…Æ€¨¶³GŽ–6I,˜Z)-¥%=w\˜•8/à»øB}îÁž3%/[_‘³õP›òÿÙ»Ó¯¦²}oôs­••†$ ( Ò(-ªh"b)¬½ëî;NyîxÆ>µëŽϽ÷Å©àìóÏ9µÇ©}w•V‰M‰” MìAA„ ’$1$„4«¹/K (…¿Ï+`%™s5¬ïš3s­‰!ŽÊ,Ê #ïÌ,õM,VFª/k4Q!R’X ,`ÏWklîŽ~ž|¿‚w绯‡ì!ÄØ_ƒyZΆû?ý Ž?~ÀYyññ‚³ßxæ½#Ãö•D¼Òìî¬kIø8C¹œNr³ø!!çr‹Ä„1a=5ç‡Ý½ˆ¶¶ÕJgD>ºüľx­fûüâó÷>¬º6süXÿ°'Äô³{óSÒe|jpr=K…¿ðŒ’²æå!Á''Bî.rcûù³¾Á2yÌž4™TŒ½P77ÜÓ­TısãY†™û…¸çH¾øAÍ-'‹!N(-Z‘ Qñ\c›ßõN[FM|©bg©Õl›>©¨ DóS}ÿtáxž¶L;¹½Ã³yOJd\½ûgʨÕiõ—kµ¯‰oŒ+²ÖI]—®¿ë&Â…±Êc#ß]Ó­ê g„¬Üpý‚Ú0Ó§ïtP,w%>™¶iˆ×›™¹üñw¦—˜Û\iZq®ogeÝHzkÈëh|j¼ã@‘»ÃÏ}I@øÆdçÇ gÃÏØyÓ¾7oW›¹roÙ™²UäùLA´²<3œ?»Pè'd󾥦mvQ¨?9seà'‹ .ÿæÆn¦$Á"÷p{\$Ï,-ŽñÁ^ í·,—ҴƦË|pO¿Ç¶¬ÂÈÇmƒö%¼ã‡t8›ïR£'éÉàÔä €Ö¹e˜)õÕ+xV~Ùn’¡(Ú5ÑUS')S”(yW«Û'(ÇHk›¬PžØD`ˆq¹(×äP‡j¦£›uYMltÁÉ_Œ¦YœpŒÜ­l›kÖ;´M·£”¿;EN[§4ëšèUé=oõŒ‰'.†ùÈÊä9žµ]i~jG˜ÏVÅ¡œm‘èp®³ªiÀ†ûïÊÏŽ’ã»õU튵ô6tç–‚±wé»TõCG2ŽâÕÔt¨e”‹réo\lMÉU–gâ,‹aÔxw}åCZ¬Vˆ‘ýA0©Ë()÷„?á»Ùuíу1x®€õ {„ËØ{­¢wÞ~þ“zÞÕÀô³Žšs ¾sZSy^ãõRÂÒ×pºoÁ%î1ñ 7°mMgÿ£i®¡nêªúkWÕ¼–û‹žÚï{æ½£ëOV Ü™õm¿òCû›×ÊÞwú}p¸Xðl|Âö€°„= ìa{ÂÆ›<ØŸw =–wû¯ïžZ“Ú‘¡ÿTôâÌÝ6­íÌ1«R.î³%Uña¬„ò8n×Þ¨k{b q¼[aO›{j/9|óW¦hÁve®«öÊ€—9oûØ ÖêZÙU^§r9þ1[™®ËgF¦„¸ÁéG$.Üqà] û†qø¼E¢Ž6Þ¯®ß åR†ÎÃÌ/αÞÇx†çþ 'ƒÖ;ìII\Ö¾d)‡v9œÓãÝ·Ÿ¿´”—‘“Äa1œƒÙ‡ÛU}/ææqÃÅQY…YÑ"Êf§»Éƾ{¡iÔ…R¸?1<"˜=~2F!Úô°¶®ûÅì ¸žùçx‚¾úûoTó¦Æ%$IæøwÖ lI !Î!lý×kïŒÎ¶Ô‰Mq Å?.‡Ãár1ËUý-­{*¾e•‹B¸ 4Y‘¹ÝcY„цGÍ-ÝcvIµšoKÛníëuÁ`½ÃÅ(’iª*®™)„qTfQVygf©ob±2R}é\£‰B‘’Äe{¾ZccBñäû¼;ß}=dG!~ÄþÒÏÄò´áþO?¨ãpV^|¼`zæŸ#Ãö•D¼Òìî¬kIø™¶iˆ×›™¹üñw¦ßfD=ã°!ÒgM¾©©)33“$IÈ{ÀûÉkG25Þñ@ ÈÝáç¾$ |c²óㄳágì¼iß›·+h6²8â-;S¶Š<Ÿ)ˆV–g†óg ý„¬sÞ€7Úf…Îôsc\?YT˜pùýÚŒÝLI‚Eî¯p‘<³´8Æç¥3üÛ–ËNiÚFã Óe>¸§ßc[Vaäã¶AûRêÅ Kû0AÊw¿ãISöíÐßcñWMÓ7nÜhhh˜žžfY&€–ý+É9¥¾zÏÊ/ÛM2E»&ºjê$eŠ%ïjuûåi­b“Ê› 1.åšêPÍtt³.«‰.8™ã‹Ñ4‹Ž‘»•-csÍz‡¶év”òw§Èië´“f]½*½ç­ž1ñ$ÂÒ0Y™ÜÁ"dz¶+ÍOíóÙª8”³-ÎuV5 Øpÿ]ùÙQr¼`·¾ªÝ@±–Þ†î¼ÂòO0ãà.}—ª~èHÆÑC¼ššµŒrB.ý‹­)¹ÊòLœe1Œﮯ|h B‹Õ !ÇsÍXÄG‡OJH–f² Þªl…¬_£–=Bhxx¸¹¹Y¡Pðù|hß ìÁeì½VÑ;ï?ÿI=ïj`úYG͹Žß9­©<¯ñz)aék8Ý·à÷˜ø…Ïݶ¦³ÿÑ4×p3uUýµ«j^KîEOí÷=óÞÑõ§G+PîÌú¶_ù¡ý-j…¨o^y|¹õqãÆ={ö$™‘‘Áãñ ïïžÞ[·nmkkëíí½}û¶Ãá€þ|Àû#€aÞ4Mïر£¹¹ùÁƒíííN§òðžØLB؃÷EQ,ËîÚµ«¡¡áÎ;<€¼¼'bR7C؃÷¥eOÓ4B(--íêÕ«7nÜèîîv¹\÷€ O˜˜aÞ—–½;ìB}ôÑ¥K—š››{{{!ïÞ ñv{ð¾´ì)Š¢(Ï­Ÿûöí«¨¨hllìë냼llþ<¼O-ûÙÆ=B¨¨¨è›o¾©¯¯òä‰û}ØJ€ éJ—¼_-ûÙÆ=Bèã?þË_þrõêÕÁÁAš¦!ïÒP·Ž[¼'a? Çñ;w ¸ÙíöË—/Ÿ8q",, ¶Øx£oÒ²'6Åœøì‹O…kUA2´à÷Ÿíõ_óî‡Õ,í8|ꟿüÍ>€kÈݦ=}ú4˲ÿøÿ¸iÓ¦ììì‚‚‚#GŽq8†a`C6fâÖ›ämtý©}…ú:Û•·ò¼WÐ>6¨5®ð$°ëU®?"=aìjã°zŒ×ºeÿìÙ³}ûö}ñÅ&“éÔ©Sv»Çq¹\½mÛ¶ÀÀ@‚ `C6¤uüÎãðy‹œ\iãýêz•Ùå"„7tÏn×Íöq ¢~ŒŒ‹D"‰Dòç?ÿ™ÇãýáøùçŸY–åñx<&Àl`‹~gOJâ²ö%K9´Ëáœï¾ýüåð ˆËÈI â°ÎÁìÃíª¾sãŸpqTVaV´ˆ²ÙiÆn²q§ï^hu!D¤îO fŸL£Bˆ6=¬­ë~1;VÚ3ÿOÀWÿjbÞÔ¸„$éàÑÿκ-‰!$Â9„­ÿzíQÇì 6Å);ü¸‡ËÅ,OTõ·´î©ø–U.Bá‚ÐdEæv?ŒeF5·t9Ø%Õ !ÄÙœ’Îï¨~æD1pä­±cÇŽ¦¤¤Põ§?ýéøÃ_|ñ?þÇÿÐét€$IØD€÷6ìqQì"™¦ªâš™BˆGee…‘wf–ú&+#Õ—Î5š(„)I,P°ç«56w»•'߯àÝùîë!;Bñ#ö—ÆxZN´áþO?¨ãpV^|ìX¨dÏüsdؾ’ˆWšÝu- ŸçEÜúö|«…F¸(¶¸43ôLè§ß6÷6VözÚθ_ÊÑl™®fȹÜr"¥é%™lkÕÙaƒ!’g(§^½Ðn¤—P+„û%|Øsíî4Bðuýš NJJ’J¥AýùÏþ—ù—;w>}úT&“q8hÖ60¯ÝøœÍ)ÉÓÍMj³»±N[úU­O)ÜsRäHÓRìMõ=&OSÞeìnPái)3}ä”IkŠOˆ ôåÈ®­?Ýðl¥¾wh›®÷Yhw>©§üçšf?(.ûð‰òòòÇ>>œ»9@´7`˜Œ°žš[Ãî^D[Ûjµ1‘ü¥Ô Å|9¤ê³Â°5†aXvvvZZZpp°»ŸžžþÇ?þ177÷Ñ£G¾¾¾°‰ïwËžÒÏî9çýÅe|jpr=K…¿ðŒ’²æW >9! rw‘ÛÏŸõ –Écö¤É¤bì…º¹ážÎ±B5ŸwO4Ë0s¿p÷É?¨ù±Åàd"Ä ¥E+R ! žklóÚ¶ŒšøRþÄÎ,R+Dl“úù8¾Ã½aeal~òØ•FŽÁUFÄÞ½{1 s+aØ–-[:400™™ßÖÞï°§-ÄÎ@nïðlÞ“YWïi¸µ:­þr­ö5ñqÅBÖ:©ëïÒõwÝD¸0Vy,cä»k:jUOíþ±rÃõ jÃLŸ¾ÓA±Ü•ødÚj`¤!B\ofæò?Äßa˜^JcÝ©­þæÿ›ù…󱯃¤_«–=—Ëuÿ0ÿ[·nݲe BˆËåâ8f‹BìÔãk5í‚Þ¥°§Í=µ—>‡ù+S´`»2×U{eÀËœ·Œ}lPku­ì*¯W¹¹žßüñÜC+ ‡€w5ìWÆáói8ÓÆûÕõ¥\à {R—µ/YÊ¡]çôx÷íç/-åÄeä$qX ç`öáöFUß‹¹ži\•U˜-¢lvš±›lÜ黚F])…ûÃ#‚Ùã'Óh„¢Mkëº_ÌÎ癎'à«¿ÿF51or8B’tðhŽgÝÀ–ÄáÂÖ½öÎèlKاPìðãr8.³¥pObs¤i)ö¦ú“§)ï2v7¨ð´”@Âóvʤ5Å'D…úò d×ÖŸnx¶RÙæÐ6]ï³Ðî|ROø“³Ë0~P\öáååå'Ž}|8;vs€h%¾­À„1a=5·†Ý½ˆ¶¶Õjc2"ùKªblº!ýí¹ùYõ$xW0 G €õmÙâ@úÙ=缿¸ŒO N®g©PâžQRöÁ¼<$øä„€@ÈÝEnl?Ö7X&Ù“&“Š±êæ†{:Ç ÕœçÆ2ÌÜ/œÀ=GòÅj~l18Y„qBiÑŠHˆˆçÛü¯ûi˨‰/àOìÌ"µzm·°><¸óÀ*ûïÿW¦×°§-ÄÎ@nïðlÞ“YWïi¸µ:­þr­ö5ñqÅBÖ:©ëïÒõwÝD¸0Vy,cä»kºU½ÝŒð•®_PÏ|N;Ë]‰O¦­F"Äõff.ÿCü†é·¹?·y g¼î¼°ú¼6-©ñŽEî?÷%á“'œ ?cçMûÞ¼]AØK¡ý–å²Sš¶ÑøÂt™îé÷Ø–Uù¸mо”zùD,/Üè¾ìÀ…ò2|º:9þXÞ¿Ìf¦ÔW¯àYùe»I†¢h×DWM¤LQ¢ä]­nŸ #­Ul²Byb!Æå¢\“Cª™ŽnÖe5±Ñ's|1šfqÂ1r·²el®-ëÐ6ÝŽRþî9mvÒ¬k¢W¥÷¼Õ3&žD¸@æ#+“;XäxÖv¥ù©a>[‡r¶E¢Ã¹Îª¦î¿+?;JŽìÖWµ(ÖÒÛÐWXþ ÆbÜ¥ïRÕÉ8zˆWSÓa –Q.BÈ¥¿q±5%WYž‰³,†QãÝõ• 4Bh±Z!d{\W‹}ü#šÁsïµê  €õ{„ËØ{­¢wÞ~þ“zÞÕÀô³Žšs ¾sZSy^ãõRÂÒ×pºoÁ%î1ñ 7°mMgÿ£i®¡nêªúkWÕ¼–û‹žÚï{æ½£ëOV Ü™õm¿òCû[Ô !—±¯õǾV8æ¬1!@Øà×+0ÂØÐÌöÀ†æš‚°68{`#û㿶BØв„=6DØ›â N|öŧ‰ÂµªZðûÏöú¯ùÉ*–Kˆ#?8TV~¢ìäÉÒŒ0VÝ›LôN›{j/9|óW¦hÁve®«öÊ€—9oûØ ÖêZÙU^¯rBÜôÒ‚ÀÞšxî`áàðî…ý Ã8|Þ"3ÁÐÆûÕõ¥\D†~´?¤ûâ¥îI˜À»ö¤$.k_²”C»ÎéñîÛÏ_nªÄeä$qX ç`öáöFUß‹¹yípqTVaV´ˆ²ÙiÆn²q§ï^hu!D¤îO fŸL£Bˆ6=¬­ë~1;å«gþ9ž €¯þþÕü¹` IÒÁ£9þu[CH„s[ÿõÚ;£³-ubSœB±ÃËáp¸\ÌòDUK랊oYå"„.MVdn÷ÃXa´áQsK÷˜»¾h­74%\«j…¤ðN…=.Š=P$ÓTU\3Sâ¨Ì¢¬0òÎÌRßÄbe¤úÒ¹FÑÈÄÝ IDAT…"%‰Êö|µÆæî¢æÉ÷+xw¾ûzÈŽBüˆý¥1ž‰åiÃýŸ~PÇ?଼øxÁîtÏüsdؾ’ˆWšÝu- ŸçEÜúö|«…F¸(¶¸43ôLè§ß6÷6VözºÉq¿”£Ù2]Ís¹å"DJÓK2ÙÖª³Ã6!B$ÏPN½z¡ÝH/¡V¸0H p„ ü0JÄØ ê¶ëzèÌ°ú¼ŽAãlNIžnnR›ÝuÚÒ¯j}JážÄæHÓRìMõ=&OSÞeìnPái)„çí”IkŠOˆ ôåÈ®­?Ýðl¥¾wh›®÷Yhw>©§üÉÙe?(.ûð‰òòòÇ>>œ»9@´ßV`˜Œ°žš[Ãî^D[Ûjµ1‘ü¥ÕŠäâÔ;w.œ>SQ­öË=˜à 7CXç–=!¤ŸÝsÎû‹ËøÔàäz– %~á%eÌËC‚ON„Ü]äÆöóg}ƒeò˜=i2©{¡nn¸§s¬PÍÙ¹F1Ë0s¿p÷É?¨ù±Åàd"Ä ¥E+R ! žkló{ái˨‰/àOìÌ"µBˆµÛ›§®ý­Cï@Ù†nÞN:ãÛÓþºõ¬cØÓ– bg ·wx6ïI‰,€«÷4ÜZV¹VûšøƸb!kÔõwéú»n"\«<–1òÝ5µš+DøÇÊ ×/¨ 3}úNÅrWâ“i«‘†q½™™Ëÿ‡azIqÍØžõOø8f°4Åв°ú¼¦ 5Þñ@ ÈÝáç¾$ |c²óãfo²§7í{óvñ°™+ñ–)[EžÏD+Ë3Ãù³ …~BÖ9oÀm³‹Bgú¹1®Ÿ,*L¸üðcìfJ,r•€‹ä™¥Å1>ØK¡ý–å²Sš¶ÑøÂt™îé÷Ø–Uù¸mо´šQã5&o&Bˆ°[ô¤ß ÍzëÛ²GÌ”úê<+¿l7ÉPíšèª©“”)J”¼«Õí”c¤µŠMV(Ol"0ĸ\”kr¨C5ÓÑͺ¬&6ºàdŽ/FÓ,N8FîV¶ŒÍ5ëÚ¦ÛQÊß"§­ÓNšuMôªôž·zÆÄ“HÃ|der‹ÏÚ®4?µ#Ìg«âPζHt8×YÕ4`ÃýwågGÉñ‚ÝúªvÅZzºó Ë?ÁXŒƒ»ô]ªú¡#Gñjj: Ô2ÊE¹ô7.¶¦ä*Ë3q–Å0j¼»¾ò¡F-V+÷Ö4?¼z;-ÿÈqÉ¡t÷jš 4`eTTTÀF¼v±âLBÒîW|ùå—©©©°xgegg‡††VTT$''{åÁƒ5— …¯¾úêÕ477Ÿ:u ¶3ï,Ng±X¼ÿG?êlÿü7€/€ €°„= ìa{@ØÂxŸ¼ÉܯĦø¼é±¼Ûý{÷ÔšÔŽ -ø§¢gþë¶imŸ!¿ åòcÊÿþ|ræ¹_)Õô¿~xì€cÀ;ö´¹§ö’Ãç0eŠlWæºj¯ x‰;Æ>6¨µºVv•×§\êÅÏm—uè=Ê Éù8ÖÆðŽ…ý Ã8|Þ" gÚx¿º~c”K=蘷Ù%QRÃã6 @ï@Ø“’¸¬}ÉRír8§Ç»o?i)7 .#')ˆÃb8³·7ªú^Ì%.ŽÊ*ÌŠQ6;ÍØM6îôÝ M£.„ˆ€”Âý‰áÁìñ“i4BѦ‡µuÝ/f›ºžùçx‚¾úûoTóÚÀ„$éàÑÿκ-‰!$Â9„­ÿzíÑÙ–:±)N¡ØáÇåp8\.fy¢ª¿¥uOÅ·¬rB„&+2·ûa,‹0Ú𨹥{ÌÁ.©V󛶅›ûî¸àø°îa‹bÉ4U×ÌB„8*³(+Œ¼3³Ô7±X©¾t®ÑD!„HIb²€=_­±±!„xòý Þï¾²#„?biŒgbyÚpÿ§ÔñÇ8+/.ü­5cÕ4œÓ 2l_IÄ+Íîκ–„Ïó"n}{¾ÕB#\[\šz¦aÔž´¹·±²—õTÒ/åh¶LW3ä\n¹‘Òô’L¶µêì°AˆÉ3”‡S¯^h7ÒK¨ÕüÍ*–GXÕ÷pøXu‡^Þ³9%yºù¢Úìn¬Ó–~U«<6Ï“ØiZŠ½élÉÓ”w»TòÒ”À'ªq!„(“ÖŸå3š&íÚúÓÚ«ºCÛt½Ïý7cRO%ù“h&V1~PÜÞôÄPœ¡\N'¹YüƒÐ²£Æd„õÔœv÷ Ú:ØV+=žùèòûⵚ ¹ÕÑ×Có¬-ʽ†=!¤ŸÝ›Ÿ’.ãSƒ“ëY*”ø…g””}0/ >9! rw‘ÛÏŸõ –Écö¤É¤bì…º¹ážn¥2Žeç~d˜¹_8{Žä‹ÔüØbp²â„Ò¢)Ï5¶ù_÷Ó–Q_*ÀŸØ™Ej5ÿ¢ÁGE÷]†ãÀ¸öóö´e‚ØÈížÍ{R" àêÝ?SF­N«¿\«}M|c\±µNêú»tý]7.ŒUËùîšnU‡¥þ±rÃõ jÃLŸ¾ÓA±Ü•ødÚj`¤!B\ofæò?Äßa˜~³ûóøa1Øã~Ö¡hÜëCu¨ñŽEî?÷%á“'œ ?cçMûÞ¼]AkÂ06Ž]¬8“´ûÕe_~ùejj*l#ÞYÙÙÙ¡¡¡ÉÉÉÞ_yðàAFóº¥_}õÕ«/hnn>uêlgÞY:Îb±xÿ~ÔÙþׯ¿‡gãºaðÂØÈþü¯g!ì€ Â€°„= ìa{¼)μ–ØŸw =–wû¯ïžZ“Ú‘¡ÿTôâÌÝ6­íãeW§\BýQþž-|–ÅÐÔUým­ žš àÝjÙÓæžÚKןÚWhÁvåÁ­<¯/aìcƒZ£keWyÊ%þá`’µ¹âôé3g¾»pÿð@‚/ô«xÇÂ~…a>XäòÂx¿º^ce6B¹ÜÍèçŸ'ÜW}O?7"€ €UößþÏÂE㆔ÄeíK–rh—Ã9=Þ}ûùË —‘“Äa1œƒÙ‡ÛU}/æ¦xÃÅQY…YÑ"Êf§»Éƾ{¡iÔ…R¸?1<"˜=~2F!Úô°¶®ûÅì ¸žùçx‚¾úûoTó¦Æ%$IæøwÖ lI !Î!lý×kïŒ:f_°)N¡ØáÇåp8\.fy¢ª¿åé0_V¹!„ B“™Ûý0–EmxÔÜÒ=æ`—T+Çè#˾dÙÀ­§6†ðÝš,3þüÈÇ €UÖvë‰÷°ÇE±ŠdšªŠkf !B•Y”FÞ™Yê›X¬ŒT_:×h¢B¤$±@YÀž¯ÖØÜý<ù~ïÎw_ÙBˆ±¿4Æ3±QÓ!D™´æ ø„(çИÑ4i×ÖŸÖ®XÕÚ¦ë}ÚÏCê©$ÍÄ*ÆŠÛ›žêƒ3”Ëé$7‹rr.·HL“ÖSs~Ø3¬Ž¶¶ÕJgD>ºüľx­pQìþl^óÿúû-4Ç?n_~îÖË Ó÷Vó>ŸÒÏîÍOI—ñ©ÁÉõ,JüÂ3JÊ>˜—‡Ÿœ¹»ÈíçÏúËä1{ÒdR1öBÝÜpOçX¡Ê³s1É2ÌÜ/œÀ=GòÅj~l18Y„qBiÑŠHˆˆçš—ÐÓ–Q_*ÀŸØ™Ej…ðM;v1·/õ[h„eêm¹s0J8Øm…´°šþÛïs½†=m™ vr{‡góž”ȸz÷Ï”Q«Óê/×j_ßW,d­“ºþ.]×M„ c•Ç2F¾»¦£VsÿX¹áúµa¦Oßé XîJ|2m50Ò!®73sùâï0L/e Fp1šžKv†¢ !a`5¶|àu4>5Þñ@ ÈÝáç¾$ |c²óㄳágì¼iß›·+ˆ‡Í4ªÅ[v¦ly>S­,Ï çÏ.ú Yç¼o´Í. õ'g® üdQaÂåßÀØÍ”$Xän‹ä™¥Å1>ØK¡ý–å²Sš¶ÑøÂt™îé÷Ø–Uù¸mо¤K“zP’ž,å"„Æß³—ûó Üg`•‰ã}€3¥¾zÏÊ/ÛM2E»&ºjê$eŠ%ïjuûåi­b“Ê› 1.åšêPÍtt³.«‰.8™ã‹Ñ4‹Ž‘»•-csÍz‡¶év”òw§Èië´“f]½*½ç­ž1ñ$ÂÒ0Y™ÜÁ"dz¶+ÍOíóÙª8”³-ÎuV5 Øpÿ]ùÙQr¼`·¾ªÝ@±–Þ†î¼ÂòO0ãà.}—ª~èHÆÑC¼ššµŒrB.ý‹­)¹ÊòLœe1Œﮯ|h B‹Õ !Dî]iÿpÿ±O|X„!ËãÖ«ÝfÈz«Mý"v±âLBÒîW—}ùå—©©©°xgegg‡††VTT$''{åÁƒ5Íë–VTT|õÕW¯¾ ¹¹ùÔ©S°xgét:‹Åâý?úQgûç¿ù<à Øà ì{@ØÂö€°„= ìð#øÏ@Ø¿„ -øýg{ýWïìH§Q„‘+ú‘˨󪯯WĦø‚Ÿ}ñi¢p­J\¯õ]•rÉÅ1EøkfFÂù›£wç.ÿÇþß? àìØ0¼=ŸH).Ý¿±Ð± M™5­ôŽ_LÔÆØǵVת¥›RÞöñÚ;+ZÀrê¼Êë»ÚÜS{Éás˜¿2'Ø®ÌuÕ^p¬ñú®W¹®±îgùE)’Ó·ô«W~Á¾¶¾ë•êÔb9œïIØÓ†ûu­Ñ$}ù¼ÆŽB¸ üƒÃG÷^;wkü¥YjiãýêúÕkß…ìÙ W62—QçÕ]ß5†qø÷W—Q«Ÿfßhå–^«×bçJd†]‰]Dˆég÷槤ËøÔàô ('„¿ðŒ’²æå!Á''BîŽcûù³¾Á2yÌž4™TŒ½P77ÜÓ9VèøyÝúzÝûËÚKÙû‹ì Ã1–fà¿aÿšSí¦˜j z UÂ?Vn¸~Am˜éÕu:(–;w~6‰ÜŸGf2Œ³)\ÂxõcèÉç6~ G“³çfÚj`¤!B\ofæ Äßa˜^îé{ɵzÛV© |÷ÞX_Zßyó‘Zz­&ˆÜÞáÙ¼'%²®Þs=bÔê´ú˵Ú×Ä7Æ Y뤮¿K×ßuáÂX屌‘ï®éVµÛûÞ_Ö>Z½Oˆù¶çþýï‹7¹‰Ê3‹¶·öY—ÐbeìfJ,r߬Œ‹ä™¥Å1>Øl“yüþCqvæ6÷rÜ'2='šÏ² …ýà +Û"šWOvJÓ6_˜.óÁ=-ßmY…‘ÛíËÝK®Õ[Ddåg}´K4i¥Þ¨VŠÜ~î 3Â7&;?nö&{ÚØyÓ¾7oWofãrÄ[v¦lÙb‚heyf8v¡ÐOÈ:çÅm³‹BýÉ™+?YT˜pùw¶{ßûË)wö>. ‹À†ÍöhÙ#„ˆ€”üLùVìØÉ BcÌ?7^¼?{ßgÌ4‰p4ÌGV&w°Èñ¬íJóS;B¬¥·¡;¯°üŒÅ8¸Kߥª:’qô¯¦¦Ã@!ÆÒS]Oî+.O'(§Ãnü¹¥åi¹`Ôuöàű~óî‹véo\lMÉU–gâ,‹aÔxw}åC=SëÔ¢ý;üýÂg{¿4ÔÛk©•·÷b>[‡r¶E¢Ã¹Îª¦î¿+?;JŽìÖWµ{Zñ¼Ç¡®m~³NtfJ}õ ž•_¶›d(ŠvMtÕÔIÊ%JÞÕêö Ê1ÒZÅ&+”'6b\.Ê59Ô¡šéèf]V]p2Ç£i'#w+[æ Lsh›nG)wŠœ¶N;iÖ5Ñ«Ò3Ë^ßEöþÛ—ë}ï/e/ ÜwûNNOõƒópQtnaZ‰ ‚„‘åÆùLõS³Öç À¯v±âLBÒîW|ùå—©©©kWAì1%qålíÕ*Šâ?.ÛrëïµÃk§Ôëkõ渲CŸ*½ñïßÜ2@›rýay¿M­¸ØcýõiŸZQQ‘œœìý•Ôh4¯[ZQQñÕW_½ú‚æææS§NÁ1À;K§ÓY,ïÿÑ:Û?ÿÍÞ‘'€sücãÙþ…oJg­ê†{¬,€|§jõÆpa°ŸAUÝI¿¾Gš$ŒºÓ ¶Âð<À{uî[DzqQLÎÔÍ„0œÖݽÚjzMҦκï\­Þcºûíßà€[oÔó»×ŸÃf@دƪ¹v^ó®m‘w³VÀÛ·caö€°„= ì°Âþû¿þ „=°‘}óoÿ¹aO†üþ³½þ«wÝ@§Q„­èCu–SçU__¯ˆMñ'>ûâÓDáZ•¸^ë»*å’!ŠcŠð×ÎÉÑ$+++?ùÉ'çl÷ãÀIðëg_äÙøÅ¥û·c£#!–¡)³¦µñÞñ‹©aûØ ÖêZ­ZþIyÛÇkï¬hË©ó*¯ï"hsOí%‡ÏaþÊ|œ`»2×U{eÀ±Æë»^åºÆºŸå¥HNÏ›ka.ÚQt@úðRÅ AÿEEYÎ5ºàLøµóö´á~]k4I_>¯±#„.ÿàðѽ×ÎÝiÚx¿º~õÚw!{v£û§'Vö)³Ë©óê®ïÃ8|³ë»^å"ÚÐÕAÿöƒ°ûÕO9×Îç ^ktÏ!D™zn}œDŽ>ƒ´lä°¥¥5=r»VSš³½ûÇ÷,·žùÉx‚¾úûoT¿dbSœB±ÃËáp¸\ÌòDUKk›;Áó¤I9Š NÑ 3Ùwk0º$›7t·¦ºã¥©Þ¹¡)Ѷ.Õ/fÕÅ¡ÉŠÌí~Ë"Œ6)‰ËÚ—,åÐ.‡sz¼ûöË|åÄeä$qX ç`öáöFUß‹¹mˆ‹£² ³¢E”ÍN3v“;}÷BÓ¨kvCEÌLø‹ ÑòÖ÷õ{YåzßûKÙ ¬u Ó¢H å>úEÚSÝçoV¿Ìò3àêk#8K~õJ?ÉyÓ°ÇI>²¿XJ×&'pÏ‘|ñƒš[ N!BœPZ4w6IèQã¼0e1:y | K³Ža±sï žklóë@[FM|©bŸû«Ë¨ÕO³o´rK¯Õk±s%² îÄ^"Äô³{óSÒe|jpz”B‰_xFIÙóòà“!wG‡±ýüYß`™`#øãÿó?ß0ì‰M11Ô@õ«„¬Üpý‚Ú0Ó«ëtP,wîül ¸?ÌdgS¸„;ñêÇГÏmü@!Ž&gO¼´ÕÀHC„¸ÞÌÌ%@ˆ¿Ã0½Üsó’kõ¶­RAøî½±¾´¾óæ#µôZM;¹½Ã³yOJd\½çzĨÕiõ—kµ¯‰oŒ+²ÖI]—®¿ë&Â…±Êc#ß]ÓQ«ydyßûËÚG+°÷ Q ßöܲP=WšVœëÛYY÷’°¼ÉM̸PžY´}¸µÏº„+c7S’`á~§HžYZãƒÍ6™Çï?ggns/Ç}"Ós¢ù,»Pز²-¢yõd§4m£ñ…é2ÜÓòÝ–Uù¸mоܱäZ½%ADV~ÖG»D“VêjÕñ@ ÈÝá¹é›ðÉΛ½Éž6vÞ´ïÍÛěٸñ–)[g¶˜ ZYžΟ]%yüÐ IDAT(ô²Îy1GÛì¢PræÊÀO&\þíÞ÷þrÊ]½ Ã"°¡Aó/Ãã‡t8›ïR£'éÉàÔä ¸Õ°,rŸ}~¦|+vìä?Ð!„1æŸ/ÞŸ½ïÎ3fšD¸@æ#+“;XäxÖv¥ù©!ÖÒÛÐWXþ ÆbÜ¥ïRÕÉ8zˆWSÓa cé©®'÷—§”Óa7þÜÒò4ƒ\0ê:{ðâX¿Žy÷E»ô7.¶¦ä*Ë3q–Å0j¼»¾ò¡ž©ujÑþþ~á3ƒ½_êíµÎ‹ÔÊÛ{1Ÿ­ŠC9Û"Ñá\gUÓ€ ÷ß•Ÿ%Ç vë«Ú=­x^H‚ŒãP׶¿Y':3¥¾zÏÊ/ÛM2E»&ºjê$eŠ%ïjuûåi­b“Ê› 1.åšêPÍtt³.«‰.8™ã‹Ñ4‹Ž‘»•-ó¦9´M·£”¿;EN[§4ëšèUé™e¯ï"{ÿíËõ¾÷—²î»}'§§úÕÁy¼ˆì‚I]FI¹'ü ßÍ®kŒQpžüÊa+Î$$í~uÁ—_~™ššºvÄSWÎöØ^­¢(þã²-·þ^;ì\óÍóúZ½9®ìÐçŸJoüû7· 0Â{ýay¿M­¸Øcýõ÷Óggg‡††VTT$''{åÁƒ5Íë–VTT|õÕW¯¾ ¹¹ùÔ©SpÌðÎÒét‹Åûô£ÎöÏó‡wäÙøÿØx¶á›ÒY«ºá+ ß©Z½1\ìgPUw@Ò¯ï‘& £î4¨­ð<à½:÷­cÙ¸(&ç@êfBNëî^m5½&iSgÝw®Voˆ1ÝýöopÀ­7êùÝëÏa3 ì× cÕ\;¯y׶ȻY+àíÛ±° €ôöÀFÀ°6´Á$„=°‘Ťn†°62ab"„=°‘½o_‰°'C ~ÿÙ^ÿÕ»n ƒÓ(ÂVô¡:˩󪯯WĦø‚Ÿ}ñi¢p­J\¯õ]•rÉÅ1EøëæäÁ}¶üCщßüæ“O>ùíoK³¶‰ábðëçù.òlüâÒýÛ±Ñ ËДYÓÚø@ïøÅÔ0Œ}lPku­V- ÿ¤¼íãµwV´€åÔy•×w´¹§ö’Ãç0e>N°]™ëª½2àXãõ]¯r]cÝÏò‹R$§ç͵0‹ã³•éº|fdšAˆœ~ô@’áÂýð´=À¯Ú•.»·°§ ÷ëZ£Iúòy!„pAø‡î½vîÖøKSƒÐÆûÕõ«×¾ Ù³Ý?=±²O™]NWw}×Æáó˜uXßõ*ц®ú·„ݯ~úÊ\ ”¡³Å0ó‹s¬÷1žáǹÿ § À¯ÙP·îMž ÇLܮՔælïþ±Ç=Ë­g~2ž €¯þþÕ/™Ø§Pìðãr8.³^ð¢d9ëûú½¿¬r½ïý¥í»^ëÌûq†ž¼–$*ÍM45ÿí»™­¿^ÌÄ­7}6>ã²#þïÎÇøAq{ÓC}p†r9äfñCBN„â¥&š¯_˜bÜ]OoÕ?Œ+X(è„Gÿôü¦&ŒÉë©9?ìé% ­ƒmµÒã‘.?±Ï{¥k=רfBŒÃb]B…—\«×^=i›®÷Yhw^ ©§’üI´ì°çlNIžn¾¨6»“‰¶ô«Zå±yžÄæHÓRìMg{LžÜr»TòÒ”À'ªq!„(“ÖŸå3š&íÚúÓÚ•»Zô²¾^öþr,aï/ºèi“C+ÂÑØÂW´Æ—¾îª Ù•£Ø¥¦g¾³üÚ½iØã$Ù—4b‰¸çH¾øAÍ-'‹!N(-šk<‰$ô¨qÞ ˜2¼>†¥Y Ç0„ع÷Ï5¶ùu -£&¾T€?±ÏýÕeÔê§ß¬Y¶ôZ½;W"Ë0+Ò*$Äô³{óSÒe|jpz”B‰_xFIÙóòà“!wG‡±ýüYß`™uê3¼³t:Åbñþý¨³ýóßüáy(Ç?6ží_ø¦tÖªn¸ÇÊÈwªVo ûTÕôë{¤I¨; j+Œ°¼Wç¾u,ÅäHÝÌAÃiÝÝ«­¦×ä mꬻñÎÕê 1¦»ßþ ¸õF=¿{ý9l„ýša¬škç5ïÚy7k¼};6a{@ØÂöX‡°'C ~ÿÙ^ÿÕ»n ƒÓ(ÂVô¡:˩󪯯WĦø‚Ÿ}ñi¢p­J\¯õ]•rÉÅ1Eø"sòࢇOýó—¿ÙÁ‡3`CXäÙøÅ¥û·c£#!–¡)³¦µñÞñÿ·wßqQ\ëÿÀÏìì.»ô&( JTf—.EE± XbSLnâ/&jŒÉ½Ü˜˜jLL¼)ßä&æÚ¢¨Ä° ‰E×F“^¤nß HÙ¥D„Ïû_/g§œ9g8Ïœ³³ó´J £’–ÜÏ©}li@i· Qé±K=z€î”ù1Ÿo˜ªŒcdº³{( EáŠc±÷d½|¾O긊’ô!3—Ò"Õ;ßÊcTýµ¤VYu9B+w‘1Ų„bÊoœ9›^Ò8á@›yLv±±”%ütXæ4ù>¿äÜþ£™ÕªËÜ~©ÚÙ–6u ›`’vüÞP—!<ÂáÒõwN»ÔC/Öç™:ùNu·ä2 ™\Rš~±å _ùfNÞn\–âp)i^òɤ[Õ!Ç`¤ïtßQúÊz)£’VÖó%—÷Ÿ*P4UÔ°Æô€­’î¯öÖïÖqÛoýδ[{/­ÆßÓŠŸ›­!ÚsyL¤Ä={ô0 ƒ=!„©‹¹¾ƒxµrBQÕŠã÷Š ÏzêÜašVÎlâá=yõ*Bh};ïðÙžG÷'«'f™òÔ?wg^òü„ˆùÕ7ŽÄì*•st tU ·핹£Rµ³-S‘vü¬ó›AÃ.lß—XÃŽ¾ÃÌ«]ñÝž‡æè;L›a+>P¥$„6é3Ãךw©ñSC—™áóì=Y©$„ðL]BÃCÙ}qâzuéØûë\Úñs¶”BÂ#쩦™›?wgŽœ&?s[ãMIwÎW{ëwë¸í·~çZAZ”#÷¶3æf·IsË1vžlž‘pYB¾®€ì‰J!%^ç¦7)…Ó„I.Vº•R!—ó\ç"'„®…§KÕéý÷êTê)ƒÜ '®;…j tzæBÙIó¡¥gïmqd_^Ã,S{ÿÜ1ËHïá7Ý•6[MY˜¸÷dŽ„%„¨d5µ©ŒÎ–JYΩӷju¼ÊΪs3á‘n{î wÉ™˜¬*udbjî$%Ú95Dl®å8é©=• qKQ‘Ÿdáa~7©”!„eeN•Åhç‘òì’ŠÊjiΉ9=võ´w¾í´~wt¢õ;lFR):èsHI«}ëÛOžt)ñ` {O@¤;ÓrÍÇÏ 1¸zä³år–ÚÀ9bFÓ‡´¾)SPѬVVåWÈu4ì†eXŠCQ„°¶5£‹ÅõÍËÀÔT ,…œ»ÒGK9E’®=aÕùRiÅ>:"«RõÈ^´9óàJó(©¨È-—7w_ÚÝÊèt©þ&á0ßß)cô«k•]*UÊU¡ £±úÆŒ6´÷ qjú‘=S‘v^:!hŒ…Ncår †ºz<ÓXcÂQá‹}lMêë±òfŒ©—ê[™ðï ŒmGZëuÿ!ôö[¿;ÇíÖçèY£²ïW!ŒFö„ÚÌ#ÄÇîjÁ’± !„Pªª›'cR›~w×ðÌ4p„–Öº¶‹ìd,‘=8{&WJ[“Ÿ4}ñ³Kq9Š¢kI'²çxÏ›¥säHJ¹’¨j2âNð¦Î\<‰VÊeÒŠ›gÏæzó4†º´ ÎLã”f¿‹Vý“è¾Ø‡Ã²¥,M?qðz9ÓXjÏÁŽ&Æ6{·xÔ»Ý2wPªö¶¥tŸñŸ0b8™(?|ê^=ÇdLˆßH;N¨WÑáä†Q¼Îg[®,ëXb^×&ÑUuYGc9¾!‹¼x*¥’Q”];rÜt‘ÿÜp£qÉeJY~âaÖÝ?|¡M•B¡TTg§$5Nt³ŠÚJvTè’CŠaX-Ë¿|ðl³Ód9§.Ž _¶œ'©•ÈVQ–™T¤êöùvÐúÿ¸í·~gZp E®ÜŒ¸í7\´©Û´W›aVªPÒ?SËqWO;*&z—³›WÛV¯^íééÙ{:,§c÷dÔ·-¢þèù‹†^Øv,OÞëÕ£½T]Ç·õæó–ý°õ¢ÇijznRAtLF?x ÏÏÏÏÊÊ*::ÚÝݽý5ÃÂÂÄb±¶O££££¢¢Ú®pæÌ™åË—ãšè³ kjjÚÿ‹¾‘–üæÒ•}äµ!\‡ÑìÍ?Jgk³â¯°¶f¼>Uª.ãè 6.OŠKA¤²Wš©µòR|¸€Õ÷=Ácsôí¦yâBq˜ÂËG+µÄA¦2íø_}®T]¤ª¼¼ý7\pOš²øòébT Ø÷U­8aŸ¸¯ÕHß,ÀßÇ¢ ìÁìÁìà {žUèŠW&˜<¾ûÞàIsü­{ô¥:Ý)óc?ßvÑF£C¾òÖó.z½uÄ'u¾å¸¼!þ üm´åäÑy&âÍמ[¼Hmá,/s.úè:x7þ̈`U_êe•8ñäÕ"Y«Ô0*iÉýœZÅã*"mâ$*=v©GÐ2?æóíS•qì€Lw¶ gv'…*ŽÅÞ“õòù>©ã*JÒ„Ìð0ÝyQs>;Eñù?ö^¯eÑ5À@ öLyêñÄQ<æÐ>±”B8B›‰³çMHØ{¡´E¦"5îÄãß ïERw–õì[f»SæÇ{¾½Œâ tTOà|ŸÔq S~-…yn¢uj\®ÿ€`ßv¤%É¿xL Jÿ#C=ôiÈO¦#4dý¾5©U@¦œüýù\.—ϧjî&¸Sÿ¨ƒ×±t ðw6å(•ªúÖ…û£æúéd_>—Ò"Õ;ßÊcTýµ¤VC-ŽÐÊÝßGdL±,¡˜ògΦ—4N8ÐfÓƒ]l,e ?–¹M~Æ@ Ï/9·ÿhfµªÃ2·_ªv¶¥MÝÂ昤¿7Ôep¸týÓÇ.õЋõy¦N¾SÝ-¹ŒB&—”¦_lùÂW¾™“w€›—¥8\Jš—|2éÖÃGuÈ1é;Ýw”¾²^ʨ¤•õ|Éåý§ M5¬1=`«¤{竽õ»uÜö[¿3­ÀÖÞK«ñ÷´âçfkˆöÜAãæ,ÃÂá0¥W'¤—+ÐKÀ€ ö„¦J,æúâeÔÊ !DU+Žß+&<ë©s‡iZ9óäÁ̆Ì1ö˜çg[x¤¡ƒå8Ï2K;ðû½:ám'GÌ^uêûC×ëZ•ÏØΚɽ(mëy–“æú°‰‡÷äÕ«¡õí¼Ãg{ÝŸ¬ž˜eÊSÿÜ5zÉó"æWß8³«TÎÑ1ÐU5Üf´WæŽJÕζLEÚñ³Îo »°}_b C8ú3#|¬vÅt;Xpô¦Í°ŽN¨RBŒô™ákÍ»Ôø©¡ËÌðáYöž¬TBx¦.¡á¡ì¾8q½ºÎtì‚ýu.íø9[J!‚aÁöTÓÌÍŸ»³FGN“Œ¹­ñ¦¤;竽õ»uÜö[¿s­ -Ê‘{Ûs³Û¤¹U­Î+ªc¡œÃfN.‹>Sˆp,Ø•BJ¼Î=3E ,œ&Lr±Ò娔 ¹œ7Èà:—9!„p-<]ªNï¿W§ROä^8qÝ)TS Ó3ÊîHší(={oëŒ#ûòf ˜ÚûçŽYFz¿qè®´ÙjÊÂĽ's$,!D%«©íLet¶TÚÈrN¾UèãUvV› t;Øsy¸KÎÄdU©#Ss')ÑÎ!¨!bs-ÇyHOíɨlˆ[ŠŠôø$»ó»I¥ !„(+sª,F;”g—TTVKsNìÌ鱫§½óm§õ»£­ßa+0’J™ÐAŸCJÚ\Þõ…ÙõM÷ª7“îÎ;˜W˜‡h,Øsx"}Ø™ü \óñsB ®ùãl¹œ%„6pŽ˜Ñô!­oÊT4ëC•Uùr »a–âP!ì£mÍèbq}ó205•K!ç®ôÑREEN‘¤kÏYu¾TZ±ŽÈªT=ò˜m`Î<¸Òò©3h(·4EùüÇÆÆÚÛÛkü(::š…Î` ëÊ°…£gç3C”—x«3?LRI«”¦ƒõiõ–úv>3íu©¦!siêu?ŸêÏ9ºÃ'Œ°¬¦`ÿ>k;T¿Y9Ù:ñ¹‚ÑÓ'ÙêrF¾#|§¿}î¾´»•ÑéRýMÂa¾!¾SÆèW×*»Tª”«Bÿ@GcõmhïâÔô#{¦"í¼tBÐ ÆÊå uõx¦±Æ„£ÂûØš>Ô3ÖcåÍÂS/Õ·2á5ÞÛŽ´ÖëþX¶ýÖïÎq{ õ9zÖèìûUm‚½î¨°ÅÓ]ÍùWûoÝki=ü3ÇF[°'„¸»»£§ÀÈ^ûÈÌ#ÄÇîjÁ’± !„Pªª›'cR›~w×ðÌ4p„–Öº¶‹ìd,‘=8{&WJ[“Ÿ4}ñ³Kq9Š¢kI'²çxÏ›¥säHJ¹’¨j2âNð¦Î\<‰VÊeÒŠ›gÏæzó4†º´ ÎLã”f¿‹Vý“è¾Ø‡Ã²¥,M?qðz9ÓXjÏÁŽ&Æ6{·xÔ»Ý2wPªö¶¥tŸñŸ0b8™(?|ê^=ÇdLˆßH;N¨WÑáä†Q¼Îg[®,ëXb^×&ÑUuYGc9¾!‹¼x*¥’Q”];rÜt‘ÿÜp£qÉeJY~âaÖÝ?|¡M•B¡TTg§$5Nt³ŠÚJvTè’CŠaX-Ë¿|ðl³Ód9§.Ž _¶œ'©•ÈVQ–™T¤êöùvÐúÿ¸í·~gZp E®ÜŒ¸’¶7\õ·ãLô‹|Á€0*JU•™—Q£zJþ’5îÕÃz*&z—³›WÛV¯^íééÙ{:,§c÷dÔ·-¢þèù‹†^Øv,¯÷­½T]Ç·õæó–ý°õB9ƒ ï‰áY=7© :&£VõÔŸ‹ŸŸŸ••Utt´zì&‹[{mµZ-**ªí gÎœY¾|9®€>«°°°¦¦¦ý¿èiÉ [¿î#OqMF³w4ÿ(­ÍŠ¿ÂÚšñúT©ºŒ£7ظ<).‘þÉ^i¦ÖÊKñYý ÒkÜcXm ¶øI¾ú›£o0Ís—ŠÃ^>šX©%2•iÇÿês¥ê"Uååí¿áš{ҔŗO÷ßÓ³··o>@Ç·õ@I¸M=É`¯ª'ì÷µJ雥èäà^ýÍ=†õÐdÖ(RÜô·Á=†õÐÜÖM?!ƒ'@Ü_½zõMÊKJÛ ö~~~¨#€§†õÐÄÅ{¼Ö`ÿÕW_¡‚ž ­^—‹·ç@‹¡ûD;Lã<Ý"##Ûù =À6îÁzý”.¾’cIDATFö Uaa!*Áú-??¿¶o‧¦ñìÁìÁž:’kúûo¸.EM ØC—0å~X6ÞQ$9zø„¿öíåªÎ¦_—\òé^ôíÒøÖÓÞ\jÅC›ôeøé]Ÿ£*OxïÿJ–ÿxr©‡©¢ôNʹ› ¦ŒRÕËe¥ÍÇÍ™&èË^{:Fö}Ž¢(í¶ÂiÉ Vz|±µsàÂùîFBaë³ö¼;g‚“Hä4~öÚÝ7ëXB$Éë¼ýßÞµtF ÏôU›ß\±·¸`çsÞcÇú¾r ˆiq+P”ðéW‘ƒÓ˜É‘›Ò%„yAüÆÝD"јÀ—·œ/gj“ֵ܃¦­šM4NãK’×yû¯þvý‹³§N¼bû- ‹¦è Î]¸‹‘}ŸÃ·ï&øßÆÿ§>Â×ÓÅÁÖ„O©?¨Kùbùç·#þ›´Ï¤ýøÊËË¿qôCBd§oŒ>ó‘€IÚÃUüïŽF¹ ZíVrýû '†­?¹-`U“›‘­Ë#ÒÌï^Z{ÎÿÛ£?M6.>±á…·ßûãçß-¸úh’«Q­·ÒFV(v;°w½ {ûç‹ÖÇïŒL£9ž´ô‹·0²ïs(#ßO÷m‰^ùeÝâà .“~›+#„HnÆÄËßyi¬)—kêõâÚ eüþL)!„k2uù,;EÕÞniN]^ÖÍœJÇÀÖÕÅœ+¹} Â÷Ý×}†èÐBÛàË­¯Æ¤Õv´•Öð̃^ ±á¢c;eŠéƒÔr´%@ Cá½>î…Ã^ÿbû‘óé×w,ãXûÖž%QVÔÚ™óÆÿævF’Â*!„6°4èx-pþ›W»øɉÎãÂþ±%±„QVæ”?LXêíííííí·à›{”²Z¢ê`+­àè™èª/'ŠÇç0R%æñú‚×W"Ø÷騯c9nÑ‹žÜÜ´B9áZéVg—7Œ—åe÷«„CŒx„BµÒS„Õf9FnÏ~¼ýø•ôó¿Î—l{gcŠÜØÚÌ"ü‡øsjIÓ.oŸ7˜Órm¶ªG³ùŸÞSÞ¿\ßáq;NÅûïäzUÍõ__#rtóaËÅŠG»Ð°óÖÛâš@°ï«šÂ<ÕÞ;gõ§|þÝK«g·Ÿ»råìç ¦ !òü#Gu^Ý—’qþÿ|o~ýA\1CHC‚Z§Ž^ÏLûsÅÁ·ß;VÚ*ì E ¿Ž½’™™óƒ?Ö®?UÉBˆüAÜ ƒ•‡Rn\úuzî¦×¿I—Æü¹þÿIÊÈHú>ðîË¿¸"³>gpúž *B©Kß{† œï¢Ûáq5쪖% ©x?ŠIH<þÉ8Uò¯W±`gòä½/+Æ+ÕÛjÛyómuq) Ø÷?´ÉÔ·_gÎãšyÍ 6Sgí8A-m66<ÄÙBHÓú£f½6Óðæ¹l)!„pô}Þ\6Ö„¦<^XåSÿç©æü¹J›ày6™{ËBjÒö&ñƒç9;û|Çq¹‚â0µÕt¨Œ%„ÚØvŽz¾ÙPƒÚì*F?wl‹ü¹ª\Ï Ûï÷ž- Gÿ% ýU$$òŽŽ«yWĤy*^euA½ÑDsžúœt­†5l­Ô¼s“N¦ñû'#22rÆ ê¯í[…y­èiKSۼ匭Í,¬¿9ñ‰§–Ô4²[ÿ]³åÁÒmÇw8Ò²ŒÏ¦½T¢þ€y˜[&#D‡"/Ï«ÑbD?ÊŸ;ŠGšåÏ¥‡,xæ›}§o%^4œ±}”!ªŽŽ«eW¤y*^®áÝú¢j%!¼õ|ªs'ö}”X,n5ˆo»äíij[F{M j›îòêZ_ñÚi!á‹ß‰7o¥Ó¸ëéS~5ËÓyì²CÖ«¾{ËUH´çÏ¥-|"G_º;xöŒá¼N·3©x)}¯w¿}¾~Óœà™s—|xÓyRÃŒ~Ç;€®CŠ[€~ )n{{@°{è7´e¤E¦Z{èkd™Cg|#®HúÇä%q•ÞâÃäï˜+¹½qªŠíá=‚=ôŠÜ#Ñ·M‡ó.î:_¡Bu ØCÏ«9ý¢Wøö|õËd$©ÿô™úE†Œ¨3½ú­üòýçæÎ õók/;mCZØÕß®qö´ÀÉãƒW;miµfª{ø[d…g³Ü?Ž>|,~÷²ºÿ¬Ðš¶!¥l¢Øí£½GOžÚ‘·e}\qëP̵ }ûÍ Áz#Ÿ]÷Š«nÃKö,=ßß3l†Eú®„"†Bjµd¤Õ¶\ãžÁ:Fù½4Ý–GÏfÚËÞõÚ³ÓJ !<ó Clø„èØN™bªNŒÛroÆ®!!£ uOžé7´áeºõé{â*]æO6›68k÷Ñ<%!Z3ÒjY®qÏ€` öƒÒ†ƒuk ³ÓÚµH)[X¥ „pôLtÕíKñøFªìxŸ­NýýDkÄd3á›f}wïáûruFÚg4d¤Õ²ì¡]W‡ÃÈuhfê«$ž’SVf«“ÒYÉýª–Ùi !-SÊv[uy×éjù•÷ÝÇŒ3vÞ9ªœ?bnKe¤%¤!#-!¤y¦Ú–ËÁÚ%ê9¸øÌ¥b%!òœ#¿%U?Š lݹ-[“+æaÊo›“ôÛÍNÛuªòs;þ¢ü¾±Ñ7ÿ§·ÖŸ­B•ôè=õýËç˜æ°ê÷óiéWb?úð§}µ–m¶Ž²ðxÔ·“<7îù|¦ uÖ{MóÊÜwÏZ,ÝräRZÚåد—XÝ8x¹Cžn0ã"Ø÷U‡¦i¾Á0Ÿ%óž‘Þ¾YÚt[Ë”'n\<ÎY$¹ú?·ñT±úO@–wüóeSÝE"‘ÓØ°Õä+ !D^¿ñ…@7‘H4&ðå-çËñ× :>LÚôí ÷X*2ò m'¿òÕ7sª·}z¸ ñ¯Džÿ纅kS½7ïÞ2¯íå¦qÛðË ¼l…B#÷™ooÙfI4u}’äuÞþ«¿]ÿâìi“ǯØ~KÒp»†.±9×Á<û>MU'NHÈá iÚlX!-ü:öJffJÌ? þX»þT%K$éß¾´öŒÝÚƒi7o¦Ý´ÈA"DšùÝKkÏ9}pôzfÚŸk,¾ýÞ1|@!Ò» )2çySÑM·×®á¾ú·Žß¨a !Ê‚CkFÝ ýþ÷¨€Gë@¯5M„eËj§4w}„YA¢Øí£½GOžÚ‘·e}œú&]b öžƒìû&yêj·1.Ns¾—ÌúèÓ¦Oh³±á!ÎBšÖ5ëµ™†7ÏeK%7£=ôû×ÛÁÄŽÀÜÁËÉ„&’›Ñ*|ß}Ýgˆ-´ ^±ÜújLZ-ª€¦¦¤^ÇÌLH5[FY0Õåõ*BˆªèÜåJ»i³‚.±=ÌNõM<çþø1ÈDG×ÄP—¦!’¦Á~õõ]Ÿ}¾ãŒ¸\Aq˜Új:TÆ*ëóëŒ=--v¡¬Ì)˜¼:Ô›OB«¨¡ì«% 1Bïm`¡+Ë­²D·)¨0UÅ5C3]!„ïµÉ3zÍó+ôw¿tTÛÀ½Ú4Z»>Báè™èªG­Ïa¤J–æa§»D•Å@¨Õ‡"û¾‰âš27iÛÇÈnýwÍ–K·ßádHË2>›öR !\cýª»ÅR2ªÙl?×ØÚÌÂú›Ÿx QŸ-Fzè¬Þ®44ÜB}÷ËÖ¦:[kÿÑT!„2·êÿ§Xòê’•üè- í¨³^kš©ž:«ö'–„În>“Ïjìú´éB—(¹6jÕ„2Ä4þSF%­– ¬FØЄ);¿+öœ"tŒœetæ³-§ò$*•¬ìVÊ͇ :. ×;þÉwgò%*ÂHÊÄI§Òð@+!„PÆÞ«ßpHýàõMÇnUHå5yçYµò½%ï…[7ŽÉäw·7={Ãâ51¹rÔY¯5Í”5+G_ý÷ò 1©ùÕIõƒ´ØoV®‰-ª×Ðõiƒ.±•ØkRû§ŒÐåÕµ¾âµÓBÂ#¿o2ÞJG½tå/ŸNÊÚææèè1}õî¬Z–¡óÊ­§ˆ?›íá(=!|Õ¶”‡ˆõjü/üåšùÕ¢‰c\¼¦­=nøòo»×z´˜M£Íý¢v~ãŸùïÅëþ,À½{«iì–þóÉäÂß^ëæ6iÞ»Ñå^ &XÑÔõiï(Ñ%6—^HÅDïrvóÂõ½ÃÞÞ^,£þ†èèèÈÈH´Ú®Kn¤%¿õöŒìú3UÙ{€~ÁÁì¡3¢›é¥CödöFIÚzÿ€Ó¥]Þ¥Àc„—êôR˜'„DEE5-Ù°a!äÉ>X öГšGú¦ÿâG,€`ßO†õ­"}Í?Z•$¯ ~Gá[yúRNiíþƾ\*R„yAü7ïþû…|‰ÀÆû¥ _¼19´xÆþEGw†›Ëo|2oç3[þÚb"½5mð»¸÷G¨÷¨¸÷SDä…׎n 3çB꯼¼ŠÚ|ü“±ºMeÊ¿Zóñþ+9U «ñÏF}¹:À’&ªšë¿½»zËÉ|Êjü¢HC†4¼FZÛò¶%œdFk]ò*möž¬ Qìv`ïzööÏ ­ Þ1˜VçgôÿöèO“‹Olxáí÷FÄnãkñåɬºð‰•ÉÉRK½Û§nKBÜ /¤ÊßÞô:oÞ°é †þ²;±tÚ\KšÔ\Ýs†ü£³nËc E ¿ŽýÁÑŒ¹óîÒµëÝãâ%ñúwÏïL~ÑQzá‹¥¯+—B©Õ²\S ô½¯ye¨Äb1¾ÀBÛA/Ãz}’–¤mó3ʬ&xòoœÉ–Te$V{uÑ Œ¿ò$•i‰%v¢fÁœ¶ Z4R}º˜!lUòî$Ýé-ÓãhÌœ›uà´*hÕš;hòk+¼î´.×TÂ-+FöüLSÒFùùãýë6_ºç/ε öõOÝöEjlK?Wãæ7r´¥ÿbçM¿%Ì Ëüý’éì•#[½XZsæÜ‚z£‰æê´Q”®Õ°†ì¸Êj-Ë5•°¶RóÊ€`ßDFFnØ°Aã×ö]z@OKÊZ†Øî‹“z¿a9ÒHTòKllžpìòÁ-›–cæ½ÄãÓ-G®ê_H2ïŸv¼–;ל9×pˆn}Qµ’!Dñ°¤VeKH;Ë5•P’|EãÊÐ{CHTA/Ðö€^—hÉÏH›ºO1Ëüí@Ç8Kžžƒÿàë¿Æ=tôk“›2·dBÙÖ¾J³]:´õMžÆ̹‡¹~lÂŽ¤2†yöá­çëT„´³\c ùZVŒìûõÞÞÞþï飽óÊ­9|6ÛãÕV`>rÜì5Ÿú·™ìaðC±³¯€ph7ïÁLŠQ £~›Í)CE~ÔÒ?>™:¤ÍLºÐåÕµ¾+ÖN 1bbjï5ÞJ‡%„Ò÷z÷Ûçß]7'xË SsGÏIæê=iY®¹„ÚV€Þ‚·ˆ2ç—Ùóνzì·™æ˜Ñn¤%¿¹t%:ýƒ­IÝþ{ùäe>fht€ýþ€ *‹}ÁÃëÅ3cÖ¾3ňB} ,øÎ~`ÜÓ™‡ýïjê#{@°‡^ГyiûÞáÁßzÚ›+C­xDr5ʧï…á¾Y*ÐßÙ÷=´ù¸9³ !D‚ºŒìû¥†yõ²¤u+öì|Î{ìXßW1„È â7¾è&‰Æ¾¼å|9Cˆ$ywÀ{¿|ñ¼³Hä5÷£„üœ#ëçOt‰Æ/Ú|¹JE!LQ§K|\DNc&GnJ×rÁ”'n\<ÎY$¹ú?·ñT1£a[ ¥ŒìáïÑŸòùw ®®âw4ÊU@!ÒÌoÛ& Dþàø“_§9VxcÞÛ‹¯¯ø*>Ý:÷×¥KÖïšuèõÊëßo81lýÉmƒ¨šÜŒl]ž¶C¶Ír;%»Õ¶Æ.­JÙC ø5%­%„Цo>ënÌØN°§¸¾/‡ÛëÓz¢Ð™6%)yRB(Z‡S——u3§RÁ1°uu1×r‡§)Ëmg·Œì¡û4&•0&„6°0  !„âêðy†–†êFå y*¹‚%Dàüÿ6¯üòëOnÉ&#–®ýð  Mif5e¹Œk³­šÁzEX¶±©4¦¸•$w¼Ž‘Û³oöcEùõßW/{gã„Ó›Æé¶YKc–[ Ûnà5/ô}˜ÆïÃhýA‚šœüÚvÈvæù8Yö‰#Wòj”,×ÀÌÜÏ¥)/ÌÕ˜åVö-KöÐ üás–ûäÿkŠ£È}qt#t^¹uãñg³=E£'„¯Ú–ò°SÏ«jnîûçÜñ£ÜÂ6×Îÿt‡PÓjB—W×úŠ×N XüN¼Éx+ÍÛ¶*š  ÏCŠ[€~ )n{{xjQ„B°èßÁ#{ŒìÁú(Eq !7Ò’QýtdÏáéPWWWWWW__/•Jë$îü9&µ@i;çí—ÇõÉ?+¹{ðËÿ»çûî[¾æt‡+KdæpŸe)`J/ÿos¬î ÿ\l/ º|Pyö›ÿ›\o<Êoáb¿¡:®ž>¯ × Ë²J¥R©TÊd²Üûwï>Úïðœ"#\-„4!¬ìáýŒ+±§®åKÙRB®¡ÈɤàFNM¯¼îY ZøÑ|aì÷Û+U¸6 ÷B;E5üKq(Š"š¢þ?pCša7K±óIEND®B`‚gnm/vignettes/fig-LCqvplot.pdf0000644000176000001440000001625512515147523016113 0ustar ripleyusers%PDF-1.1 %âãÏÓ\r 1 0 obj << /CreationDate (D:20061218174133) /ModDate (D:20061218174133) /Title (R Graphics Output) /Producer (R 2.4.1) /Creator (R) >> endobj 2 0 obj << /Type /Catalog /Pages 3 0 R >> endobj 5 0 obj << /Type /Font /Subtype /Type1 /Name /F1 /BaseFont /ZapfDingbats >> endobj 6 0 obj << /Type /Page /Parent 3 0 R /Contents 7 0 R /Resources 4 0 R >> endobj 7 0 obj << /Length 8 0 R >> stream q Q q 59.04 73.44 378.72 335.52 re W n 0.000 0.000 0.000 RG 0.75 w [] 0 d 1 J 1 j 10.00 M BT /F1 1 Tf 1 Tr 7.48 0 0 7.48 70.10 359.09 Tm (l) Tj 0 Tr /F1 1 Tf 1 Tr 7.48 0 0 7.48 78.07 339.04 Tm (l) Tj 0 Tr /F1 1 Tf 1 Tr 7.48 0 0 7.48 86.04 340.06 Tm (l) Tj 0 Tr /F1 1 Tf 1 Tr 7.48 0 0 7.48 94.01 340.98 Tm (l) Tj 0 Tr /F1 1 Tf 1 Tr 7.48 0 0 7.48 101.98 336.39 Tm (l) Tj 0 Tr /F1 1 Tf 1 Tr 7.48 0 0 7.48 109.95 346.70 Tm (l) Tj 0 Tr /F1 1 Tf 1 Tr 7.48 0 0 7.48 117.92 313.89 Tm (l) Tj 0 Tr /F1 1 Tf 1 Tr 7.48 0 0 7.48 125.89 337.12 Tm (l) Tj 0 Tr /F1 1 Tf 1 Tr 7.48 0 0 7.48 133.86 326.43 Tm (l) Tj 0 Tr /F1 1 Tf 1 Tr 7.48 0 0 7.48 141.83 318.49 Tm (l) Tj 0 Tr /F1 1 Tf 1 Tr 7.48 0 0 7.48 149.80 317.46 Tm (l) Tj 0 Tr /F1 1 Tf 1 Tr 7.48 0 0 7.48 157.77 310.69 Tm (l) Tj 0 Tr /F1 1 Tf 1 Tr 7.48 0 0 7.48 165.74 302.80 Tm (l) Tj 0 Tr /F1 1 Tf 1 Tr 7.48 0 0 7.48 173.71 314.78 Tm (l) Tj 0 Tr /F1 1 Tf 1 Tr 7.48 0 0 7.48 181.68 303.23 Tm (l) Tj 0 Tr /F1 1 Tf 1 Tr 7.48 0 0 7.48 189.65 307.08 Tm (l) Tj 0 Tr /F1 1 Tf 1 Tr 7.48 0 0 7.48 197.62 268.32 Tm (l) Tj 0 Tr /F1 1 Tf 1 Tr 7.48 0 0 7.48 205.59 294.42 Tm (l) Tj 0 Tr /F1 1 Tf 1 Tr 7.48 0 0 7.48 213.56 293.54 Tm (l) Tj 0 Tr /F1 1 Tf 1 Tr 7.48 0 0 7.48 221.53 285.33 Tm (l) Tj 0 Tr /F1 1 Tf 1 Tr 7.48 0 0 7.48 229.50 292.74 Tm (l) Tj 0 Tr /F1 1 Tf 1 Tr 7.48 0 0 7.48 237.47 251.85 Tm (l) Tj 0 Tr /F1 1 Tf 1 Tr 7.48 0 0 7.48 245.44 259.33 Tm (l) Tj 0 Tr /F1 1 Tf 1 Tr 7.48 0 0 7.48 253.41 271.84 Tm (l) Tj 0 Tr /F1 1 Tf 1 Tr 7.48 0 0 7.48 261.38 261.43 Tm (l) Tj 0 Tr /F1 1 Tf 1 Tr 7.48 0 0 7.48 269.35 256.54 Tm (l) Tj 0 Tr /F1 1 Tf 1 Tr 7.48 0 0 7.48 277.32 221.27 Tm (l) Tj 0 Tr /F1 1 Tf 1 Tr 7.48 0 0 7.48 285.29 259.30 Tm (l) Tj 0 Tr /F1 1 Tf 1 Tr 7.48 0 0 7.48 293.26 252.63 Tm (l) Tj 0 Tr /F1 1 Tf 1 Tr 7.48 0 0 7.48 301.23 242.81 Tm (l) Tj 0 Tr /F1 1 Tf 1 Tr 7.48 0 0 7.48 309.20 241.76 Tm (l) Tj 0 Tr /F1 1 Tf 1 Tr 7.48 0 0 7.48 317.17 237.95 Tm (l) Tj 0 Tr /F1 1 Tf 1 Tr 7.48 0 0 7.48 325.14 223.36 Tm (l) Tj 0 Tr /F1 1 Tf 1 Tr 7.48 0 0 7.48 333.10 236.22 Tm (l) Tj 0 Tr /F1 1 Tf 1 Tr 7.48 0 0 7.48 341.07 220.15 Tm (l) Tj 0 Tr /F1 1 Tf 1 Tr 7.48 0 0 7.48 349.04 200.67 Tm (l) Tj 0 Tr /F1 1 Tf 1 Tr 7.48 0 0 7.48 357.01 156.14 Tm (l) Tj 0 Tr /F1 1 Tf 1 Tr 7.48 0 0 7.48 364.98 191.07 Tm (l) Tj 0 Tr /F1 1 Tf 1 Tr 7.48 0 0 7.48 372.95 188.09 Tm (l) Tj 0 Tr /F1 1 Tf 1 Tr 7.48 0 0 7.48 380.92 192.73 Tm (l) Tj 0 Tr /F1 1 Tf 1 Tr 7.48 0 0 7.48 388.89 165.53 Tm (l) Tj 0 Tr /F1 1 Tf 1 Tr 7.48 0 0 7.48 396.86 171.70 Tm (l) Tj 0 Tr /F1 1 Tf 1 Tr 7.48 0 0 7.48 404.83 162.30 Tm (l) Tj 0 Tr /F1 1 Tf 1 Tr 7.48 0 0 7.48 412.80 134.21 Tm (l) Tj 0 Tr /F1 1 Tf 1 Tr 7.48 0 0 7.48 420.77 131.07 Tm (l) Tj 0 Tr ET Q q 0.000 0.000 0.000 RG 0.75 w [] 0 d 1 J 1 j 10.00 M 112.92 73.44 m 431.70 73.44 l S 112.92 73.44 m 112.92 66.24 l S 192.61 73.44 m 192.61 66.24 l S 272.31 73.44 m 272.31 66.24 l S 352.01 73.44 m 352.01 66.24 l S 431.70 73.44 m 431.70 66.24 l S BT 0.000 0.000 0.000 rg /F2 1 Tf 12.00 0.00 -0.00 12.00 106.24 47.52 Tm (50) Tj /F2 1 Tf 12.00 0.00 -0.00 12.00 185.94 47.52 Tm (60) Tj /F2 1 Tf 12.00 0.00 -0.00 12.00 265.64 47.52 Tm (70) Tj /F2 1 Tf 12.00 0.00 -0.00 12.00 345.33 47.52 Tm (80) Tj /F2 1 Tf 12.00 0.00 -0.00 12.00 425.03 47.52 Tm (90) Tj ET 59.04 98.78 m 59.04 361.69 l S 59.04 98.78 m 51.84 98.78 l S 59.04 186.42 m 51.84 186.42 l S 59.04 274.05 m 51.84 274.05 l S 59.04 361.69 m 51.84 361.69 l S BT /F2 1 Tf 0.00 12.00 -12.00 0.00 41.76 86.94 Tm (-1.5) Tj /F2 1 Tf 0.00 12.00 -12.00 0.00 41.76 174.57 Tm (-1.0) Tj /F2 1 Tf 0.00 12.00 -12.00 0.00 41.76 262.21 Tm (-0.5) Tj /F2 1 Tf 0.00 12.00 -12.00 0.00 41.76 353.35 Tm (0.0) Tj ET 59.04 73.44 m 437.76 73.44 l 437.76 408.96 l 59.04 408.96 l 59.04 73.44 l S Q q BT 0.000 0.000 0.000 rg /F3 1 Tf 14.00 0.00 -0.00 14.00 44.73 442.09 Tm (Canada, males over 45, Lee-Carter model: relative sensitivity) Tj /F3 1 Tf 14.00 0.00 -0.00 14.00 102.47 424.81 Tm (of different ages to change in total mortality) Tj /F2 1 Tf 12.00 0.00 -0.00 12.00 237.73 18.72 Tm (Age) Tj /F2 1 Tf 0.00 12.00 -12.00 0.00 12.96 217.52 Tm (log\(beta\)) Tj ET Q q 59.04 73.44 378.72 335.52 re W n 0.000 0.000 0.000 RG 0.75 w [] 0 d 1 J 1 j 10.00 M 73.07 352.73 m 73.07 370.64 l S 81.04 331.98 m 81.04 351.29 l S 89.01 333.36 m 89.01 351.94 l S 96.98 334.69 m 96.98 352.45 l S 104.95 330.15 m 104.95 347.82 l S 112.92 341.22 m 112.92 357.37 l S 120.88 307.06 m 120.88 325.91 l S 128.85 331.73 m 128.85 347.69 l S 136.82 320.80 m 136.82 337.27 l S 144.79 312.80 m 144.79 329.37 l S 152.76 311.90 m 152.76 328.21 l S 160.73 305.08 m 160.73 321.49 l S 168.70 297.04 m 168.70 313.75 l S 176.67 309.76 m 176.67 324.99 l S 184.64 297.95 m 184.64 313.70 l S 192.61 302.22 m 192.61 317.14 l S 200.58 261.85 m 200.58 279.97 l S 208.55 289.47 m 208.55 304.56 l S 216.52 288.73 m 216.52 303.53 l S 224.49 280.41 m 224.49 295.45 l S 232.46 288.37 m 232.46 302.29 l S 240.43 245.83 m 240.43 263.07 l S 248.40 253.92 m 248.40 269.92 l S 256.37 267.16 m 256.37 281.72 l S 264.34 256.50 m 264.34 271.56 l S 272.31 251.58 m 272.31 266.68 l S 280.28 214.69 m 280.28 233.05 l S 288.25 254.70 m 288.25 269.09 l S 296.22 247.83 m 296.22 262.62 l S 304.19 237.67 m 304.19 253.14 l S 312.16 236.61 m 312.16 252.10 l S 320.13 232.67 m 320.13 248.43 l S 328.10 217.35 m 328.10 234.56 l S 336.07 230.82 m 336.07 246.81 l S 344.04 213.92 m 344.04 231.58 l S 352.01 193.33 m 352.01 213.20 l S 359.98 145.60 m 359.98 171.87 l S 367.95 182.70 m 367.95 204.63 l S 375.92 179.22 m 375.92 202.15 l S 383.88 183.77 m 383.88 206.88 l S 391.85 154.03 m 391.85 182.23 l S 399.82 160.09 m 399.82 188.51 l S 407.79 149.06 m 407.79 180.73 l S 415.76 116.95 m 415.76 156.66 l S 423.73 111.76 m 423.73 155.58 l S BT 0.000 0.000 0.000 rg /F2 1 Tf 12.00 0.00 -0.00 12.00 340.96 130.85 Tm (Age 81) Tj /F2 1 Tf 12.00 0.00 -0.00 12.00 261.26 200.96 Tm (Age 71) Tj /F2 1 Tf 12.00 0.00 -0.00 12.00 181.57 248.28 Tm (Age 61) Tj /F2 1 Tf 12.00 0.00 -0.00 12.00 101.87 292.10 Tm (Age 51) Tj ET Q endstream endobj 8 0 obj 5932 endobj 3 0 obj << /Type /Pages /Kids [ 6 0 R ] /Count 1 /MediaBox [0 0 468 468] >> endobj 4 0 obj << /ProcSet [/PDF /Text] /Font << /F1 5 0 R /F2 10 0 R /F3 11 0 R >> /ExtGState << >> >> endobj 9 0 obj << /Type /Encoding /BaseEncoding /WinAnsiEncoding /Differences [ 45/minus 96/quoteleft 144/dotlessi /grave /acute /circumflex /tilde /macron /breve /dotaccent /dieresis /.notdef /ring /cedilla /.notdef /hungarumlaut /ogonek /caron /space] >> endobj 10 0 obj << /Type /Font /Subtype /Type1 /Name /F2 /BaseFont /Helvetica /Encoding 9 0 R >> endobj 11 0 obj << /Type /Font /Subtype /Type1 /Name /F3 /BaseFont /Helvetica-Bold /Encoding 9 0 R >> endobj xref 0 12 0000000000 65535 f 0000000021 00000 n 0000000163 00000 n 0000006380 00000 n 0000006463 00000 n 0000000212 00000 n 0000000295 00000 n 0000000375 00000 n 0000006360 00000 n 0000006567 00000 n 0000006824 00000 n 0000006921 00000 n trailer << /Size 12 /Info 1 0 R /Root 2 0 R >> startxref 7023 %%EOF gnm/vignettes/fig-deaths1921-1940.pdf0000644000176000001440000001701312515147523016520 0ustar ripleyusers%PDF-1.1 %âãÏÓ\r 1 0 obj << /CreationDate (D:20061218211720) /ModDate (D:20061218211720) /Title (R Graphics Output) /Producer (R 2.4.1) /Creator (R) >> endobj 2 0 obj << /Type /Catalog /Pages 3 0 R >> endobj 5 0 obj << /Type /Font /Subtype /Type1 /Name /F1 /BaseFont /ZapfDingbats >> endobj 6 0 obj << /Type /Page /Parent 3 0 R /Contents 7 0 R /Resources 4 0 R >> endobj 7 0 obj << /Length 8 0 R >> stream q Q q 59.04 73.44 378.72 335.52 re W n 0.000 0.000 0.000 RG 0.75 w [] 0 d 1 J 1 j 10.00 M BT /F1 1 Tf 1 Tr 7.48 0 0 7.48 70.10 164.14 Tm (l) Tj 0 Tr /F1 1 Tf 1 Tr 7.48 0 0 7.48 74.54 166.20 Tm (l) Tj 0 Tr /F1 1 Tf 1 Tr 7.48 0 0 7.48 78.98 164.41 Tm (l) Tj 0 Tr /F1 1 Tf 1 Tr 7.48 0 0 7.48 83.42 165.75 Tm (l) Tj 0 Tr /F1 1 Tf 1 Tr 7.48 0 0 7.48 87.86 161.13 Tm (l) Tj 0 Tr /F1 1 Tf 1 Tr 7.48 0 0 7.48 92.30 160.36 Tm (l) Tj 0 Tr /F1 1 Tf 1 Tr 7.48 0 0 7.48 96.74 160.11 Tm (l) Tj 0 Tr /F1 1 Tf 1 Tr 7.48 0 0 7.48 101.18 160.08 Tm (l) Tj 0 Tr /F1 1 Tf 1 Tr 7.48 0 0 7.48 105.62 163.97 Tm (l) Tj 0 Tr /F1 1 Tf 1 Tr 7.48 0 0 7.48 110.05 157.73 Tm (l) Tj 0 Tr /F1 1 Tf 1 Tr 7.48 0 0 7.48 114.49 161.26 Tm (l) Tj 0 Tr 1.000 0.000 0.000 rg 1.000 0.000 0.000 RG /F1 1 Tf 2 Tr 7.48 0 0 7.48 118.93 150.88 Tm (l) Tj 0 Tr 0.000 0.000 0.000 RG /F1 1 Tf 1 Tr 7.48 0 0 7.48 123.37 162.44 Tm (l) Tj 0 Tr /F1 1 Tf 1 Tr 7.48 0 0 7.48 127.81 159.00 Tm (l) Tj 0 Tr /F1 1 Tf 1 Tr 7.48 0 0 7.48 132.25 158.34 Tm (l) Tj 0 Tr /F1 1 Tf 1 Tr 7.48 0 0 7.48 136.69 173.81 Tm (l) Tj 0 Tr /F1 1 Tf 1 Tr 7.48 0 0 7.48 141.13 168.11 Tm (l) Tj 0 Tr /F1 1 Tf 1 Tr 7.48 0 0 7.48 145.56 168.35 Tm (l) Tj 0 Tr /F1 1 Tf 1 Tr 7.48 0 0 7.48 150.00 181.93 Tm (l) Tj 0 Tr /F1 1 Tf 1 Tr 7.48 0 0 7.48 154.44 175.15 Tm (l) Tj 0 Tr /F1 1 Tf 1 Tr 7.48 0 0 7.48 158.88 194.09 Tm (l) Tj 0 Tr 1.000 0.000 0.000 RG /F1 1 Tf 2 Tr 7.48 0 0 7.48 163.32 171.99 Tm (l) Tj 0 Tr 0.000 0.000 0.000 RG /F1 1 Tf 1 Tr 7.48 0 0 7.48 167.76 201.84 Tm (l) Tj 0 Tr /F1 1 Tf 1 Tr 7.48 0 0 7.48 172.20 191.00 Tm (l) Tj 0 Tr /F1 1 Tf 1 Tr 7.48 0 0 7.48 176.64 190.95 Tm (l) Tj 0 Tr /F1 1 Tf 1 Tr 7.48 0 0 7.48 181.08 214.42 Tm (l) Tj 0 Tr /F1 1 Tf 1 Tr 7.48 0 0 7.48 185.51 201.68 Tm (l) Tj 0 Tr /F1 1 Tf 1 Tr 7.48 0 0 7.48 189.95 209.12 Tm (l) Tj 0 Tr /F1 1 Tf 1 Tr 7.48 0 0 7.48 194.39 223.31 Tm (l) Tj 0 Tr /F1 1 Tf 1 Tr 7.48 0 0 7.48 198.83 224.41 Tm (l) Tj 0 Tr /F1 1 Tf 1 Tr 7.48 0 0 7.48 203.27 248.61 Tm (l) Tj 0 Tr 1.000 0.000 0.000 RG /F1 1 Tf 2 Tr 7.48 0 0 7.48 207.71 221.00 Tm (l) Tj 0 Tr 0.000 0.000 0.000 RG /F1 1 Tf 1 Tr 7.48 0 0 7.48 212.15 255.09 Tm (l) Tj 0 Tr /F1 1 Tf 1 Tr 7.48 0 0 7.48 216.59 248.91 Tm (l) Tj 0 Tr /F1 1 Tf 1 Tr 7.48 0 0 7.48 221.02 254.08 Tm (l) Tj 0 Tr /F1 1 Tf 1 Tr 7.48 0 0 7.48 225.46 265.68 Tm (l) Tj 0 Tr /F1 1 Tf 1 Tr 7.48 0 0 7.48 229.90 268.95 Tm (l) Tj 0 Tr /F1 1 Tf 1 Tr 7.48 0 0 7.48 234.34 265.89 Tm (l) Tj 0 Tr /F1 1 Tf 1 Tr 7.48 0 0 7.48 238.78 287.94 Tm (l) Tj 0 Tr /F1 1 Tf 1 Tr 7.48 0 0 7.48 243.22 283.84 Tm (l) Tj 0 Tr /F1 1 Tf 1 Tr 7.48 0 0 7.48 247.66 311.59 Tm (l) Tj 0 Tr 1.000 0.000 0.000 RG /F1 1 Tf 2 Tr 7.48 0 0 7.48 252.10 277.42 Tm (l) Tj 0 Tr 0.000 0.000 0.000 RG /F1 1 Tf 1 Tr 7.48 0 0 7.48 256.54 316.29 Tm (l) Tj 0 Tr /F1 1 Tf 1 Tr 7.48 0 0 7.48 260.97 326.30 Tm (l) Tj 0 Tr /F1 1 Tf 1 Tr 7.48 0 0 7.48 265.41 329.90 Tm (l) Tj 0 Tr /F1 1 Tf 1 Tr 7.48 0 0 7.48 269.85 356.85 Tm (l) Tj 0 Tr /F1 1 Tf 1 Tr 7.48 0 0 7.48 274.29 326.34 Tm (l) Tj 0 Tr /F1 1 Tf 1 Tr 7.48 0 0 7.48 278.73 342.85 Tm (l) Tj 0 Tr /F1 1 Tf 1 Tr 7.48 0 0 7.48 283.17 372.90 Tm (l) Tj 0 Tr /F1 1 Tf 1 Tr 7.48 0 0 7.48 287.61 363.43 Tm (l) Tj 0 Tr /F1 1 Tf 1 Tr 7.48 0 0 7.48 292.05 386.06 Tm (l) Tj 0 Tr 1.000 0.000 0.000 RG /F1 1 Tf 2 Tr 7.48 0 0 7.48 296.48 343.44 Tm (l) Tj 0 Tr 0.000 0.000 0.000 RG /F1 1 Tf 1 Tr 7.48 0 0 7.48 300.92 393.94 Tm (l) Tj 0 Tr /F1 1 Tf 1 Tr 7.48 0 0 7.48 305.36 387.99 Tm (l) Tj 0 Tr /F1 1 Tf 1 Tr 7.48 0 0 7.48 309.80 386.23 Tm (l) Tj 0 Tr /F1 1 Tf 1 Tr 7.48 0 0 7.48 314.24 385.71 Tm (l) Tj 0 Tr /F1 1 Tf 1 Tr 7.48 0 0 7.48 318.68 382.43 Tm (l) Tj 0 Tr /F1 1 Tf 1 Tr 7.48 0 0 7.48 323.12 355.25 Tm (l) Tj 0 Tr /F1 1 Tf 1 Tr 7.48 0 0 7.48 327.56 366.53 Tm (l) Tj 0 Tr /F1 1 Tf 1 Tr 7.48 0 0 7.48 332.00 339.33 Tm (l) Tj 0 Tr /F1 1 Tf 1 Tr 7.48 0 0 7.48 336.43 335.97 Tm (l) Tj 0 Tr 1.000 0.000 0.000 RG /F1 1 Tf 2 Tr 7.48 0 0 7.48 340.87 288.44 Tm (l) Tj 0 Tr 0.000 0.000 0.000 RG /F1 1 Tf 1 Tr 7.48 0 0 7.48 345.31 299.00 Tm (l) Tj 0 Tr /F1 1 Tf 1 Tr 7.48 0 0 7.48 349.75 279.67 Tm (l) Tj 0 Tr /F1 1 Tf 1 Tr 7.48 0 0 7.48 354.19 265.61 Tm (l) Tj 0 Tr /F1 1 Tf 1 Tr 7.48 0 0 7.48 358.63 240.83 Tm (l) Tj 0 Tr /F1 1 Tf 1 Tr 7.48 0 0 7.48 363.07 220.97 Tm (l) Tj 0 Tr /F1 1 Tf 1 Tr 7.48 0 0 7.48 367.51 198.96 Tm (l) Tj 0 Tr /F1 1 Tf 1 Tr 7.48 0 0 7.48 371.94 175.64 Tm (l) Tj 0 Tr /F1 1 Tf 1 Tr 7.48 0 0 7.48 376.38 157.71 Tm (l) Tj 0 Tr /F1 1 Tf 1 Tr 7.48 0 0 7.48 380.82 144.98 Tm (l) Tj 0 Tr /F1 1 Tf 1 Tr 7.48 0 0 7.48 385.26 123.93 Tm (l) Tj 0 Tr /F1 1 Tf 1 Tr 7.48 0 0 7.48 389.70 118.38 Tm (l) Tj 0 Tr /F1 1 Tf 1 Tr 7.48 0 0 7.48 394.14 108.48 Tm (l) Tj 0 Tr /F1 1 Tf 1 Tr 7.48 0 0 7.48 398.58 101.92 Tm (l) Tj 0 Tr /F1 1 Tf 1 Tr 7.48 0 0 7.48 403.02 96.13 Tm (l) Tj 0 Tr /F1 1 Tf 1 Tr 7.48 0 0 7.48 407.46 91.49 Tm (l) Tj 0 Tr /F1 1 Tf 1 Tr 7.48 0 0 7.48 411.89 89.19 Tm (l) Tj 0 Tr /F1 1 Tf 1 Tr 7.48 0 0 7.48 416.33 86.12 Tm (l) Tj 0 Tr /F1 1 Tf 1 Tr 7.48 0 0 7.48 420.77 83.27 Tm (l) Tj 0 Tr ET Q q 0.000 0.000 0.000 RG 0.75 w [] 0 d 1 J 1 j 10.00 M 73.07 73.44 m 428.17 73.44 l S 73.07 73.44 m 73.07 66.24 l S 161.84 73.44 m 161.84 66.24 l S 250.62 73.44 m 250.62 66.24 l S 339.40 73.44 m 339.40 66.24 l S 428.17 73.44 m 428.17 66.24 l S BT 0.000 0.000 0.000 rg /F2 1 Tf 12.00 0.00 -0.00 12.00 66.39 47.52 Tm (20) Tj /F2 1 Tf 12.00 0.00 -0.00 12.00 155.17 47.52 Tm (40) Tj /F2 1 Tf 12.00 0.00 -0.00 12.00 243.95 47.52 Tm (60) Tj /F2 1 Tf 12.00 0.00 -0.00 12.00 332.72 47.52 Tm (80) Tj /F2 1 Tf 12.00 0.00 -0.00 12.00 418.16 47.52 Tm (100) Tj ET 59.04 82.29 m 59.04 388.86 l S 59.04 82.29 m 51.84 82.29 l S 59.04 158.93 m 51.84 158.93 l S 59.04 235.58 m 51.84 235.58 l S 59.04 312.22 m 51.84 312.22 l S 59.04 388.86 m 51.84 388.86 l S BT /F2 1 Tf 0.00 12.00 -12.00 0.00 41.76 78.95 Tm (0) Tj /F2 1 Tf 0.00 12.00 -12.00 0.00 41.76 145.59 Tm (5000) Tj /F2 1 Tf 0.00 12.00 -12.00 0.00 41.76 218.90 Tm (10000) Tj /F2 1 Tf 0.00 12.00 -12.00 0.00 41.76 295.54 Tm (15000) Tj /F2 1 Tf 0.00 12.00 -12.00 0.00 41.76 372.18 Tm (20000) Tj ET 59.04 73.44 m 437.76 73.44 l 437.76 408.96 l 59.04 408.96 l 59.04 73.44 l S Q q BT 0.000 0.000 0.000 rg /F3 1 Tf 14.00 0.00 -0.00 14.00 93.98 433.45 Tm (Canada, males: Total deaths 1921-1940 by age) Tj /F2 1 Tf 12.00 0.00 -0.00 12.00 237.73 18.72 Tm (Age) Tj /F2 1 Tf 0.00 12.00 -12.00 0.00 12.96 208.90 Tm (Total deaths) Tj ET Q endstream endobj 8 0 obj 6282 endobj 3 0 obj << /Type /Pages /Kids [ 6 0 R ] /Count 1 /MediaBox [0 0 468 468] >> endobj 4 0 obj << /ProcSet [/PDF /Text] /Font << /F1 5 0 R /F2 10 0 R /F3 11 0 R >> /ExtGState << >> >> endobj 9 0 obj << /Type /Encoding /BaseEncoding /WinAnsiEncoding /Differences [ 45/minus 96/quoteleft 144/dotlessi /grave /acute /circumflex /tilde /macron /breve /dotaccent /dieresis /.notdef /ring /cedilla /.notdef /hungarumlaut /ogonek /caron /space] >> endobj 10 0 obj << /Type /Font /Subtype /Type1 /Name /F2 /BaseFont /Helvetica /Encoding 9 0 R >> endobj 11 0 obj << /Type /Font /Subtype /Type1 /Name /F3 /BaseFont /Helvetica-Bold /Encoding 9 0 R >> endobj xref 0 12 0000000000 65535 f 0000000021 00000 n 0000000163 00000 n 0000006730 00000 n 0000006813 00000 n 0000000212 00000 n 0000000295 00000 n 0000000375 00000 n 0000006710 00000 n 0000006917 00000 n 0000007174 00000 n 0000007271 00000 n trailer << /Size 12 /Info 1 0 R /Root 2 0 R >> startxref 7373 %%EOF gnm/vignettes/gnm.bib0000644000176000001440000001612112515147523014340 0ustar ripleyusers%% This BibTeX bibliography file was created using BibDesk. %% http://www.cs.ucsd.edu/~mmccrack/bibdesk.html %% Created for David at 2005-07-14 15:46:53 +0100 @book{Agre02, author={A Agresti}, title={Categorical Data Analysis}, publisher={New York: Wiley}, year={2002}, edition={2nd} } @article{Alho00, author={Alho, J. M.}, title={Discussion of {Lee} (2000)}, journal={North American Actuarial Journal}, volume={4}, year={2000}, pages={91--93} } @article{Ande84, author={J. A. Anderson}, title={Regression and Ordered Categorical Variables}, journal={J. R. Statist. Soc. B}, volume={46}, year={1984}, pages={1-30}, number={1} } @article{BrouDenuVerm02, author={Brouhns, N. and Denuit, M. and Vermunt, J. K.}, title={A Poisson log-bilinear regression approach to the construction of projected lifetables}, journal={Insurance Mathematics and Economics}, volume= 31, year= 2002, pages={373-393} } @article{CatcMorg97, author={Catchpole, E.A. and Morgan, B.J.T.}, title={Detecting parameter redundancy}, journal={Biometrika}, volume={84}, year={1997}, pages={187--196} } @article{Caut98, author={B Cautres and A F Heath and D Firth}, title={Class, religion and vote in {B}ritain and {F}rance}, journal={La Lettre de la Maison Fran{\c{c}}aise}, volume={8}, year={1998} } @article{Clif93, author={P Clifford and A F Heath}, title={The Political Consequences of Social Mobility}, journal={J. Roy. Stat. Soc. A}, volume={156}, year={1993}, pages={51-61}, number={1} } @article{deLe06, author={J de Leeuw}, title={Principal component analysis of binary data by iterated singular value decomposition}, journal={Comp. Stat. Data Anal.}, volume={50}, year={2006}, pages={21-39}, number={1} } @article{Erik82, author={Erikson, R and Goldthorpe, J H and Portocarero, L}, title={Social fluidity in Industrial nations: {E}ngland, {F}rance and {S}weden}, journal={British Journal of Sociology}, volume={33}, year={1982}, pages={1--34} } @book{Erik92, author={R Erikson and J H Goldthorpe}, title={The Constant Flux}, publisher={Oxford: Clarendon Press}, year={1992} } @article{Firt03, author={D. Firth}, title={Overcoming the reference category problem in the presentation of statistical models}, journal={Sociological Methodology}, volume={33}, year={2003}, pages={1--18} } @inproceedings{Firt98, author={Firth, David}, title={LLAMA: An object-oriented system for log multiplicative models}, booktitle={COMPSTAT 1998, Proceedings in Computational Statistics}, year={1998}, publisher={Heidelberg: Physica-Verlag}, editor={Payne, Roger and Green, Peter}, pages={305--310} } @article{FirtMene04, author={D. Firth and R. X. {de Menezes}}, title={Quasi-variances}, journal={Biometrika}, volume={91}, year={2004}, pages={65--80} } @Book{Franc93, ALTauthor = {}, editor = {Francis, B J and Green, M and Payne, C D}, title = {The GLIM System, Release 4 Manual}, publisher = {Oxford: Clarendon Press}, year = {1993}, OPTkey = {}, OPTvolume = {}, OPTnumber = {}, OPTseries = {}, OPTaddress = {}, OPTedition = {}, OPTmonth = {}, OPTnote = {}, OPTannote = {} } @article{Gabr98, author={K. R. Gabriel}, title={Generalised bilinear regression}, journal={Biometrika}, volume={85}, year={1998}, pages={689-700} } @article{Good79, author={L A Goodman}, title={Simple models for the analysis of association in cross-classifications having ordered categories}, journal={J. Amer. Statist. Assoc.}, volume={74}, year={1979}, pages={537-552} } @article{Good85, author={L A Goodman}, title={The analysis of cross-classified data having ordered and/or unordered categories: Association models, correlation models, and asymmetry models for contingency tables with or without missing entries.}, journal={Ann Statist}, volume={13}, year={1985}, pages={10-69} } @techreport{Hatz04, author={Hatzinger, R and Francis, B J}, title={Fitting Paired Comparison Models in {R}}, institution={Department of Statistics and Mathematics, Wirtschaftsuniversit{\"a}t Wien}, year={2004}, number={3} } @article{LeeCart92, author={Lee, R. D. and Carter, L.}, title={Modelling and forecasting the time series of {US} mortality}, journal={Journal of the American Statistical Association}, volume={87}, year={1992}, pages={659-671} } @book{McCu89, author={McCullagh, P. and Nelder, J. A.}, title={Generalized Linear Models ({S}econd Edition)}, publisher={Chapman \& Hall Ltd}, year={1989}, pages={500} } @article{RensHabe03, author={Renshaw, A. and Haberman, S.}, title={Lee-Carter mortality forecasting: a parallel generalized linear modelling approach for {England} and {Wales} mortality projections}, journal={Applied Statistics}, volume={52}, year={2003}, pages={119--137} } @book{Sebe89, author={Seber, G. A. F. and Wild, C. J.}, title={Nonlinear Regression}, publisher={Wiley}, year={1989} } @article{Sobe81, author={M. E. Sobel}, title={Diagonal mobility models: A substantively motivated class of designs for the analysis of mobility effects}, journal={Amer. Soc. Rev.}, volume={46}, year={1981}, pages={893-906} } @article{Sobe85, author={M. E. Sobel}, title={Social mobility and fertility revisited: Some new models for the analysis of the mobility effects hypothesis}, journal={Amer. Soc. Rev.}, volume={50}, year={1985}, pages={699-712} } @article{Vand02, author={F. W. P. {van der Slik} and N. D. {de Graaf} and J. R. M. Gerris}, title={Conformity to Parental Rules: Asymmetric Influences of Father's and Mother's Levels of Education}, journal={Europ. Soc. Rev.}, volume={18}, year={2002}, pages={489-502}, month={4} } @article{vanE, author={F A {van Eeuwijk}}, title={Multiplicative interaction in generalized linear models}, journal={Biometrics}, volume={51}, year={1995}, pages={1017-1032} } @article{Varg01, author={M Vargas and J Crossa and F {van Eeuwijk} and K D Sayre and M P Reynolds}, title={Interpreting treatment by environment interaction in agronomy trials}, journal={Agronomy Journal}, volume={93}, year={2001}, pages={949-960} } @article{Wedd74, author={Wedderburn, R. W. M.}, title={Quasi-likelihood Functions, Generalized Linear Models, and the {G}auss-{N}ewton Method}, journal={Biometrika}, volume={61}, year={1974}, pages={439--447}, keywords={Estimation; Exponential family; Maximum likelihood} } @techreport{Wilm93, author={Wilmoth, J. R.}, title={Computational methods for fitting and extrapolating the {Lee-Carter} model of mortality change}, institution={Department of Demography, University of California, Berkeley}, year={1993} } @article{Xie92, author={Y Xie}, title={The log-multiplicative layer effect model for comparing mobility tables}, journal={American Sociological Review}, volume={57}, year={1992}, pages={380-395} } @book{Yais04, author={Yaish, Meir}, title={Class Mobility Trends in Israeli Society, 1974-1991}, publisher={Edwin Mellen Press}, year={2004}, address={Lewiston} } @phdthesis{Yais98, author={M Yaish}, title={Opportunities, Little Change. Class Mobility in {I}sraeli Society, 1974--1991}, school={Nuffield College, University of Oxford}, year={1998} } gnm/vignettes/fig-Effect_plots.pdf0000644000176000001440000001241512515147523016756 0ustar ripleyusers%PDF-1.4 %âãÏÓ\r 1 0 obj << /CreationDate (D:20100416145338) /ModDate (D:20100416145338) /Title (R Graphics Output) /Producer (R 2.10.1) /Creator (R) >> endobj 2 0 obj << /Type /Catalog /Pages 3 0 R >> endobj 5 0 obj << /Type /Page /Parent 3 0 R /Contents 6 0 R /Resources 4 0 R >> endobj 6 0 obj << /Length 7 0 R >> stream q Q q 28.34 31.10 180.06 156.56 re W n BT 0.000 0.000 0.000 rg /F2 1 Tf 7.00 0.00 -0.00 7.00 44.97 51.31 Tm (A) Tj ET BT /F2 1 Tf 7.00 0.00 -0.00 7.00 67.68 72.28 Tm (B) Tj ET BT /F2 1 Tf 7.00 0.00 -0.00 7.00 93.33 85.85 Tm (C) Tj ET BT /F2 1 Tf 7.00 0.00 -0.00 7.00 94.07 69.87 Tm (D) Tj ET BT /F2 1 Tf 7.00 0.00 -0.00 7.00 99.71 83.19 Tm (E) Tj ET BT /F2 1 Tf 7.00 0.00 -0.00 7.00 106.20 78.95 Tm (F) Tj ET BT /F2 1 Tf 7.00 0.00 -0.00 7.00 113.86 73.01 Tm (G) Tj ET BT /F2 1 Tf 7.00 0.00 -0.00 7.00 133.51 60.60 Tm (H) Tj ET BT /F2 1 Tf 7.00 0.00 -0.00 7.00 159.70 56.30 Tm (I) Tj ET Q q 0.000 0.000 0.000 RG 0.75 w [] 0 d 1 J 1 j 10.00 M 51.68 31.10 m 185.06 31.10 l S 51.68 31.10 m 51.68 27.65 l S 85.02 31.10 m 85.02 27.65 l S 118.37 31.10 m 118.37 27.65 l S 151.71 31.10 m 151.71 27.65 l S 185.06 31.10 m 185.06 27.65 l S BT 0.000 0.000 0.000 rg /F2 1 Tf 7.00 0.00 -0.00 7.00 47.69 18.66 Tm (-4) Tj ET BT /F2 1 Tf 7.00 0.00 -0.00 7.00 81.03 18.66 Tm (-2) Tj ET BT /F2 1 Tf 7.00 0.00 -0.00 7.00 116.42 18.66 Tm (0) Tj ET BT /F2 1 Tf 7.00 0.00 -0.00 7.00 149.77 18.66 Tm (2) Tj ET BT /F2 1 Tf 7.00 0.00 -0.00 7.00 183.11 18.66 Tm (4) Tj ET 28.34 51.40 m 28.34 167.37 l S 28.34 51.40 m 24.88 51.40 l S 28.34 80.39 m 24.88 80.39 l S 28.34 109.38 m 24.88 109.38 l S 28.34 138.37 m 24.88 138.37 l S 28.34 167.37 m 24.88 167.37 l S BT /F2 1 Tf 0.00 7.00 -7.00 0.00 20.04 47.41 Tm (-4) Tj ET BT /F2 1 Tf 0.00 7.00 -7.00 0.00 20.04 76.40 Tm (-2) Tj ET BT /F2 1 Tf 0.00 7.00 -7.00 0.00 20.04 107.44 Tm (0) Tj ET BT /F2 1 Tf 0.00 7.00 -7.00 0.00 20.04 136.43 Tm (2) Tj ET BT /F2 1 Tf 0.00 7.00 -7.00 0.00 20.04 165.42 Tm (4) Tj ET 28.34 31.10 m 208.40 31.10 l 208.40 187.66 l 28.34 187.66 l 28.34 31.10 l S Q q 0.00 0.00 216.00 216.00 re W n BT 0.000 0.000 0.000 rg /F3 1 Tf 9.00 0.00 -0.00 9.00 93.91 198.60 Tm [(Site Eff) 10 (ects)] TJ ET BT /F2 1 Tf 7.00 0.00 -0.00 7.00 97.36 4.84 Tm (Component 1) Tj ET BT /F2 1 Tf 0.00 7.00 -7.00 0.00 6.22 88.37 Tm (Component 2) Tj ET Q q 244.34 31.10 180.06 156.56 re W n Q q 244.34 31.10 180.06 156.56 re W n BT 0.000 0.000 0.000 rg /F2 1 Tf 7.00 0.00 -0.00 7.00 347.42 123.78 Tm (1) Tj ET BT /F2 1 Tf 7.00 0.00 -0.00 7.00 360.67 127.41 Tm (2) Tj ET BT /F2 1 Tf 7.00 0.00 -0.00 7.00 360.77 126.05 Tm (3) Tj ET BT /F2 1 Tf 7.00 0.00 -0.00 7.00 347.84 119.97 Tm (4) Tj ET BT /F2 1 Tf 7.00 0.00 -0.00 7.00 347.93 116.64 Tm (5) Tj ET BT /F2 1 Tf 7.00 0.00 -0.00 7.00 358.95 113.00 Tm (6) Tj ET BT /F2 1 Tf 7.00 0.00 -0.00 7.00 347.20 111.99 Tm (7) Tj ET BT /F2 1 Tf 7.00 0.00 -0.00 7.00 348.58 106.53 Tm (8) Tj ET BT /F2 1 Tf 7.00 0.00 -0.00 7.00 350.60 106.68 Tm (9) Tj ET BT /F2 1 Tf 7.00 0.00 -0.00 7.00 347.28 103.77 Tm (X) Tj ET Q q 0.000 0.000 0.000 RG 0.75 w [] 0 d 1 J 1 j 10.00 M 267.68 31.10 m 401.06 31.10 l S 267.68 31.10 m 267.68 27.65 l S 301.02 31.10 m 301.02 27.65 l S 334.37 31.10 m 334.37 27.65 l S 367.71 31.10 m 367.71 27.65 l S 401.06 31.10 m 401.06 27.65 l S BT 0.000 0.000 0.000 rg /F2 1 Tf 7.00 0.00 -0.00 7.00 263.69 18.66 Tm (-4) Tj ET BT /F2 1 Tf 7.00 0.00 -0.00 7.00 297.03 18.66 Tm (-2) Tj ET BT /F2 1 Tf 7.00 0.00 -0.00 7.00 332.42 18.66 Tm (0) Tj ET BT /F2 1 Tf 7.00 0.00 -0.00 7.00 365.77 18.66 Tm (2) Tj ET BT /F2 1 Tf 7.00 0.00 -0.00 7.00 399.11 18.66 Tm (4) Tj ET 244.34 51.40 m 244.34 167.37 l S 244.34 51.40 m 240.88 51.40 l S 244.34 80.39 m 240.88 80.39 l S 244.34 109.38 m 240.88 109.38 l S 244.34 138.37 m 240.88 138.37 l S 244.34 167.37 m 240.88 167.37 l S BT /F2 1 Tf 0.00 7.00 -7.00 0.00 236.04 47.41 Tm (-4) Tj ET BT /F2 1 Tf 0.00 7.00 -7.00 0.00 236.04 76.40 Tm (-2) Tj ET BT /F2 1 Tf 0.00 7.00 -7.00 0.00 236.04 107.44 Tm (0) Tj ET BT /F2 1 Tf 0.00 7.00 -7.00 0.00 236.04 136.43 Tm (2) Tj ET BT /F2 1 Tf 0.00 7.00 -7.00 0.00 236.04 165.42 Tm (4) Tj ET 244.34 31.10 m 424.40 31.10 l 424.40 187.66 l 244.34 187.66 l 244.34 31.10 l S Q q 216.00 0.00 216.00 216.00 re W n BT 0.000 0.000 0.000 rg /F3 1 Tf 9.00 0.00 -0.00 9.00 303.42 198.60 Tm [(V) 60 (ariety Eff) 10 (ects)] TJ ET BT /F2 1 Tf 7.00 0.00 -0.00 7.00 313.36 4.84 Tm (Component 1) Tj ET BT /F2 1 Tf 0.00 7.00 -7.00 0.00 222.22 88.37 Tm (Component 2) Tj ET Q endstream endobj 7 0 obj 4095 endobj 3 0 obj << /Type /Pages /Kids [ 5 0 R ] /Count 1 /MediaBox [0 0 432 216] >> endobj 4 0 obj << /ProcSet [/PDF /Text] /Font <> /ExtGState << >> >> endobj 8 0 obj << /Type /Encoding /BaseEncoding /WinAnsiEncoding /Differences [ 45/minus 96/quoteleft 144/dotlessi /grave /acute /circumflex /tilde /macron /breve /dotaccent /dieresis /.notdef /ring /cedilla /.notdef /hungarumlaut /ogonek /caron /space] >> endobj 9 0 obj << /Type /Font /Subtype /Type1 /Name /F2 /BaseFont /Helvetica /Encoding 8 0 R >> endobj 10 0 obj << /Type /Font /Subtype /Type1 /Name /F3 /BaseFont /Helvetica-Bold /Encoding 8 0 R >> endobj xref 0 11 0000000000 65535 f 0000000021 00000 n 0000000164 00000 n 0000004461 00000 n 0000004544 00000 n 0000000213 00000 n 0000000293 00000 n 0000004441 00000 n 0000004636 00000 n 0000004893 00000 n 0000004989 00000 n trailer << /Size 11 /Info 1 0 R /Root 2 0 R >> startxref 5091 %%EOF gnm/vignettes/gnmOverview.Rnw0000644000176000001440000037057012515147523016114 0ustar ripleyusers%\VignetteIndexEntry{Generalized nonlinear models in R: An overview of the gnm package} %\VignetteKeywords{Generalized Nonlinear Models} %\VignettePackage{gnm} \documentclass[a4paper]{article} \usepackage[english]{babel} % to avoid et~al with texi2pdf \usepackage{Sweave} %\usepackage{alltt} % now replaced by environments Sinput, Soutput, Scode \usepackage{amsmath} %\usepackage{times} %\usepackage[scaled]{couriers} \usepackage{txfonts} % Times, with Belleek math font and txtt for monospaced \usepackage[scaled=0.92]{helvet} %\usepackage[T1]{fontenc} %\usepackage[expert,altbullet,lucidasmallerscale]{lucidabr} \usepackage{booktabs} \usepackage[round,authoryear]{natbib} \usepackage[left=2cm,top=2.5cm,nohead]{geometry} \usepackage{hyperref} \usepackage{array} % for paragraph columns in tables %\usepackage{moreverb} \setkeys{Gin}{width=0.6\textwidth} %% The next few definitions from "Writing Vignettes for Bioconductor Packages" %% by R Gentleman \newcommand{\Robject}[1]{{\emph{\texttt{#1}}}} \newcommand{\Rfunction}[1]{{\emph{\texttt{#1}}}} \newcommand{\Rcode}[1]{{\emph{\texttt{#1}}}} \newcommand{\Rpackage}[1]{{\textsf{#1}}} \newcommand{\Rclass}[1]{{\emph{#1}}} \newcommand{\Rmethod}[1]{{\emph{\texttt{#1}}}} \newcommand{\Rfunarg}[1]{{\emph{\texttt{#1}}}} \newcommand{\R}{\textsf{R}} \newcommand\twiddle{{\char'176}} %\setlength{\oddsidemargin}{0.5in} %\setlength{\evensidemargin}{0.5in} %\setlength{\textwidth}{5.5in} \setlength{\itemindent}{1cm} \title{Generalized nonlinear models in \R: An overview of the \Rpackage{gnm} package} \author{Heather Turner and David Firth\footnote{ This work was supported by the Economic and Social Research Council (UK) through Professorial Fellowship RES-051-27-0055.}\\ \emph{University of Warwick, UK} } \date{For \Rpackage{gnm} version \Sexpr{packageDescription("gnm")[["Version"]]} , \Sexpr{Sys.Date()}} \begin{document} \maketitle {\small \tableofcontents } <>= options(SweaveHooks = list(eval = function() options(show.signif.stars = FALSE))) @ \section{Introduction} The \Rpackage{gnm} package provides facilities for fitting \emph{generalized nonlinear models}, i.e., regression models in which the link-transformed mean is described as a sum of predictor terms, some of which may be non-linear in the unknown parameters. Linear and generalized linear models, as handled by the \Rfunction{lm} and \Rfunction{glm} functions in \R, are included in the class of generalized nonlinear models, as the special case in which there is no nonlinear term. This document gives an extended overview of the \Rpackage{gnm} package, with some examples of applications. The primary package documentation in the form of standard help pages, as viewed in \R\ by, for example, \Rcode{?gnm} or \Rcode{help(gnm)}, is supplemented rather than replaced by the present document. We begin below with a preliminary note (Section \ref{sec:glms}) on some ways in which the \Rpackage{gnm} package extends \R's facilities for specifying, fitting and working with generalized \emph{linear} models. Then (Section \ref{sec:nonlinear} onwards) the facilities for nonlinear terms are introduced, explained and exemplified. The \Rpackage{gnm} package is installed in the standard way for CRAN packages, for example by using \Rfunction{install.packages}. Once installed, the package is loaded into an \R\ session by <>= library(gnm) @ \section{Generalized linear models} \label{sec:glms} \subsection{Preamble} Central to the facilities provided by the \Rpackage{gnm} package is the model-fitting function \Rfunction{gnm}, which interprets a model formula and returns a model object. The user interface of \Rfunction{gnm} is patterned after \Rfunction{glm} (which is included in \R's standard \Rpackage{stats} package), and indeed \Rfunction{gnm} can be viewed as a replacement for \Rfunction{glm} for specifying and fitting generalized linear models. In general there is no reason to prefer \Rfunction{gnm} to \Rfunction{glm} for fitting generalized linear models, except perhaps when the model involves a large number of incidental parameters which are treatable by \Rfunction{gnm}'s \emph{eliminate} mechanism (see Section \ref{sec:eliminate}). While the main purpose of the \Rpackage{gnm} package is to extend the class of models to include nonlinear terms, some of the new functions and methods can be used also with the familiar \Rfunction{lm} and \Rfunction{glm} model-fitting functions. These are: three new data-manipulation functions \Rfunction{Diag}, \Rfunction{Symm} and \Rfunction{Topo}, for setting up structured interactions between factors; a new \Rclass{family} function, \Rfunction{wedderburn}, for modelling a continuous response variable in $[0,1]$ with the variance function $V(\mu) = \mu^2(1-\mu)^2$ as in \citet{Wedd74}; and a new generic function \Rfunction{termPredictors} which extracts the contribution of each term to the predictor from a fitted model object. These functions are briefly introduced here, before we move on to the main purpose of the package, nonlinear models, in Section \ref{sec:nonlinear}. \subsection{\Rfunction{Diag} and \Rfunction{Symm}} When dealing with \emph{homologous} factors, that is, categorical variables whose levels are the same, statistical models often involve structured interaction terms which exploit the inherent symmetry. The functions \Rfunction{Diag} and \Rfunction{Symm} facilitate the specification of such structured interactions. As a simple example of their use, consider the log-linear models of \emph{quasi-independence}, \emph{quasi-symmetry} and \emph{symmetry} for a square contingency table. \citet{Agre02}, Section 10.4, gives data on migration between regions of the USA between 1980 and 1985: <>= count <- c(11607, 100, 366, 124, 87, 13677, 515, 302, 172, 225, 17819, 270, 63, 176, 286, 10192 ) region <- c("NE", "MW", "S", "W") row <- gl(4, 4, labels = region) col <- gl(4, 1, length = 16, labels = region) @ The comparison of models reported by Agresti can be achieved as follows: <>= independence <- glm(count ~ row + col, family = poisson) quasi.indep <- glm(count ~ row + col + Diag(row, col), family = poisson) symmetry <- glm(count ~ Symm(row, col), family = poisson) quasi.symm <- glm(count ~ row + col + Symm(row, col), family = poisson) comparison1 <- anova(independence, quasi.indep, quasi.symm) print(comparison1, digits = 7) comparison2 <- anova(symmetry, quasi.symm) print(comparison2) @ The \Rfunction{Diag} and \Rfunction{Symm} functions also generalize the notions of diagonal and symmetric interaction to cover situations involving more than two homologous factors. \subsection{\Rfunction{Topo}} More general structured interactions than those provided by \Rfunction{Diag} and \Rfunction{Symm} can be specified using the function \Rfunction{Topo}. (The name of this function is short for `topological interaction', which is the nomenclature often used in sociology for factor interactions with structure derived from subject-matter theory.) The \Rfunction{Topo} function operates on any number ($k$, say) of input factors, and requires an argument named \Rfunarg{spec} which must be an array of dimension $L_1 \times \ldots \times L_k$, where $L_i$ is the number of levels for the $i$th factor. The \Rfunarg{spec} argument specifies the interaction level corresponding to every possible combination of the input factors, and the result is a new factor representing the specified interaction. As an example, consider fitting the `log-multiplicative layer effects' models described in \citet{Xie92}. The data are 7 by 7 versions of social mobility tables from \citet{Erik82}: <>= ### Collapse to 7 by 7 table as in Erikson et al. (1982) erikson <- as.data.frame(erikson) lvl <- levels(erikson$origin) levels(erikson$origin) <- levels(erikson$destination) <- c(rep(paste(lvl[1:2], collapse = " + "), 2), lvl[3], rep(paste(lvl[4:5], collapse = " + "), 2), lvl[6:9]) erikson <- xtabs(Freq ~ origin + destination + country, data = erikson) @ From sociological theory --- for which see \citet{Erik82} or \citet{Xie92} --- the log-linear interaction between origin and destination is assumed to have a particular structure: \begin{Sinput} > levelMatrix <- matrix(c(2, 3, 4, 6, 5, 6, 6, + 3, 3, 4, 6, 4, 5, 6, + 4, 4, 2, 5, 5, 5, 5, + 6, 6, 5, 1, 6, 5, 2, + 4, 4, 5, 6, 3, 4, 5, + 5, 4, 5, 5, 3, 3, 5, + 6, 6, 5, 3, 5, 4, 1), 7, 7, byrow = TRUE) \end{Sinput} The models of table 3 of \citet{Xie92} can now be fitted as follows: \begin{Sinput} > ## Null association between origin and destination > nullModel <- gnm(Freq ~ country:origin + country:destination, + family = poisson, data = erikson, verbose = FALSE) > > ## Interaction specified by levelMatrix, common to all countries > commonTopo <- update(nullModel, ~ . + + Topo(origin, destination, spec = levelMatrix), + verbose = FALSE) > > ## Interaction specified by levelMatrix, different multiplier for each country > multTopo <- update(nullModel, ~ . + + Mult(Exp(country), Topo(origin, destination, spec = levelMatrix)), + verbose = FALSE) > > ## Interaction specified by levelMatrix, different effects for each country > separateTopo <- update(nullModel, ~ . + + country:Topo(origin, destination, spec = levelMatrix), + verbose = FALSE) > > anova(nullModel, commonTopo, multTopo, separateTopo) \end{Sinput} \begin{Soutput} Analysis of Deviance Table Model 1: Freq ~ country:origin + country:destination Model 2: Freq ~ Topo(origin, destination, spec = levelMatrix) + country:origin + country:destination Model 3: Freq ~ Mult(country, Topo(origin, destination, spec = levelMatrix)) + country:origin + country:destination Model 4: Freq ~ country:origin + country:destination + country:Topo(origin, destination, spec = levelMatrix) Resid. Df Resid. Dev Df Deviance 1 108 4860.0 2 103 244.3 5 4615.7 3 101 216.4 2 28.0 4 93 208.5 8 7.9 \end{Soutput} Here we have used \Rfunction{gnm} to fit all of these log-link models; the first, second and fourth are log-linear and could equally well have been fitted using \Rfunction{glm}. \subsection{The \Rfunction{wedderburn} family} In \citet{Wedd74} it was suggested to represent the mean of a continuous response variable in $[0,1]$ using a quasi-likelihood model with logit link and the variance function $\mu^2(1-\mu)^2$. This is not one of the variance functions made available as standard in \R's \Rfunction{quasi} family. The \Rfunction{wedderburn} family provides it. As an example, Wedderburn's analysis of data on leaf blotch on barley can be reproduced as follows: <>= ## data from Wedderburn (1974), see ?barley logitModel <- glm(y ~ site + variety, family = wedderburn, data = barley) fit <- fitted(logitModel) print(sum((barley$y - fit)^2 / (fit * (1-fit))^2)) @ This agrees with the chi-squared value reported on page 331 of \citet{McCu89}, which differs slightly from Wedderburn's own reported value. \subsection{\Rfunction{termPredictors}} \label{sec:termPredictors} The generic function \Rfunction{termPredictors} extracts a term-by-term decomposition of the predictor function in a linear, generalized linear or generalized nonlinear model. As an illustrative example, we can decompose the linear predictor in the above quasi-symmetry model as follows: <>= print(temp <- termPredictors(quasi.symm)) rowSums(temp) - quasi.symm$linear.predictors @ Such a decomposition might be useful, for example, in assessing the relative contributions of different terms or groups of terms. \section{Nonlinear terms} \label{sec:nonlinear} The main purpose of the \Rpackage{gnm} package is to provide a flexible framework for the specification and estimation of generalized models with nonlinear terms. The facility provided with \Rfunction{gnm} for the specification of nonlinear terms is designed to be compatible with the symbolic language used in \Rclass{formula} objects. Primarily, nonlinear terms are specified in the model formula as calls to functions of the class \Rclass{nonlin}. There are a number of \Rclass{nonlin} functions included in the \Rpackage{gnm} package. Some of these specify simple mathematical functions of predictors: \Rfunction{Exp}, \Rfunction{Mult}, and \Rfunction{Inv}. %\Rfunction{Log}, \Rfunction{Raise} (to raise to a constant power), and \Rfunction{Logit}. Others specify more specialized nonlinear terms, in particular \Rfunction{MultHomog} specifies homogeneous multiplicative interactions and \Rfunction{Dref} specifies diagonal reference terms. Users may also define their own \Rclass{nonlin} functions. \subsection{Basic mathematical functions of predictors} \label{sec:Basic} Most of the \Rclass{nonlin} functions included in \Rpackage{gnm} are basic mathematical functions of predictors: \begin{description} \setlength{\itemindent}{-0.5cm} \item[\Rfunction{Exp}:] the exponential of a predictor \item[\Rfunction{Inv}:] the reciprocal of a predictor %\item[\Rfunction{Log}:] the natural logarithm of a predictor %\item[\Rfunction{Logit}:] the logit of a predictor \item[\Rfunction{Mult}:] the product of predictors %\item[\Rfunction{Raise}:] a predictor raised to a constant power \end{description} Predictors are specified by symbolic expressions that are interpreted as the right-hand side of a \Rclass{formula} object, except that an intercept is \textbf{not} added by default. The predictors may contain nonlinear terms, allowing more complex functions to be built up. For example, suppose we wanted to specify a logistic predictor with the same form as that used by \Rfunction{SSlogis} (a selfStart model for use with \Rfunction{nls} --- see section~\ref{sec:gnmVnls} for more on \Rfunction{gnm} vs.\ \Rfunction{nls}): \[\frac{\text{Asym}}{1 + \exp((\text{xmid} - x)/\text{scal})}.\] This expression could be simplified by re-parameterizing in terms of xmid/scal and 1/scal, however we shall continue with this form for illustration. We could express this predictor symbolically as follows \begin{Scode} ~ -1 + Mult(1, Inv(Const(1) + Exp(Mult(1 + offset(-x), Inv(1))))) \end{Scode} where \Rfunction{Const} is a convenience function to specify a constant in a \Rclass{nonlin} term, equivalent to \Rcode{offset(rep(1, nObs))} where \Robject{nObs} is the number of observations. However, this is rather convoluted and it may be preferable to define a specialized \Rclass{nonlin} function in such a case. Section \ref{sec:nonlin.functions} explains how users can define custom \Rclass{nonlin} functions, with a function to specify logistic terms as an example. One family of models usefully specified with the basic functions is the family of models with multiplicative interactions. For example, the row-column association model \[ \log \mu_{rc} = \alpha_r + \beta_c + \gamma_r\delta_c, \] also known as the Goodman RC model \citep{Good79}, would be specified as a log-link model (for response variable \Robject{resp}, say), with formula \begin{Scode} resp ~ R + C + Mult(R, C) \end{Scode} where \Robject{R} and \Robject{C} are row and column factors respectively. In some contexts, it may be desirable to constrain one or more of the constituent multipliers\footnote{ A note on terminology: the rather cumbersome phrase `constituent multiplier', or sometimes the abbreviation `multiplier', will be used throughout this document in preference to the more elegant and standard mathematical term `factor'. This will avoid possible confusion with the completely different meaning of the word `factor' --- that is, a categorical variable --- in \R. } in a multiplicative interaction to be nonnegative . This may be achieved by specifying the multiplier as an exponential, as in the following `uniform difference' model \citep{Xie92, Erik92} \[ \log \mu_{rct} = \alpha_{rt} + \beta_{ct} + e^{\gamma_t}\delta_{rc}, \] which would be represented by a formula of the form \begin{Scode} resp ~ R:T + C:T + Mult(Exp(T), R:C) \end{Scode} \subsection{\Rfunction{MultHomog}} \Rfunction{MultHomog} is a \Rclass{nonlin} function to specify multiplicative interaction terms in which the constituent multipliers are the effects of two or more factors and the effects of these factors are constrained to be equal when the factor levels are equal. The arguments of \Rfunction{MultHomog} are the factors in the interaction, which are assumed to be objects of class \Rclass{factor}. As an example, consider the following association model with homogeneous row-column effects: \[\log \mu_{rc} = \alpha_r + \beta_c + \theta_{r}I(r=c) + \gamma_r\gamma_c.\] To fit this model, with response variable named \Robject{resp}, say, the formula argument to \Rfunction{gnm} would be \begin{Scode} resp ~ R + C + Diag(R, C) + MultHomog(R, C) \end{Scode} If the factors passed to \Rfunction{MultHomog} do not have exactly the same levels, a common set of levels is obtained by taking the union of the levels of each factor, sorted into increasing order. \subsection{\Rfunction{Dref}} \label{sec:Dref function} \Rfunction{Dref} is a \Rclass{nonlin} function to fit diagonal reference terms \citep{Sobe81, Sobe85} involving two or more factors with a common set of levels. A diagonal reference term comprises an additive component for each factor. The component for factor $f$ is given by \[ w_f\gamma_l \] for an observation with level $l$ of factor $f$, where $w_f$ is the weight for factor $f$ and $\gamma_l$ is the ``diagonal effect'' for level $l$. The weights are constrained to be nonnegative and to sum to one so that a ``diagonal effect'', say $\gamma_l$, is the value of the diagonal reference term for data points with level $l$ across the factors. \Rfunction{Dref} specifies the constraints on the weights by defining them as \[ w_f = \frac{e^{\delta_f}}{\sum_i e^{\delta_i}} \] where the $\delta_f$ are the parameters to be estimated. Factors defining the diagonal reference term are passed as unspecified arguments to \Rfunction{Dref}. For example, the following diagonal reference model for a contingency table classified by the row factor \Robject{R} and the column factor \Robject{C}, \[ \mu_{rc} =\frac{e^{\delta_1}}{e^{\delta_1} + e^{\delta_2}}\gamma_r + \frac{e^{\delta_2}}{e^{\delta_1} + e^{\delta_2}}\gamma_c, \] would be specified by a formula of the form \begin{Scode} resp ~ -1 + Dref(R, C) \end{Scode} The \Rfunction{Dref} function has one specified argument, \Rfunarg{delta}, which is a formula with no left-hand side, specifying the dependence (if any) of $\delta_f$ on covariates. For example, the formula \begin{Scode} resp ~ -1 + x + Dref(R, C, delta = ~ 1 + x) \end{Scode} specifies the generalized diagonal reference model \[ \mu_{rci} = \beta x_i + \frac{e^{\xi_{01} + \xi_{11}x_i}}{e^{\xi_{01} + \xi_{11}x_i} + e^{\xi_{02} + \xi_{12}x_i}}\gamma_r + \frac{e^{\xi_{02} + \xi_{12}x_i}}{e^{\xi_{01} + \xi_{11}x_i} + e^{\xi_{02} + \xi_{12}x_i}}\gamma_c. \] The default value of \Rfunarg{delta} is \Robject{\twiddle 1}, so that constant weights are estimated. The coefficients returned by \Rfunction{gnm} are those that are directly estimated, i.e. the $\delta_f$ or the $\xi_{.f}$, rather than the implied weights $w_f$. However, these weights may be obtained from a fitted model using the \Rfunction{DrefWeights} function, which computes the corresponding standard errors using the delta method. \subsection{\Rfunction{instances}} \label{sec:instances} Multiple instances of a linear term will be aliased with each other, but this is not necessarily the case for nonlinear terms. Indeed, there are certain types of model where adding further instances of a nonlinear term is a natural way to extend the model. For example, Goodman's RC model, introduced in section \ref{sec:Basic} \[ \log \mu_{rc} = \alpha_r + \beta_c + \gamma_r\delta_c, \] is naturally extended to the RC(2) model, with a two-component interaction \[ \log \mu_{rc} = \alpha_r + \beta_c + \gamma_r\delta_c + \theta_r\phi_c. \] Currently all of the \Rclass{nonlin} functions in \Rpackage{gnm} except \Rpackage{Dref} have an \Rfunarg{inst} argument to allow the specification of multiple instances. So the RC(2) model could be specified as follows \begin{Scode} resp ~ R + C + Mult(R, C, inst = 1) + Mult(R, C, inst = 2) \end{Scode} The convenience function \Rfunction{instances} allows multiple instances of a term to be specified at once \begin{Scode} resp ~ R + C + instances(Mult(R, C), 2) \end{Scode} The formula is expanded by \Rfunction{gnm}, so that the instances are treated as separate terms. The \Rfunction{instances} function may be used with any function with an \Rfunarg{inst} argument. \subsection{Custom \Rclass{nonlin} functions} \label{sec:nonlin.functions} \subsubsection{General description} Users may write their own \Rclass{nonlin} functions to specify nonlinear terms which can not (easily) be specified using the \Rclass{nonlin} functions in the \Rpackage{gnm} package. A function of class \Rclass{nonlin} should return a list of arguments for the internal function \Rfunction{nonlinTerms}. The following arguments must be specified in all cases: \begin{description} \setlength{\itemindent}{-0.5cm} \item[\Robject{predictors}:] a list of symbolic expressions or formulae with no left hand side which represent (possibly nonlinear) predictors that form part of the term. \item[\Robject{term}:] a function that takes the arguments \Rfunarg{predLabels} and \Rfunarg{varLabels}, which are labels generated by \Rfunction{gnm} for the specified predictors and variables (see below), and returns a deparsed mathematical expression of the nonlinear term. Only functions recognised by \Rfunction{deriv} should be used in the expression, e.g. \Rfunction{+} rather than \Rfunction{sum}. \end{description} Intercepts are added by default to predictors that are specified by formulae. If predictors are named, these names are used as a prefix for parameter labels or as the parameter label itself in the single-parameter case. The following arguments of \Rfunction{nonlinTerms} must be specified whenever applicable to the nonlinear term: \begin{description} \setlength{\itemindent}{-0.5cm} \item[\Robject{variables}:] a list of expressions representing variables in the term (variables with a coefficient of 1). \item[\Robject{common}:] a numeric index of \Rfunarg{predictors} with duplicated indices identifying single factor predictors for which homologous effects are to be estimated. \end{description} The arguments below are optional: \begin{description} \setlength{\itemindent}{-0.5cm} \item[\Robject{call}:] a call to be used as a prefix for parameter labels. \item[\Robject{match}:] (if \Robject{call} is non-\Rcode{NULL}) a numeric index of \Robject{predictors} specifying which arguments of \Robject{call} the predictors match to --- zero indicating no match. If \Rcode{NULL}, predictors will not be matched to the arguments of \Robject{call}. \item[\Robject{start}:] a function which takes a named vector of parameters corresponding to the predictors and returns a vector of starting values for those parameters. This function is ignored if the term is nested within another nonlinear term. \end{description} Predictors which are matched to a specified argument of \Robject{call} should be given the same name as the argument. Matched predictors are labelled using ``dot-style'' labelling, e.g. the label for the intercept in the first constituent multiplier of the term \Rcode{Mult(A, B)} would be \Rcode{"Mult(.\ + A, 1 + B).(Intercept)"}. It is recommended that matches are specified wherever possible, to ensure parameter labels are well-defined. The arguments of \Rclass{nonlin} functions are as suited to the particular term, but will usually include symbolic representations of predictors in the term and/or the names of variables in the term. The function may also have an \Rfunarg{inst} argument to allow specification of multiple instances (see \ref{sec:instances}). \subsubsection{Example: a logistic function} As an example, consider writing a \Rclass{nonlin} function for the logistic term discussed in \ref{sec:Basic}: \[\frac{\text{Asym}}{1 + \exp((\text{xmid} - x)/\text{scal})}.\] We can consider \emph{Asym}, \emph{xmid} and \emph{scal} as the parameters of three separate predictors, each with a single intercept term. Thus we specify the \Rfunarg{predictors} argument to \Rfunction{nonlinTerms} as \begin{Scode} predictors = list(Asym = 1, xmid = 1, scal = 1) \end{Scode} The term also depends on the variable $x$, which would need to be specified by the user. Suppose this is specified to our \Rclass{nonlin} function through an argument named \Rfunarg{x}. Then our \Rclass{nonlin} function would specify the following \Rfunarg{variables} argument \begin{Scode} variables = list(substitute(x)) \end{Scode} We need to use \Rfunction{substitute} here to list the variable specified by the user rather than the variable named \Rcode{``x''} (if it exists). Our \Rclass{nonlin} function must also specify the \Rfunarg{term} argument to \Rfunction{nonlinTerms}. This is a function that will paste together an expression for the term, given labels for the predictors and the variables: \begin{Scode} term = function(predLabels, varLabels) { paste(predLabels[1], "/(1 + exp((", predLabels[2], "-", varLabels[1], ")/", predLabels[3], "))") } \end{Scode} We now have all the necessary ingredients of a \Rclass{nonlin} function to specify the logistic term. Since the parameterization does not depend on user-specified values, it does not make sense to use call-matched labelling in this case. The labels for our parameters will be taken from the labels of the \Rfunarg{predictors} argument. Since we do not anticipate fitting models with multiple logistic terms, our \Rclass{nonlin} function will not specify a \Rfunarg{call} argument with which to prefix the parameter labels. We do however, have some idea of useful starting values, so we will specify the \Rfunarg{start} argument as \begin{Scode} start = function(theta){ theta[3] <- 1 theta } \end{Scode} which sets the initial scale parameter to one. Putting all these ingredients together we have \begin{Scode} Logistic <- function(x){ list(predictors = list(Asym = 1, xmid = 1, scal = 1), variables = list(substitute(x)), term = function(predLabels, varLabels) { paste(predLabels[1], "/(1 + exp((", predLabels[2], "-", varLabels[1], ")/", predLabels[3], "))") }, start = function(theta){ theta[3] <- 1 theta }) } class(Logistic) <- "nonlin" \end{Scode} \subsubsection{Example: \Rfunction{MultHomog}} The \Rfunction{MultHomog} function included in the \Rpackage{gnm} package provides a further example of a \Rclass{nonlin} function, showing how to specify a term with quite different features from the preceding example. The definition is \begin{Scode} MultHomog <- function(..., inst = NULL){ dots <- match.call(expand.dots = FALSE)[["..."]] list(predictors = dots, common = rep(1, length(dots)), term = function(predLabels, ...) { paste("(", paste(predLabels, collapse = ")*("), ")", sep = "")}, call = as.expression(match.call())) } class(MultHomog) <- "nonlin" \end{Scode} Firstly, the interaction may be based on any number of factors, hence the use of the special ``\Rfunarg{...}'' argument. The use of \Rfunction{match.call} is analogous to the use of \Rfunction{substitute} in the \Rfunction{Logistic} function: to obtain expressions for the factors as specified by the user. The returned \Rfunarg{common} argument specifies that homogeneous effects are to be estimated across all the specified factors. The term only depends on these factors, but the \Rfunarg{term} function allows for the empty \Robject{varLabels} vector that will be passed to it, by having a ``\Rfunarg{...}'' argument. Since the user may wish to specify multiple instances, the \Rfunarg{call} argument to \Rfunction{nonlinTerms} is specified, so that parameters in different instances of the term will have unique labels (due to the \Rfunarg{inst} argument in the call). However as the expressions passed to ``\Rfunarg{...}'' may only represent single factors, rather than general predictors, it is not necessary to use call-matched labelling, so the \Rfunarg{match} argument is not specified here. % Dref starting values as example of ensuring the arbitrariness of the final % parameterization is emphasised (see old plug-in section)? \section{Controlling the fitting procedure} The \Rfunction{gnm} function has a number of arguments which affect the way a model will be fitted. Basic control parameters can be set using the arguments %\Rfunarg{checkLinear}, \Rfunarg{lsMethod}, \Rfunarg{ridge}, \Rfunarg{tolerance}, \Rfunarg{iterStart} and \Rfunarg{iterMax}. Starting values for the parameter estimates can be set by \Rfunarg{start} or they can be generated from starting values for the predictors on the link or response scale via \Rfunarg{etastart} or \Rfunarg{mustart} respectively. Parameters can be constrained via \Rfunarg{constrain} and \Rfunarg{constrainTo} arguments, while parameters of a stratification factor can be handled more efficiently by specifying the factor in an \Rfunarg{eliminate} argument. These options are described in more detail below. \subsection{Basic control parameters} %By default, \Rfunction{gnm} will use \Rfunction{glm.fit} to fit models where the %predictor is linear and \Rfunarg{eliminate} is \Rcode{NULL}. This behaviour can %be overridden by setting \Rfunarg{checkLinear} to \Rcode{FALSE}. %%% At present there is no advantage to doing this! Parameterization would be %%% the same. The arguments \Rfunarg{iterStart} and \Rfunarg{iterMax} control respectively the number of starting iterations (where applicable) and the number of main iterations used by the fitting algorithm. The progress of these iterations can be followed by setting either \Rfunarg{verbose} or \Rfunarg{trace} to \Robject{TRUE}. If \Rfunarg{verbose} is \Robject{TRUE} and \Rfunarg{trace} is \Robject{FALSE}, which is the default setting, progress is indicated by printing the character ``.'' at the beginning of each iteration. If \Rfunarg{trace} is \Robject{TRUE}, the deviance is printed at the beginning of each iteration (over-riding the printing of ``.'' if necessary). Whenever \Rfunarg{verbose} is \Robject{TRUE}, additional messages indicate each stage of the fitting process and diagnose any errors that cause that cause the algorithm to restart. Prior to solving the (typically rank-deficient) least squares problem at the heart of the \Rfunction{gnm} fitting algorithm, the design matrix is standardized and regularized (in the Levenberg-Marquardt sense); the \Rfunarg{ridge} argument provides a degree of control over the regularization performed (smaller values may sometimes give faster convergence but can lead to numerical instability). The fitting algorithm will terminate before the number of main iterations has reached \Rfunarg{iterMax} if the convergence criteria have been met, with tolerance specified by \Rfunarg{tolerance}. Convergence is judged by comparing the squared components of the score vector with corresponding elements of the diagonal of the Fisher information matrix. If, for all components of the score vector, the ratio is less than \Robject{tolerance\^{}2}, or the corresponding diagonal element of the Fisher information matrix is less than 1e-20, the algorithm is deemed to have converged. \subsection{Specifying starting values} \label{sec:start} \subsubsection{Using \Rfunarg{start}} In some contexts, the default starting values may not be appropriate and the fitting algorithm will fail to converge, or perhaps only converge after a large number of iterations. Alternative starting values may be passed on to \Rfunction{gnm} by specifying a \Rfunarg{start} argument. This should be a numeric vector of length equal to the number of parameters (or possibly the non-eliminated parameters, see Section \ref{sec:eliminate}), however missing starting values (\Robject{NA}s) are allowed. If there is no user-specified starting value for a parameter, the default value is used. This feature is particularly useful when adding terms to a model, since the estimates from the original model can be used as starting values, as in this example: \begin{Scode} model1 <- gnm(mu ~ R + C + Mult(R, C)) model2 <- gnm(mu ~ R + C + instances(Mult(R, C), 2), start = c(coef(model1), rep(NA, 10))) \end{Scode} The \Rfunction{gnm} call can be made with \Rcode{method = "coefNames"} to identify the parameters of a model prior to estimation, to assist with the specification of arguments such as \Rfunarg{start}. For example, to get the number \Rcode{10} for the value of \Rfunarg{start} above, we could have done \begin{Scode} gnm(mu ~ R + C + instances(Mult(R, C), 2), method = "coefNames") \end{Scode} from whose output it would be seen that there are 10 new coefficients in \Robject{model2}. When called with \Rcode{method = "coefNames"}, \Rfunction{gnm} makes no attempt to fit the specified model; instead it returns just the names that the coefficients in the fitted model object would have. The starting procedure used by \Rfunction{gnm} is as follows: \begin{enumerate} \item Begin with all parameters set to \Rcode{NA}. \item \label{i:nonlin} Replace \Rcode{NA} values with any starting values set by \Rclass{nonlin} functions. \item \label{i:start} Replace current values with any (non-\Rcode{NA}) starting values specified by the \Rfunarg{start} argument of \Rfunction{gnm}. \item \label{i:constrain} Set any values specified by the \Rfunarg{constrain} argument to the values specified by the \Rfunarg{constrainTo} argument (see Section \ref{sec:constrain}). \item \label{i:gnmStart} Categorise remaining \Rcode{NA} parameters as linear or nonlinear, treating non-\Rcode{NA} parameters as fixed. Initialise the nonlinear parameters by generating values $\theta_i$ from the Uniform($-0.1$, $0.1$) distribution and shifting these values away from zero as follows \begin{equation*} \theta_i = \begin{cases} \theta_i - 0.1 & \text{if } \theta_i < 1 \\ \theta_i + 0.1 & \text{otherwise} \end{cases} \end{equation*} \item Compute the \Rfunction{glm} estimate of the linear parameters, offsetting the contribution to the predictor of any terms fully determined by steps \ref{i:nonlin} to \ref{i:gnmStart}. \item \label{i:iter} Run starting iterations: update nonlinear parameters one at a time, jointly re-estimating linear parameters after each round of updates. \end{enumerate} Note that no starting iterations (step \ref{i:iter}) will be run if all parameters are linear, or if all nonlinear parameters are specified by \Rfunarg{start}, \Rfunarg{constrain} or a \Rclass{nonlin} function. \subsubsection{Using \Rfunarg{etastart} or \Rfunarg{mustart}} An alternative way to set starting values for the parameters is to specify starting values for the predictors. If there are linear parameters in the model, the predictor starting values are first used to fit a model with only the linear terms (offsetting any terms fully specified by starting values given by \Rfunarg{start}, \Rfunarg{constrain} or a \Rclass{nonlin} function). In this case the parameters corresponding to the predictor starting values can be computed analytically. If the fitted model reproduces the predictor starting values, then these values contain no further information and they are replaced using the \Rfunction{initialize} function of the specified \Rfunarg{family}. The predictor starting values or their replacement are then used as the response variable in a nonlinear least squares model with only the unspecified nonlinear terms, offsetting the contribution of any other terms. Since the model is over-parameterized, the model is approximated using \Rfunarg{iterStart} iterations of the ``L-BFGS-B'' algorithm of \Rfunction{optim}, assuming parameters lie in the range (-10, 10). Starting values for the predictors can be specified explicitly via \Rfunarg{etastart} or implicitly by passing starting values for the fitted means to \Rfunarg{mustart}. For example, when extending a model, the fitted predictors from the first model can be used to find starting values for the parameters of the second model: \begin{Scode} model1 <- gnm(mu ~ R + C + Mult(R, C)) model2 <- gnm(mu ~ R + C + instances(Mult(R, C), 2), etastart = model1$predictors) \end{Scode} %$ Using \Rfunction{etastart} avoids the one-parameter-at-a-time starting iterations, so is quicker than using \Rfunction{start} to pass on information from a nested model. However \Rfunction{start} will generally produce better starting values so should be used when feasible. For multiplicative terms, the \Rfunction{residSVD} functions provides a better way to avoid starting iterations. \subsection{Using \Rfunarg{constrain}} \label{sec:constrain} By default, \Rfunction{gnm} only imposes identifiability constraints according to the general conventions used by \Robject{R} to handle linear aliasing. Therefore models that have any nonlinear terms will be typically be over-parameterized, and \Rfunction{gnm} will return a random parameterization for unidentified coefficients (determined by the randomly chosen starting values for the iterative algorithm, step 5 above). To illustrate this point, consider the following application of \Rfunction{gnm}, discussed later in Section \ref{sec:RCmodels}: <>= set.seed(1) RChomog1 <- gnm(Freq ~ origin + destination + Diag(origin, destination) + MultHomog(origin, destination), family = poisson, data = occupationalStatus, verbose = FALSE) @ Running the analysis again from a different seed <>= set.seed(2) RChomog2 <- update(RChomog1) @ gives a different representation of the same model: <>= compareCoef <- cbind(coef(RChomog1), coef(RChomog2)) colnames(compareCoef) <- c("RChomog1", "RChomog2") round(compareCoef, 4) @ Even though the linear terms are constrained, the parameter estimates for the main effects of \Robject{origin} and \Robject{destination} still change, because these terms are aliased with the higher order multiplicative interaction, which is unconstrained. Standard errors are only meaningful for identified parameters and hence the output of \Rmethod{summary.gnm} will show clearly which coefficients are estimable: <>= summary(RChomog2) @ Additional constraints may be specified through the \Rfunarg{constrain} and \Rfunarg{constrainTo} arguments of \Rfunction{gnm}. These arguments specify respectively parameters that are to be constrained in the fitting process and the values to which they should be constrained. Parameters may be specified by a regular expression to match against the parameter names, a numeric vector of indices, a character vector of names, or, if \Rcode{constrain = "[?]"} they can be selected through a \emph{Tk} dialog. The values to constrain to should be specified by a numeric vector; if \Rfunarg{constrainTo} is missing, constrained parameters will be set to zero. In the case above, constraining one level of the homogeneous multiplicative factor is sufficient to make the parameters of the nonlinear term identifiable, and hence all parameters in the model identifiable. Figure~\ref{fig:Tk} illustrates how the coefficient to be constrained may be specified via a \emph{Tk} dialog, an approach which can be helpful in interactive R sessions. % here illustrate TclTk dialog, but explain other methods better for reproducibility \begin{figure}[tp] \centering \begin{tabular}[!h]{m{0.6\linewidth}m{0.4\linewidth}} \scalebox{0.9}{\includegraphics{screenshot1.png}} & When \Rfunction{gnm} is called with \Rcode{constrain = "[?]"}, a \emph{Tk} dialog is shown listing the coefficients in the model.\\ \scalebox{0.9}{\includegraphics{screenshot2.png}} & Scroll through the coefficients and click to select a single coefficient to constrain. To select multiple coefficients, hold down the \texttt{Ctrl} key whilst clicking. The \texttt{Add} button will become active when coefficient(s) have been selected.\\ \scalebox{0.9}{\includegraphics{screenshot3.png}} & Click the \texttt{Add} button to add the selected coefficients to the list of coefficients to be constrained. To remove coefficients from the list, select the coefficients in the right pane and click \texttt{Remove}. Click \texttt{OK} when you have finalised the list.\\ \end{tabular} \caption{Selecting coefficients to constrain with the \emph{Tk} dialog.} \label{fig:Tk} \end{figure} However for reproducible code, it is best to specify the constrained coefficients directly. For example, the following code specifies that the last level of the homogeneous multiplicative factor should be constrained to zero, <>= set.seed(1) RChomogConstrained1 <- update(RChomog1, constrain = length(coef(RChomog1))) @ Since all the parameters are now constrained, re-fitting the model will give the same results, regardless of the random seed set beforehand: <>= set.seed(2) RChomogConstrained2 <- update(RChomogConstrained1) identical(coef(RChomogConstrained1), coef(RChomogConstrained2)) @ It is not usually so straightforward to constrain all the parameters in a generalized nonlinear model. However use of \Rfunarg{constrain} in conjunction with \Rfunarg{constrainTo} is usually sufficient to make coefficients of interest identifiable . The functions \Rfunction{checkEstimable} or \Rfunction{getContrasts}, described in Section \ref{sec:Methods}, may be used to check whether particular combinations of parameters are estimable. \subsection{Using \Rfunarg{eliminate}} \label{sec:eliminate} When a model contains the additive effect of a factor which has a large number of levels, the iterative algorithm by which maximum likelihood estimates are computed can usually be accelerated by use of the \Rfunarg{eliminate} argument to \Rfunction{gnm}. A factor passed to \Rfunarg{eliminate} specifies the first term in the model, replacing any intercept term. So, for example \begin{Scode} gnm(mu ~ A + B + Mult(A, B), eliminate = strata1:strata2) \end{Scode} is equivalent, in terms of the structure of the model, to \begin{Scode} gnm(mu ~ -1 + strata1:strata2 + A + B + Mult(A, B)) \end{Scode} However, specifying a factor through \Rfunarg{eliminate} has two advantages over the standard specification. First, the structure of the eliminated factor is exploited so that computational speed is improved --- substantially so if the number of eliminated parameters is large. Second, eliminated parameters are returned separately from non-eliminated parameters (as an attribute of the \Robject{coefficients} component of the returned object). Thus eliminated parameters are excluded from printed model summaries by default and disregarded by \Rclass{gnm} methods that would not be relevant to such parameters (see Section \ref{sec:Methods}). The \Rfunarg{eliminate} feature is useful, for example, when multinomial-response models are fitted by using the well known equivalence between multinomial and (conditional) Poisson likelihoods. In such situations the sufficient statistic involves a potentially large number of fixed multinomial row totals, and the corresponding parameters are of no substantive interest. For an application see Section \ref{sec:Stereotype} below. Here we give an artificial illustration: 1000 randomly-generated trinomial responses, and a single predictor variable (whose effect on the data generation is null): <>= set.seed(1) n <- 1000 x <- rep(rnorm(n), rep(3, n)) counts <- as.vector(rmultinom(n, 10, c(0.7, 0.1, 0.2))) rowID <- gl(n, 3, 3 * n) resp <- gl(3, 1, 3 * n) @ The logistic model for dependence on \Robject{x} can be fitted as a Poisson log-linear model\footnote{For this particular example, of course, it would be more economical to fit the model directly using \Rfunction{multinom} (from the recommended package \Rpackage{nnet}). But fitting as here via the `Poisson trick' allows the model to be elaborated within the \Rpackage{gnm} framework using \Rfunction{Mult} or other \Rclass{nonlin} terms.}, using either \Rfunction{glm} or \Rfunction{gnm}: \begin{Sinput} > ## Timings on a Xeon 2.33GHz, under Linux > system.time(temp.glm <- glm(counts ~ rowID + resp + resp:x, family = poisson))[1] \end{Sinput} \begin{Soutput} user.self 37.126 \end{Soutput} \begin{Sinput} > system.time(temp.gnm <- gnm(counts ~ resp + resp:x, eliminate = rowID, family = poisson, verbose = FALSE))[1] \end{Sinput} \begin{Soutput} user.self 0.04 \end{Soutput} \begin{Sinput} > c(deviance(temp.glm), deviance(temp.gnm)) \end{Sinput} \begin{Soutput} [1] 2462.556 2462.556 \end{Soutput} Here the use of \Rfunarg{eliminate} causes the \Rfunction{gnm} calculations to run much more quickly than \Rfunction{glm}. The speed advantage increases with the number of eliminated parameters (here 1000). By default,the eliminated parameters do not appear in printed model summaries as here: \begin{Sinput} > summary(temp.gnm) \end{Sinput} \begin{Soutput} Call: gnm(formula = counts ~ resp + resp:x, eliminate = rowID, family = poisson, verbose = FALSE) Deviance Residuals: Min 1Q Median 3Q Max -2.852038 -0.786172 -0.004534 0.645278 2.755013 Coefficients of interest: Estimate Std. Error z value Pr(>|z|) resp2 -1.961448 0.034007 -57.678 <2e-16 resp3 -1.255846 0.025359 -49.523 <2e-16 resp1:x -0.007726 0.024517 -0.315 0.753 resp2:x -0.023340 0.037611 -0.621 0.535 resp3:x 0.000000 NA NA NA (Dispersion parameter for poisson family taken to be 1) Std. Error is NA where coefficient has been constrained or is unidentified Residual deviance: 2462.6 on 1996 degrees of freedom AIC: 12028 Number of iterations: 4 \end{Soutput} although the \Rmethod{summary} method has a logical \Rfunarg{with.eliminate} that can toggled so that the eliminated parameters are included if desired. The \Rfunarg{eliminate} feature as implemented in \Rpackage{gnm} extends the earlier work of \cite{Hatz04} to a broader class of models and to over-parameterized model representations. \section{Methods and accessor functions} \label{sec:Methods} \subsection{Methods} \label{sec:specificMethods} The \Rfunction{gnm} function returns an object of class \Robject{c("gnm", "glm", "lm")}. There are several methods that have been written for objects of class \Rclass{glm} or \Rclass{lm} to facilitate inspection of fitted models. Out of the generic functions in the \Rpackage{base}, \Rpackage{stats} and \Rpackage{graphics} packages for which methods have been written for \Rclass{glm} or \Rclass{lm} objects, Figure \ref{fig:glm.lm} shows those that can be used to analyse \Rclass{gnm} objects, whilst Figure \ref{fig:!glm.lm} shows those that are not implemented for \Rclass{gnm} objects. \begin{figure}[!tbph] \centering \begin{fbox} { \begin{tabular*}{7.5cm}{@{\extracolsep{\fill}}lll@{\extracolsep{\fill}}} add1$^*$ & family & print \\ anova & formula & profile \\ case.names & hatvalues & residuals \\ coef & labels & rstandard \\ cooks.distance & logLik & summary \\ confint & model.frame & variable.names \\ deviance & model.matrix & vcov \\ drop1$^*$ & plot & weights \\ extractAIC & predict & \\ \end{tabular*} } \end{fbox} \caption{Generic functions in the \Rpackage{base}, \Rpackage{stats} and \Rpackage{graphics} packages that can be used to analyse \Rclass{gnm} objects. Starred functions are implemented for models with linear terms only.} \label{fig:glm.lm} \end{figure} \begin{figure}[!tbph] \centering \begin{fbox} { \begin{tabular*}{4.5cm}{@{\extracolsep{\fill}}ll@{\extracolsep{\fill}}} alias & effects \\ dfbeta & influence \\ dfbetas & kappa \\ dummy.coef & proj \\ \end{tabular*} } \end{fbox} \caption{Generic functions in the \Rpackage{base}, \Rpackage{stats} and \Rpackage{graphics} packages for which methods have been written for \Rclass{glm} or \Rclass{lm} objects, but which are \emph{not} implemented for \Rclass{gnm} objects.} \label{fig:!glm.lm} \end{figure} In addition to the accessor functions shown in Figure \ref{fig:glm.lm}, the \Rpackage{gnm} package provides a new generic function called \Rfunction{termPredictors} that has methods for objects of class \Rclass{gnm}, \Rclass{glm} and \Rclass{lm}. This function returns the additive contribution of each term to the predictor. See Section \ref{sec:termPredictors} for an example of its use. Most of the functions listed in Figure \ref{fig:glm.lm} can be used as they would be for \Rclass{glm} or \Rclass{lm} objects, however care must be taken with \Rmethod{vcov.gnm}, as the variance-covariance matrix will depend on the parameterization of the model. In particular, standard errors calculated using the variance-covariance matrix will only be valid for parameters or contrasts that are estimable! Similarly, \Rmethod{profile.gnm} and \Rmethod{confint.gnm} are only applicable to estimable parameters. The deviance function of a generalized nonlinear model can sometimes be far from quadratic and \Rmethod{profile.gnm} attempts to detect assymetry or asymptotic behaviour in order to return a sufficient profile for a given parameter. As an example, consider the following model, described later in Section \ref{sec:Unidiff}: \begin{Scode} unidiff <- gnm(Freq ~ educ*orig + educ*dest + Mult(Exp(educ), orig:dest), constrain = "[.]educ1", family = poisson, data = yaish, subset = (dest != 7)) prof <- profile(unidiff, which = 61:65, trace = TRUE) \end{Scode} If the deviance is quadratic in a given parameter, the profile trace will be linear. We can plot the profile traces as follows: \begin{figure}[!tbph] \begin{center} \scalebox{1.1}{\includegraphics{fig-profilePlot.pdf}} \end{center} \caption{Profile traces for the multipliers of the orig:dest association} \label{fig:profilePlot} \end{figure} From these plots we can see that the deviance is approximately quadratic in \Robject{Mult(Exp(.), orig:dest).educ2}, assymetric in \Robject{Mult(Exp(.), orig:dest).educ3} and \Robject{Mult(Exp(.), orig:dest).educ4} and asymptotic in \Robject{Mult(Exp(.), orig:dest).educ5}. When the deviance is approximately quadratic in a given parameter, \Rmethod{profile.gnm} uses the same stepsize for profiling above and below the original estimate: \begin{Sinput} > diff(prof[[2]]$par.vals[, "Mult(Exp(.), orig:dest).educ2"]) \end{Sinput} \begin{Soutput} [1] 0.1053072 0.1053072 0.1053072 0.1053072 0.1053072 0.1053072 0.1053072 [8] 0.1053072 0.1053072 0.1053072 \end{Soutput} When the deviance is asymmetric, \Rmethod{profile.gnm} uses different step sizes to accommodate the skew: \begin{Sinput} > diff(prof[[4]]$par.vals[, "Mult(Exp(.), orig:dest).educ4"]) \end{Sinput} \begin{Soutput} [1] 0.2018393 0.2018393 0.2018393 0.2018393 0.2018393 0.2018393 0.2018393 [8] 0.2018393 0.2018393 0.2243673 0.2243673 0.2243673 0.2243673 0.2243673 \end{Soutput} Finally, the presence of an asymptote is recorded in the \Robject{"asymptote"} attribute of the returned profile: \begin{Sinput} > attr(prof[[5]], "asymptote") \end{Sinput} \begin{Soutput} [1] TRUE FALSE \end{Soutput} This information is used by \Rmethod{confint.gnm} to return infinite limits for confidence intervals, as appropriate: \begin{Sinput} > confint(prof, level = 0.95) \end{Sinput} \begin{Soutput} 2.5 % 97.5 % Mult(Exp(.), orig:dest).educ1 NA NA Mult(Exp(.), orig:dest).educ2 -0.5978901 0.1022447 Mult(Exp(.), orig:dest).educ3 -1.4836854 -0.2362378 Mult(Exp(.), orig:dest).educ4 -2.5792398 -0.2953420 Mult(Exp(.), orig:dest).educ5 -Inf -0.7006889 \end{Soutput} \subsection{\Rfunction{ofInterest} and \Rfunction{pickCoef}} \label{sec:ofInterest} It is quite common for a statistical model to have a large number of parameters, but for only a subset of these parameters be of interest when it comes to interpreting the model. The \Rfunarg{ofInterest} argument to \Rfunction{gnm} allows the user to specify a subset of the parameters which are of interest, so that \Rclass{gnm} methods will focus on these parameters. In particular, printed model summaries will only show the parameters of interest, whilst methods for which a subset of parameters may be selected will by default select the parameters of interest, or where this may not be appropriate, provide a \emph{Tk} dialog for selection from the parameters of interest. Parameters may be specified to the \Rfunarg{ofInterest} argument by a regular expression to match against parameter names, by a numeric vector of indices, by a character vector of names, or, if \Rcode{ofInterest = "[?]"} they can be selected through a \emph{Tk} dialog. The information regarding the parameters of interest is held in the \Robject{ofInterest} component of \Rclass{gnm} objects, which is a named vector of numeric indices, or \Robject{NULL} if all parameters are of interest. This component may be accessed or replaced using \Rfunction{ofInterest} or \Rfunction{ofInterest<-} respectively. The \Rfunction{pickCoef} function provides a simple way to obtain the indices of coefficients from any model object. It takes the model object as its first argument and has an optional \Rfunarg{regexp} argument. If a regular expression is passed to \Rfunarg{regexp}, the coefficients are selected by matching this regular expression against the coefficient names. Otherwise, coefficients may be selected via a \emph{Tk} dialog. So, returning to the example from the last section, if we had set \Robject{ofInterest} to index the education multipliers as follows \begin{Scode} ofInterest(unidiff) <- pickCoef(unidiff, "[.]educ") \end{Scode} then it would not have been necessary to specify the \Rfunarg{which} argument of \Rfunction{profile} as these parameters would have been selected by default. \subsection{\Rfunction{checkEstimable}} \label{sec:checkEstimable} The \Rfunction{checkEstimable} function can be used to check the estimability of a linear combination of parameters. For non-linear combinations the same function can be used to check estimability based on the (local) vector of partial derivatives. The \Rfunction{checkEstimable} function provides a numerical version of the sort of algebraic test described in \citet{CatcMorg97}. Consider the following model, which is described later in Section \ref{sec:Unidiff}: <>= doubleUnidiff <- gnm(Freq ~ election:vote + election:class:religion + Mult(Exp(election), religion:vote) + Mult(Exp(election), class:vote), family = poisson, data = cautres) @ The effects of the first constituent multiplier in the first multiplicative interaction are identified when the parameter for one of the levels --- say for the first level --- is constrained to zero. The parameters to be estimated are then the differences between each other level and the first. These differences can be represented by a contrast matrix as follows: <>= coefs <- names(coef(doubleUnidiff)) contrCoefs <- coefs[grep(", religion:vote", coefs)] nContr <- length(contrCoefs) contrMatrix <- matrix(0, length(coefs), nContr, dimnames = list(coefs, contrCoefs)) contr <- contr.sum(contrCoefs) # switch round to contrast with first level contr <- rbind(contr[nContr, ], contr[-nContr, ]) contrMatrix[contrCoefs, 2:nContr] <- contr contrMatrix[contrCoefs, 2:nContr] @ Then their estimability can be checked using \Rfunction{checkEstimable} <>= checkEstimable(doubleUnidiff, contrMatrix) @ which confirms that the effects for the other three levels are estimable when the parameter for the first level is set to zero. However, applying the equivalent constraint to the second constituent multiplier in the interaction is not sufficient to make the parameters in that multiplier estimable: <>= coefs <- names(coef(doubleUnidiff)) contrCoefs <- coefs[grep("[.]religion", coefs)] nContr <- length(contrCoefs) contrMatrix <- matrix(0, length(coefs), length(contrCoefs), dimnames = list(coefs, contrCoefs)) contr <- contr.sum(contrCoefs) contrMatrix[contrCoefs, 2:nContr] <- rbind(contr[nContr, ], contr[-nContr, ]) checkEstimable(doubleUnidiff, contrMatrix) @ \subsection{\Rfunction{getContrasts}, \Rfunction{se}} \label{sec:getContrasts} To investigate simple ``sum to zero'' contrasts such as those above, it is easiest to use the \Rfunction{getContrasts} function, which checks the estimability of possibly scaled contrasts and returns the parameter estimates with their standard errors. Returning to the example of the first constituent multiplier in the first multiplicative interaction term, the differences between each election and the first can be obtained as follows: <>= myContrasts <- getContrasts(doubleUnidiff, pickCoef(doubleUnidiff, ", religion:vote")) myContrasts @ %def Visualization of estimated contrasts using `quasi standard errors' \citep{Firt03,FirtMene04} is achieved by plotting the resulting object: <>= plot(myContrasts, main = "Relative strength of religion-vote association, log scale", xlab = "Election", levelNames = 1:4) @ \begin{figure}[!tbph] \begin{center} \includegraphics{gnmOverview-qvplot.pdf} \end{center} \caption{Relative strength of religion-vote association, log scale} \label{fig:qvplot} \end{figure} %Attempting to obtain the equivalent contrasts for the second %(religion-vote association) multiplier produces the %following result: %<>= %coefs.of.interest <- grep("[.]religion", names(coef(doubleUnidiff))) %getContrasts(doubleUnidiff, coefs.of.interest) %@ %def By default, \Rfunction{getContrasts} uses the first parameter of the specified set as the reference level; alternatives may be set via the \Rfunarg{ref} argument. In the above example, the simple contrasts are estimable without scaling. In certain other applications, for example row-column association models (see Section~\ref{sec:RCmodels}), the contrasts are identified only after fixing their scale. A more general family of \emph{scaled} contrasts for a set of parameters $\gamma_r, r = 1, \ldots, R$ is given by \begin{equation*} \gamma^*_r = \frac{\gamma_r - \overline{\gamma}_w}{ \sqrt{\sum_r v_r (\gamma_r - \overline{\gamma}_u)^2}} \end{equation*} where $\overline{\gamma}_w = \sum w_r \gamma_r$ is the reference level against which the contrasts are taken, $\overline{\gamma}_u = \sum u_r \gamma_r$ is a possibly different weighted mean of the parameters to be used as reference level for a set of ``scaling contrasts'', and $v_r$ is a further set of weights. Thus, for example, the choice \[ w_r= \begin{cases} 1&(r=1)\\ 0&\hbox{(otherwise)} \end{cases}, \qquad u_r=v_r=1/R \] specifies contrasts with the first level, with the coefficients scaled to have variance 1\null. This general type of scaling can be obtained by specifying the form of $\overline{\gamma}_u$ and $v_r$ via the \Rfunarg{scaleRef} and \Rfunarg{scaleWeights} arguments of \Rfunction{getContrasts}. As an example, consider the following model, described in Section~\ref{sec:RCmodels}: @ <>= mentalHealth$MHS <- C(mentalHealth$MHS, treatment) mentalHealth$SES <- C(mentalHealth$SES, treatment) RC1model <- gnm(count ~ SES + MHS + Mult(SES, MHS), family = poisson, data = mentalHealth) @ %def The effects of the constituent multipliers of the multiplicative interaction are identified when both their scale and location are constrained. A simple way to achieve this is to set the first parameter to zero and the last parameter to one: @ <>= RC1model2 <- gnm(count ~ SES + MHS + Mult(1, SES, MHS), constrain = "[.]SES[AF]", constrainTo = c(0, 1), ofInterest = "[.]SES", family = poisson, data = mentalHealth) summary(RC1model2) @ %def Note that a constant multiplier must be incorporated into the interaction term, i.e., the multiplicative term \Rcode{Mult(SES, MHS)} becomes \Rcode{Mult(1, SES, MHS)}, in order to maintain equivalence with the original model specification. The constraints specified for \Robject{RC1model2} result in the estimation of scaled contrasts with level \Rcode{A} of \Rcode{SES}, in which the scaling fixes the magnitude of the contrast between level \Rcode{F} and level \Rcode{A} to be equal to 1\null. The equivalent use of \Rfunction{getContrasts}, together with the \emph{unconstrained} fit (\Robject{RC1model}), in this case is as follows: @ <>= getContrasts(RC1model, pickCoef(RC1model, "[.]SES"), ref = "first", scaleRef = "first", scaleWeights = c(rep(0, 5), 1)) @ %def Quasi-variances and standard errors are not returned here as they can not (currently) be computed for scaled contrasts. When the scaling uses the same reference level as the contrasts, equal scale weights produce ``spherical'' contrasts, whilst unequal weights produce ``elliptical'' contrasts. Further examples are given in Sections~\ref{sec:RCmodels} and \ref{sec:GAMMI}. For more general linear combinations of parameters than contrasts, the lower-level \Rfunction{se} function (which is called internally by \Rfunction{getContrasts} and by the \Rmethod{summary} method) can be used directly. See \Rcode{help(se)} for details. \subsection{\Rfunction{residSVD}} \label{sec:residSVD} Sometimes it is useful to operate on the residuals of a model in order to create informative summaries of residual variation, or to obtain good starting values for additional parameters in a more elaborate model. The relevant arithmetical operations are weighted means of the so-called \emph{working residuals}. The \Rfunction{residSVD} function facilitates one particular residual analysis that is often useful when considering multiplicative interaction between factors as a model elaboration: in effect, \Rfunction{residSVD} provides a direct estimate of the parameters of such an interaction, by performing an appropriately weighted singular value decomposition on the working residuals. As an illustration, consider the barley data from \citet{Wedd74}. These data have the following two-way structure: <>= xtabs(y ~ site + variety, barley) @ In Section~\ref{sec:biplot} a biplot model is proposed for these data, which comprises a two-component interaction between the cross-classifying factors. In order to fit this model, we can proceed by fitting a smaller model, then use \Rfunction{residSVD} to obtain starting values for the parameters in the bilinear term: @ <>= emptyModel <- gnm(y ~ -1, family = wedderburn, data = barley) biplotStart <- residSVD(emptyModel, barley$site, barley$variety, d = 2) biplotModel <- gnm(y ~ -1 + instances(Mult(site, variety), 2), family = wedderburn, data = barley, start = biplotStart) @ %def In this instance, the use of purposive (as opposed to the default, random) starting values had little effect: the fairly large number of iterations needed in this example is caused by a rather flat (quasi-)likelihood surface near the maximum, not by poor starting values. In other situations, the use of \Rfunction{residSVD} may speed the calculations dramatically (see for example Section \ref{sec:GAMMI}), or it may be crucial to success in locating the MLE (for example see \Rcode{help(House2001)}, where the number of multiplicative parameters is in the hundreds). The \Rfunction{residSVD} result in this instance provides a crude approximation to the MLE of the enlarged model, as can be seen in Figure \ref{fig:residSVDplot}: @ <>= plot(coef(biplotModel), biplotStart, main = "Comparison of residSVD and MLE for a 2-dimensional biplot model", ylim = c(-2, 2), xlim = c(-4, 4)) abline(a = 0, b = 1, lty = 2) @ %def \begin{figure}[!tbph] \begin{center} \includegraphics{gnmOverview-residSVDplot} \end{center} \caption{Comparison of residSVD and the MLE for a 2-dimensional biplot model} \label{fig:residSVDplot} \end{figure} \section{\Rfunction{gnm} or \Rfunction{(g)nls}?} \label{sec:gnmVnls} The \Rfunction{nls} function in the \Rpackage{stats} package may be used to fit a nonlinear model via least-squares estimation. Statistically speaking, \Rfunction{gnm} is to \Rfunction{nls} as \Rfunction{glm} is to \Rfunction{lm}, in that a nonlinear least-squares model is equivalent to a generalized nonlinear model with \Rcode{family = gaussian}. A \Rfunction{nls} model assumes that the responses are distributed either with constant variance or with fixed relative variances (specified via the \Rfunarg{weights} argument). The \Rfunction{gnls} function in the \Rpackage{nlme} package extends \Rfunction{nls} to allow correlated responses. On the other hand, \Rfunction{gnm} allows for responses distributed with variances that are a specified (via the \Rfunarg{family} argument) function of the mean; as with \Rfunction{nls}, no correlation is allowed. The \Rfunction{gnm} function also differs from \Rfunction{nls}/\Rfunction{gnls} in terms of the interface. Models are specified to \Rfunction{nls} and \Rfunction{gnls} in terms of a mathematical formula or a \Rclass{selfStart} function based on such a formula, which is convenient for models that have a small number of parameters. For models that have a large number of parameters, or can not easily be represented by a mathematical formula, the symbolic model specification used by \Rfunction{gnm} may be more convenient. This would usually be the case for models involving factors, which would need to be represented by dummy variables in a \Rfunction{nls} formula. When working with artificial data, \Rfunction{gnm} has the minor advantage that it does not fail when a model is an exact fit to the data (see \Rcode{help(nls)})\null. Therefore it is not necessary with \Rfunction{gnm} to add noise to artificial data, which can be useful when testing methods. \section{Examples} \label{sec:Examples} \subsection{Row-column association models} \label{sec:RCmodels} There are several models that have been proposed for modelling the relationship between the cell means of a contingency table and the cross-classifying factors. The following examples consider the row-column association models proposed by \citet{Good79}. The examples shown use data from two-way contingency tables, but the \Rpackage{gnm} package can also be used to fit the equivalent models for higher order tables. \subsubsection{RC(1) model} The RC(1) model is a row and column association model with the interaction between row and column factors represented by one component of the multiplicative interaction. If the rows are indexed by $r$ and the columns by $c$, then the log-multiplicative form of the RC(1) model for the cell means $\mu_{rc}$ is given by \[\log \mu_{rc} = \alpha_r + \beta_c + \gamma_r\delta_c. \] We shall fit this model to the \Robject{mentalHealth} data set from \citet[][page 381]{Agre02}, which is a two-way contingency table classified by the child's mental impairment (MHS) and the parents' socioeconomic status (SES). Although both of these factors are ordered, we do not wish to use polynomial contrasts in the model, so we begin by setting the contrasts attribute of these factors to \Rcode{treatment}: <>= set.seed(1) mentalHealth$MHS <- C(mentalHealth$MHS, treatment) mentalHealth$SES <- C(mentalHealth$SES, treatment) @ The \Rclass{gnm} model is then specified as follows, using the poisson family with a log link function: <>= RC1model <- gnm(count ~ SES + MHS + Mult(SES, MHS), family = poisson, data = mentalHealth) RC1model @ %def The row scores (parameters 10 to 15) and the column scores (parameters 16 to 19) of the multiplicative interaction can be normalized as in Agresti's eqn (9.15): <>= rowProbs <- with(mentalHealth, tapply(count, SES, sum) / sum(count)) colProbs <- with(mentalHealth, tapply(count, MHS, sum) / sum(count)) rowScores <- coef(RC1model)[10:15] colScores <- coef(RC1model)[16:19] rowScores <- rowScores - sum(rowScores * rowProbs) colScores <- colScores - sum(colScores * colProbs) beta1 <- sqrt(sum(rowScores^2 * rowProbs)) beta2 <- sqrt(sum(colScores^2 * colProbs)) assoc <- list(beta = beta1 * beta2, mu = rowScores / beta1, nu = colScores / beta2) assoc @ %def Alternatively, the elliptical contrasts \Robject{mu} and \Robject{nu} can be obtained using \Rfunction{getContrasts}, with the advantage that the standard errors for the contrasts will also be computed: @ <>= mu <- getContrasts(RC1model, pickCoef(RC1model, "[.]SES"), ref = rowProbs, scaleWeights = rowProbs) nu <- getContrasts(RC1model, pickCoef(RC1model, "[.]MHS"), ref = colProbs, scaleWeights = colProbs) mu nu @ %def Since the value of \Robject{beta} is dependent upon the particular scaling used for the contrasts, it is typically not of interest to conduct inference on this parameter directly. The standard error for \Robject{beta} could be obtained, if desired, via the delta method. \subsubsection{RC(2) model} The RC(1) model can be extended to an RC($m$) model with $m$ components of the multiplicative interaction. For example, the RC(2) model is given by \[ \log \mu_{rc} = \alpha_r + \beta_c + \gamma_r\delta_c + \theta_r\phi_c. \] Extra instances of the multiplicative interaction can be specified by the \Rfunarg{multiplicity} argument of \Rfunction{Mult}, so the RC(2) model can be fitted to the \Robject{mentalHealth} data as follows <>= RC2model <- gnm(count ~ SES + MHS + instances(Mult(SES, MHS), 2), family = poisson, data = mentalHealth) RC2model @ \subsubsection{Homogeneous effects} If the row and column factors have the same levels, or perhaps some levels in common, then the row-column interaction could be modelled by a multiplicative interaction with homogeneous effects, that is \[\log \mu_{rc} = \alpha_r + \beta_c + \gamma_r\gamma_c.\] For example, the \Robject{occupationalStatus} data set from \citet{Good79} is a contingency table classified by the occupational status of fathers (origin) and their sons (destination). \citet{Good79} fits a row-column association model with homogeneous effects to these data after deleting the cells on the main diagonal. Equivalently we can account for the diagonal effects by a separate \Rfunction{Diag} term: @ <>= RChomog <- gnm(Freq ~ origin + destination + Diag(origin, destination) + MultHomog(origin, destination), family = poisson, data = occupationalStatus) RChomog @ %def To determine whether it would be better to allow for heterogeneous effects on the association of the fathers' occupational status and the sons' occupational status, we can compare this model to the RC(1) model for these data: <>= RCheterog <- gnm(Freq ~ origin + destination + Diag(origin, destination) + Mult(origin, destination), family = poisson, data = occupationalStatus) anova(RChomog, RCheterog) @ In this case there is little gain in allowing heterogeneous effects. \subsection{Diagonal reference models} \label{sec:Dref} Diagonal reference models, proposed by \citet{Sobe81, Sobe85}, are designed for contingency tables classified by factors with the same levels. The cell means are modelled as a function of the diagonal effects, i.e., the mean responses of the `diagonal' cells in which the levels of the row and column factors are the same. \subsubsection*{\Rfunction{Dref} example 1: Political consequences of social mobility} To illustrate the use of diagonal reference models we shall use the \Robject{voting} data from \citet{Clif93}. The data come from the 1987 British general election and are the percentage voting Labour in groups cross-classified by the class of the head of household (\Robject{destination}) and the class of their father (\Robject{origin}). In order to weight these percentages by the group size, we first back-transform them to the counts of those voting Labour and those not voting Labour: @ <>= set.seed(1) count <- with(voting, percentage/100 * total) yvar <- cbind(count, voting$total - count) @ %def The grouped percentages may be modelled by a basic diagonal reference model, that is, a weighted sum of the diagonal effects for the corresponding origin and destination classes. This model may be expressed as \[ \mu_{od} = \frac{e^{\delta_1}}{e^{\delta_1} + e^{\delta_2}}\gamma_o + \frac{e^{\delta_2}}{e^{\delta_1} + e^{\delta_2}}\gamma_d . \] See Section \ref{sec:Dref function} for more detail on the parameterization. The basic diagonal reference model may be fitted using \Rfunction{gnm} as follows @ <>= classMobility <- gnm(yvar ~ Dref(origin, destination), family = binomial, data = voting) classMobility @ %def and the origin and destination weights can be evaluated as below @ <>= DrefWeights(classMobility) @ %def These results are slightly different from those reported by \citet{Clif93}. The reason for this is unclear: we are confident that the above results are correct for the data as given in \citet{Clif93}, but have not been able to confirm that the data as printed in the journal were exactly as used in Clifford and Heath's analysis. \citet{Clif93} suggest that movements in and out of the salariat (class 1) should be treated differently from movements between the lower classes (classes 2 - 5), since the former has a greater effect on social status. Thus they propose the following model \begin{equation*} \mu_{od} = \begin{cases} \dfrac{e^{\delta_1}}{e^{\delta_1} + e^{\delta_2}}\gamma_o + \dfrac{e^{\delta_2}}{e^{\delta_1} + e^{\delta_2}}\gamma_d & \text{if } o = 1\\ \\ \dfrac{e^{\delta_3}}{e^{\delta_3} + e^{\delta_4}}\gamma_o + \dfrac{e^{\delta_4}}{e^{\delta_3} + e^{\delta_4}}\gamma_d & \text{if } d = 1\\ \\ \dfrac{e^{\delta_5}}{e^{\delta_5} + e^{\delta_6}}\gamma_o + \dfrac{e^{\delta_6}}{e^{\delta_5} + e^{\delta_6}}\gamma_d & \text{if } o \ne 1 \text{ and } d \ne 1 \end{cases} \end{equation*} To fit this model we define factors indicating movement in (upward) and out (downward) of the salariat @ <>= upward <- with(voting, origin != 1 & destination == 1) downward <- with(voting, origin == 1 & destination != 1) @ %def Then the diagonal reference model with separate weights for socially mobile groups can be estimated as follows @ <>= socialMobility <- gnm(yvar ~ Dref(origin, destination, delta = ~ 1 + downward + upward), family = binomial, data = voting) socialMobility @ %def The weights for those moving into the salariat, those moving out of the salariat and those in any other group, can be evaluated as below @ <>= DrefWeights(socialMobility) @ %def Again, the results differ slightly from those reported by \citet{Clif93}, but the essence of the results is the same: the origin weight is much larger for the downwardly mobile group than for the other groups. The weights for the upwardly mobile group are very similar to the base level weights, so the model may be simplified by only fitting separate weights for the downwardly mobile group: @ <>= downwardMobility <- gnm(yvar ~ Dref(origin, destination, delta = ~ 1 + downward), family = binomial, data = voting) downwardMobility DrefWeights(downwardMobility) @ %def \subsubsection*{\Rfunction{Dref} example 2: conformity to parental rules} %\SweaveInput{vanDerSlikEg.Rnw} Another application of diagonal reference models is given by \citet{Vand02}. The data from this paper are not publicly available\footnote{ We thank Frans van der Slik for his kindness in sending us the data.}, but we shall show how the models presented in the paper may be estimated using \Rfunction{gnm}. The data relate to the value parents place on their children conforming to their rules. There are two response variables: the mother's conformity score (MCFM) and the father's conformity score (FCFF). The data are cross-classified by two factors describing the education level of the mother (MOPLM) and the father (FOPLF), and there are six further covariates (AGEM, MRMM, FRMF, MWORK, MFCM and FFCF). In their baseline model for the mother's conformity score, \citet{Vand02} include five of the six covariates (leaving out the father's family conflict score, FCFF) and a diagonal reference term with constant weights based on the two education factors. This model may be expressed as \[ \mu_{rci} = \beta_1x_{1i} + \beta_2x_{2i} + \beta_3x_{3i} +\beta_4x_{4i} +\beta_5x_{5i} + \frac{e^{\delta_1}}{e^{\delta_1} + e^{\delta_2}}\gamma_r + \frac{e^{\delta_2}}{e^{\delta_1} + e^{\delta_2}}\gamma_c . \] The baseline model can be fitted as follows: \begin{Sinput} > set.seed(1) > A <- gnm(MCFM ~ -1 + AGEM + MRMM + FRMF + MWORK + MFCM + + Dref(MOPLM, FOPLF), family = gaussian, data = conformity, + verbose = FALSE) > A \end{Sinput} \begin{Soutput} Call: gnm(formula = MCFM ~ -1 + AGEM + MRMM + FRMF + MWORK + MFCM + Dref(MOPLM, FOPLF), family = gaussian, data = conformity, verbose = FALSE) Coefficients: AGEM MRMM FRMF 0.06363 -0.32425 -0.25324 MWORK MFCM Dref(MOPLM, FOPLF)delta1 -0.06430 -0.06043 -0.33731 Dref(MOPLM, FOPLF)delta2 Dref(., .).MOPLM|FOPLF1 Dref(., .).MOPLM|FOPLF2 -0.02505 4.95121 4.86329 Dref(., .).MOPLM|FOPLF3 Dref(., .).MOPLM|FOPLF4 Dref(., .).MOPLM|FOPLF5 4.86458 4.72343 4.43516 Dref(., .).MOPLM|FOPLF6 Dref(., .).MOPLM|FOPLF7 4.18873 4.43378 Deviance: 425.3389 Pearson chi-squared: 425.3389 Residual df: 576 \end{Soutput} The coefficients of the covariates are not aliased with the parameters of the diagonal reference term and thus the basic identifiability constraints that have been imposed are sufficient for these parameters to be identified. The diagonal effects do not need to be constrained as they represent contrasts with the off-diagonal cells. Therefore the only unidentified parameters in this model are the weight parameters. This is confirmed in the summary of the model: \begin{Sinput} > summary(A) \end{Sinput} \begin{Soutput} Call: gnm(formula = MCFM ~ -1 + AGEM + MRMM + FRMF + MWORK + MFCM + Dref(MOPLM, FOPLF), family = gaussian, data = conformity, verbose = FALSE) Deviance Residuals: Min 1Q Median 3Q Max -3.63688 -0.50383 0.01714 0.56753 2.25139 Coefficients: Estimate Std. Error t value Pr(>|t|) AGEM 0.06363 0.07375 0.863 0.38859 MRMM -0.32425 0.07766 -4.175 3.44e-05 FRMF -0.25324 0.07681 -3.297 0.00104 MWORK -0.06430 0.07431 -0.865 0.38727 MFCM -0.06043 0.07123 -0.848 0.39663 Dref(MOPLM, FOPLF)delta1 -0.33731 NA NA NA Dref(MOPLM, FOPLF)delta2 -0.02505 NA NA NA Dref(., .).MOPLM|FOPLF1 4.95121 0.16639 29.757 < 2e-16 Dref(., .).MOPLM|FOPLF2 4.86329 0.10436 46.602 < 2e-16 Dref(., .).MOPLM|FOPLF3 4.86458 0.12855 37.842 < 2e-16 Dref(., .).MOPLM|FOPLF4 4.72343 0.13523 34.929 < 2e-16 Dref(., .).MOPLM|FOPLF5 4.43516 0.19314 22.963 < 2e-16 Dref(., .).MOPLM|FOPLF6 4.18873 0.17142 24.435 < 2e-16 Dref(., .).MOPLM|FOPLF7 4.43378 0.16903 26.231 < 2e-16 --- (Dispersion parameter for gaussian family taken to be 0.7384355) Std. Error is NA where coefficient has been constrained or is unidentified Residual deviance: 425.34 on 576 degrees of freedom AIC: 1507.8 Number of iterations: 15 \end{Soutput} The weights have been constrained to sum to one as described in Section \ref{sec:Dref function}, so the weights themselves may be estimated as follows: \begin{Sinput} > prop.table(exp(coef(A)[6:7])) \end{Sinput} \begin{Soutput} Dref(MOPLM, FOPLF)delta1 Dref(MOPLM, FOPLF)delta2 0.4225638 0.5774362 \end{Soutput} However, in order to estimate corresponding standard errors, the parameters of one of the weights must be constrained. If no such constraints were applied when the model was fitted, \Rfunction{DrefWeights} will refit the model constraining the parameters of the first weight to zero: \begin{Sinput} > DrefWeights(A) \end{Sinput} \begin{Soutput} Refitting with parameters of first Dref weight constrained to zero $MOPLM weight se 0.4225636 0.1439829 $FOPLF weight se 0.5774364 0.1439829 \end{Soutput} giving the values reported by \citet{Vand02}. All the other coefficients of model A are the same as those reported by \citet{Vand02} except the coefficients of the mother's gender role (MRMM) and the father's gender role (FRMF). \citet{Vand02} reversed the signs of the coefficients of these factors since they were coded in the direction of liberal values, unlike the other covariates. However, simply reversing the signs of these coefficients does not give the same model, since the estimates of the diagonal effects depend on the estimates of these coefficients. For consistent interpretation of the covariate coefficients, it is better to recode the gender role factors as follows: \begin{Sinput} > MRMM2 <- as.numeric(!conformity$MRMM) > FRMF2 <- as.numeric(!conformity$FRMF) > A <- gnm(MCFM ~ -1 + AGEM + MRMM2 + FRMF2 + MWORK + MFCM + + Dref(MOPLM, FOPLF), family = gaussian, data = conformity, + verbose = FALSE) > A \end{Sinput} \begin{Soutput} Call: gnm(formula = MCFM ~ -1 + AGEM + MRMM2 + FRMF2 + MWORK + MFCM + Dref(MOPLM, FOPLF), family = gaussian, data = conformity, verbose = FALSE) Coefficients: AGEM MRMM2 FRMF2 0.06363 0.32425 0.25324 MWORK MFCM Dref(MOPLM, FOPLF)delta1 -0.06430 -0.06043 0.08440 Dref(MOPLM, FOPLF)delta2 Dref(., .).MOPLM|FOPLF1 Dref(., .).MOPLM|FOPLF2 0.39666 4.37371 4.28579 Dref(., .).MOPLM|FOPLF3 Dref(., .).MOPLM|FOPLF4 Dref(., .).MOPLM|FOPLF5 4.28708 4.14593 3.85767 Dref(., .).MOPLM|FOPLF6 Dref(., .).MOPLM|FOPLF7 3.61123 3.85629 Deviance: 425.3389 Pearson chi-squared: 425.3389 Residual df: 576 \end{Soutput} The coefficients of the covariates are now as reported by \citet{Vand02}, but the diagonal effects have been adjusted appropriately. \citet{Vand02} compare the baseline model for the mother's conformity score to several other models in which the weights in the diagonal reference term are dependent on one of the covariates. One particular model they consider incorporates an interaction of the weights with the mother's conflict score as follows: \[ \mu_{rci} = \beta_1x_{1i} + \beta_2x_{2i} + \beta_3x_{3i} +\beta_4x_{4i} +\beta_5x_{5i} + \frac{e^{\xi_{01} + \xi_{11}x_{5i}}}{e^{\xi_{01} + \xi_{11}x_{5i}} + e^{\xi_{02} + \xi_{12}x_{5i}}}\gamma_r + \frac{e^{\xi_{02} + \xi_{12}x_{5i}}}{e^{\xi_{01} + \xi_{11}x_{5i}} + e^{\xi_{02} + \xi_{12}x_{5i}}}\gamma_c. \] This model can be fitted as below, using the original coding for the gender role factors for ease of comparison to the results reported by \citet{Vand02}, \begin{Sinput} > F <- gnm(MCFM ~ -1 + AGEM + MRMM + FRMF + MWORK + MFCM + + Dref(MOPLM, FOPLF, delta = ~ 1 + MFCM), family = gaussian, + data = conformity, verbose = FALSE) > F \end{Sinput} \begin{Soutput} Call: gnm(formula = MCFM ~ -1 + AGEM + MRMM + FRMF + MWORK + MFCM + Dref(MOPLM, FOPLF, delta = ~1 + MFCM), family = gaussian, data = conformity, verbose = FALSE) Coefficients: AGEM 0.05818 MRMM -0.32701 FRMF -0.25772 MWORK -0.07847 MFCM -0.01694 Dref(MOPLM, FOPLF, delta = ~ . + MFCM).delta1(Intercept) 1.03515 Dref(MOPLM, FOPLF, delta = ~ 1 + .).delta1MFCM -1.77756 Dref(MOPLM, FOPLF, delta = ~ . + MFCM).delta2(Intercept) -0.03515 Dref(MOPLM, FOPLF, delta = ~ 1 + .).delta2MFCM 2.77756 Dref(., ., delta = ~ 1 + MFCM).MOPLM|FOPLF1 4.82476 Dref(., ., delta = ~ 1 + MFCM).MOPLM|FOPLF2 4.88066 Dref(., ., delta = ~ 1 + MFCM).MOPLM|FOPLF3 4.83969 Dref(., ., delta = ~ 1 + MFCM).MOPLM|FOPLF4 4.74850 Dref(., ., delta = ~ 1 + MFCM).MOPLM|FOPLF5 4.42020 Dref(., ., delta = ~ 1 + MFCM).MOPLM|FOPLF6 4.17957 Dref(., ., delta = ~ 1 + MFCM).MOPLM|FOPLF7 4.40819 Deviance: 420.9022 Pearson chi-squared: 420.9022 Residual df: 575 \end{Soutput} In this case there are two sets of weights, one for when the mother's conflict score is less than average (coded as zero) and one for when the score is greater than average (coded as one). These can be evaluated as follows: \begin{Sinput} > DrefWeights(F) \end{Sinput} \begin{Soutput} Refitting with parameters of first Dref weight constrained to zero $MOPLM MFCM weight se 1 1 0.02974675 0.2277711 2 0 0.74465224 0.2006916 $FOPLF MFCM weight se 1 1 0.9702532 0.2277711 2 0 0.2553478 0.2006916 \end{Soutput} giving the same weights as in Table 4 of \citet{Vand02}, though we obtain a lower standard error in the case where MFCM is equal to one. \subsection{Uniform difference (UNIDIFF) models} \label{sec:Unidiff} Uniform difference models \citep{Xie92, Erik92} use a simplified three-way interaction to provide an interpretable model of contingency tables classified by three or more variables. For example, the uniform difference model for a three-way contingency table, also known as the UNIDIFF model, is given by \[ \mu_{ijk} = \alpha_{ik} + \beta_{jk} + \exp(\delta_k)\gamma_{ij}. \] The $\gamma_{ij}$ represent a pattern of association that varies in strength over the dimension indexed by $k$, and $\exp(\delta_k)$ represents the relative strength of that association at level $k$. This model can be applied to the \Robject{yaish} data set \citep{Yais98,Yais04}, which is a contingency table cross-classified by father's social class (\Robject{orig}), son's social class (\Robject{dest}) and son's education level (\Robject{educ}). In this case, we can consider the importance of the association between the social class of father and son across the education levels. We omit the sub-table which corresponds to level 7 of \Robject{dest}, because its information content is negligible: @ <>= set.seed(1) unidiff <- gnm(Freq ~ educ*orig + educ*dest + Mult(Exp(educ), orig:dest), ofInterest = "[.]educ", family = poisson, data = yaish, subset = (dest != 7)) coef(unidiff) @ %def The \Robject{ofInterest} component has been set to index the multipliers of the association between the social class of father and son. We can contrast each multiplier to that of the lowest education level and obtain the standard errors for these parameters as follows: @ <>= getContrasts(unidiff, ofInterest(unidiff)) @ %def Four-way contingency tables may sometimes be described by a ``double UNIDIFF'' model \[ \mu_{ijkl} = \alpha_{il} + \beta_{jkl} + \exp(\delta_l)\gamma_{ij} + \exp(\phi_l)\theta_{ik}, \] where the strengths of two, two-way associations with a common variable are estimated across the levels of the fourth variable. The \Robject{cautres} data set, from \citet{Caut98}, can be used to illustrate the application of the double UNIDIFF model. This data set is classified by the variables vote, class, religion and election. Using a double UNIDIFF model, we can see how the association between class and vote, and the association between religion and vote, differ between the most recent election and the other elections: @ <>= set.seed(1) doubleUnidiff <- gnm(Freq ~ election*vote + election*class*religion + Mult(Exp(election), religion:vote) + Mult(Exp(election), class:vote), family = poisson, data = cautres) getContrasts(doubleUnidiff, rev(pickCoef(doubleUnidiff, ", class:vote"))) getContrasts(doubleUnidiff, rev(pickCoef(doubleUnidiff, ", religion:vote"))) @ %def \subsection{Generalized additive main effects and multiplicative interaction (GAMMI) models} \label{sec:GAMMI} Generalized additive main effects and multiplicative interaction models, or GAMMI models, were motivated by two-way contingency tables and comprise the row and column main effects plus one or more components of the multiplicative interaction. The singular value corresponding to each multiplicative component is often factored out, as a measure of the strength of association between the row and column scores, indicating the importance of the component, or axis. For cell means $\mu_{rc}$ a GAMMI-K model has the form \begin{equation} \label{eq:GAMMI} g(\mu_{rc}) = \alpha_r + \beta_c + \sum_{k=1}^K \sigma_k\gamma_{kr}\delta_{kc}, \end{equation} in which $g$ is a link function, $\alpha_r$ and $\beta_c$ are the row and column main effects, $\gamma_{kr}$ and $\delta_{kc}$ are the row and column scores for multiplicative component $k$ and $\sigma_k$ is the singular value for component $k$. The number of multiplicative components, $K$, is less than or equal to the rank of the matrix of residuals from the main effects. The row-column association models discussed in Section \ref{sec:RCmodels} are examples of GAMMI models, with a log link and poisson variance. Here we illustrate the use of an AMMI model, which is a GAMMI model with an identity link and a constant variance. We shall use the \Robject{wheat} data set taken from \citet{Varg01}, which gives wheat yields measured over ten years. First we scale these yields and create a new treatment factor, so that we can reproduce the analysis of \citet{Varg01}: @ <>= set.seed(1) yield.scaled <- wheat$yield * sqrt(3/1000) treatment <- interaction(wheat$tillage, wheat$summerCrop, wheat$manure, wheat$N, sep = "") @ %def Now we can fit the AMMI-1 model, to the scaled yields using the combined treatment factor and the year factor from the \Robject{wheat} dataset. We will proceed by first fitting the main effects model, then using \Rfunction{residSVD} (see Section \ref{sec:residSVD}) for the parameters of the multiplicative term: @ <>= mainEffects <- gnm(yield.scaled ~ year + treatment, family = gaussian, data = wheat) svdStart <- residSVD(mainEffects, year, treatment, 3) bilinear1 <- update(mainEffects, . ~ . + Mult(year, treatment), start = c(coef(mainEffects), svdStart[,1])) @ %def We can compare the AMMI-1 model to the main effects model, @ <>= anova(mainEffects, bilinear1, test = "F") @ %def giving the same results as in Table 1 of \citet{Varg01} (up to error caused by rounding). Thus the significance of the multiplicative interaction can be tested without applying constraints to this term. If the multiplicative interaction is significant, we may wish to apply constraints to obtain estimates of the row and column scores. We illustrate this using the \Robject{barleyHeights} data, which records the average height for 15 genotypes of barley over 9 years. For this small dataset the AMMI-1 model is easily estimated with the default settings: @ <>= set.seed(1) barleyModel <- gnm(height ~ year + genotype + Mult(year, genotype), data = barleyHeights) @ %def To obtain the parameterization of Equation \ref{eq:GAMMI} in which $\sigma_k$ is the singular value for component $k$, the row and column scores must be constrained so that the scores sum to zero and the squared scores sum to one. These contrasts can be obtained using \Robject{getContrasts}: @ <>= gamma <- getContrasts(barleyModel, pickCoef(barleyModel, "[.]y"), ref = "mean", scaleWeights = "unit") delta <- getContrasts(barleyModel, pickCoef(barleyModel, "[.]g"), ref = "mean", scaleWeights = "unit") gamma delta @ %def Confidence intervals based on the assumption of asymptotic normality can be computed as follows: @ <>= gamma[[2]][,1] + (gamma[[2]][,2]) %o% c(-1.96, 1.96) delta[[2]][,1] + (delta[[2]][,2]) %o% c(-1.96, 1.96) @ %def which broadly agree with Table 8 of Chadouef and Denis (1991), allowing for the change in sign. On the basis of such confidence intervals we can investigate simplifications of the model such as combining levels of the factors or fitting an additive model to a subset of the data. The singular value $\sigma_k$ may be obtained as follows @ <>= svd(termPredictors(barleyModel)[, "Mult(year, genotype)"])$d @ %def This parameter is of little interest in itself, given that the significance of the term as a whole can be tested using ANOVA. The SVD representation can also be obtained quite easily for AMMI and GAMMI models with interaction rank greater than 1\null. See \Rcode{example(wheat)} for an example of this in an AMMI model with rank 2\null. (The calculation of \emph{standard errors} and \emph{confidence regions} for the SVD representation with rank greater than 1 is not yet implemented, though.) \subsection{Biplot models} \label{sec:biplot} Biplots are graphical displays of two-dimensional arrays, which represent the objects that index both dimensions of the array on the same plot. Here we consider the case of a two-way table, where a biplot may be used to represent both the row and column categories simultaneously. A two-dimensional biplot is constructed from a rank-2 representation of the data. For two-way tables, the generalized bilinear model defines one such representation: \begin{equation*} g(\mu_{ij}) = \eta_{ij} = \alpha_{1i}\beta_{1j} + \alpha_{2i}\beta_{2j} \end{equation*} since we can alternatively write \begin{align*} \boldsymbol{\eta} &= \begin{pmatrix} \alpha_{11} & \alpha_{21} \\ \vdots & \vdots \\ \alpha_{1n} & \alpha_{2n} \\ \end{pmatrix} \begin{pmatrix} \beta_{11} & \dots & \beta_{1p} \\ \beta_{21} & \dots & \beta_{2p} \\ \end{pmatrix} \\ &= \boldsymbol{AB}^T \end{align*} where the columns of $A$ and $B$ are linearly independent by definition. To demonstrate how the biplot is obtained from this model, we shall use the \Robject{barley} data set which gives the percentage of leaf area affected by leaf blotch for ten varieties of barley grown at nine sites \citep{Wedd74,Gabr98}. As suggested by \citet{Wedd74} we model these data using a logit link and a variance proportional to the square of that of the binomial, implemented as the \Rfunction{wedderburn} family in \Rpackage{gnm} (see also Section \ref{sec:glms}): @ <>= set.seed(83) biplotModel <- gnm(y ~ -1 + instances(Mult(site, variety), 2), family = wedderburn, data = barley) @ %def The effect of site $i$ can be represented by the point \[ (\alpha_{1i}, \alpha_{2i}) \] in the space spanned by the linearly independent basis vectors \begin{align*} a_1 = (\alpha_{11}, \alpha_{12}, \ldots \alpha_{19})^T\\ a_2 = (\alpha_{21}, \alpha_{22}, \ldots \alpha_{29})^T\\ \end{align*} and the variety effects can be similarly represented. Thus we can represent the sites and varieties separately as follows \begin{Sinput} sites <- pickCoef(biplotModel, "[.]site") coefs <- coef(biplotModel) A <- matrix(coefs[sites], nc = 2) B <- matrix(coefs[-sites], nc = 2) par(mfrow = c(1, 2)) plot(A, pch = levels(barley$site), xlim = c(-5, 5), ylim = c(-5, 5), main = "Site Effects", xlab = "Component 1", ylab = "Component 2") plot(B, pch = levels(barley$variety), xlim = c(-5, 5), ylim = c(-5, 5), main = "Variety Effects", xlab = "Component 1", ylab = "Component 2") \end{Sinput} \begin{figure}[!tbph] \begin{center} \includegraphics[width = 6in]{fig-Effect_plots.pdf} \end{center} \caption{Plots of site and variety effects from the generalized bilinear model of the barley data.} \label{fig:Effect_plots} \end{figure} Of course the parameterization of the bilinear model is not unique and therefore the scale and rotation of the points in these plots will depend on the random seed. By rotation and reciprocal scaling of the matrices $A$ and $B$, we can obtain basis vectors with desirable properties without changing the fitted model. In particular, if we rotate the matrices $A$ and $B$ so that their columns are orthogonal, then the corresponding plots will display the euclidean distances between sites and varieties respectively. If we also scale the matrices $A$ and $B$ so that the corresponding plots have the same units, then we can combine the two plots to give a conventional biplot display. The required rotation and scaling can be performed via singular value decomposition of the fitted predictors: @ <>= barleyMatrix <- xtabs(biplotModel$predictors ~ site + variety, data = barley) barleySVD <- svd(barleyMatrix) A <- sweep(barleySVD$u, 2, sqrt(barleySVD$d), "*")[, 1:2] B <- sweep(barleySVD$v, 2, sqrt(barleySVD$d), "*")[, 1:2] rownames(A) <- levels(barley$site) rownames(B) <- levels(barley$variety) colnames(A) <- colnames(B) <- paste("Component", 1:2) A B @ %def These matrices are essentially the same as in \citet{Gabr98}. From these the biplot can be produced, for sites $A \ldots I$ and varieties $1 \dots 9, X$: @ <>= barleyCol <- c("red", "blue") plot(rbind(A, B), pch = c(levels(barley$site), levels(barley$variety)), col = rep(barleyCol, c(nlevels(barley$site), nlevels(barley$variety))), xlim = c(-4, 4), ylim = c(-4, 4), main = "Biplot for barley data", xlab = "Component 1", ylab = "Component 2") text(c(-3.5, -3.5), c(3.9, 3.6), c("sites: A-I","varieties: 1-9, X"), col = barleyCol, adj = 0) @ %def \begin{figure}[!tbph] \begin{center} \includegraphics{gnmOverview-Biplot1.pdf} \end{center} \caption{Biplot for barley data} \label{fig:Biplot1} \end{figure} The biplot gives an idea of how the sites and varieties are related to one another. It also allows us to consider whether the data can be represented by a simpler model than the generalized bilinear model. We see that the points in the biplot approximately align with the rotated axes shown in Figure \ref{fig:Biplot2}, such that the sites fall about a line parallel to the ``h-axis'' and the varieties group about two lines roughly parallel to the ``v-axis''. @ <>= plot(rbind(A, B), pch = c(levels(barley$site), levels(barley$variety)), col = rep(barleyCol, c(nlevels(barley$site), nlevels(barley$variety))), xlim = c(-4, 4), ylim = c(-4, 4), main = "Biplot for barley data", xlab = "Component 1", ylab = "Component 2") text(c(-3.5, -3.5), c(3.9, 3.6), c("sites: A-I","varieties: 1-9, X"), col = barleyCol, adj = 0) abline(a = 0, b = tan(pi/3)) abline(a = 0, b = -tan(pi/6)) abline(a = 2.6, b = tan(pi/3), lty = 2) abline(a = 4.5, b = tan(pi/3), lty = 2) abline(a = 1.3, b = -tan(pi/6), lty = 2) text(2.8, 3.9, "v-axis", font = 3) text(3.8, -2.7, "h-axis", font = 3) @ %def %abline(a = 0, b = tan(3*pi/10), lty = 4) %abline(a = 0, b = -tan(pi/5), lty = 4) \begin{figure}[!tbph] \begin{center} \includegraphics{gnmOverview-Biplot2.pdf} \end{center} \caption{Biplot for barley data, showing approximate alignment with rotated axes.} \label{fig:Biplot2} \end{figure} This suggests that the sites could be represented by points along a line, with co-ordinates \begin{equation*} (\gamma_i, \delta_0). \end{equation*} and the varieties by points on two lines perpendicular to the site line: \begin{equation*} (\nu_0 + \nu_1I(i \in \{2, 3, 6\}), \omega_j) \end{equation*} This corresponds to the following simplification of the bilinear model: \begin{align*} &\alpha_{1i}\beta_{1j} + \alpha_{2i}\beta_{2j} \\ \approx &\gamma_i(\nu_0 + \nu_1I(i \in \{2, 3, 6\})) + \delta_0\omega_j \end{align*} or equivalently \begin{equation*} \gamma_i(\nu_0 + \nu_1I(i \in \{2, 3, 6\})) + \omega_j, \end{equation*} the double additive model proposed by \citet{Gabr98}. We can fit this model as follows: @ <>= variety.binary <- factor(match(barley$variety, c(2,3,6), nomatch = 0) > 0, labels = c("rest", "2,3,6")) doubleAdditive <- gnm(y ~ variety + Mult(site, variety.binary), family = wedderburn, data = barley) @ %def Comparing the chi-squared statistics, we see that the double additive model is an adequate model for the leaf blotch incidence: @ <>= biplotModChiSq <- sum(residuals(biplotModel, type = "pearson")^2) doubleAddChiSq <- sum(residuals(doubleAdditive, type = "pearson")^2) c(doubleAddChiSq - biplotModChiSq, doubleAdditive$df.residual - biplotModel$df.residual) @ %def \subsection{Stereotype model for multinomial response} \label{sec:Stereotype} The stereotype model was proposed by \citet{Ande84} for ordered categorical data. It is a special case of the multinomial logistic model, in which the covariate coefficients are common to all categories but the scale of association is allowed to vary between categories such that \[ p_{ic} = \frac{\exp(\beta_{0c} + \gamma_c \boldsymbol{\beta}^T\boldsymbol{x}_{i})}{\sum_{k = 1}^K \exp(\beta_{0k} + \gamma_k \boldsymbol{\beta}^T\boldsymbol{x}_{i})} \] where $p_{ic}$ is the probability that the response for individual $i$ is category $c$ and $K$ is the number of categories. Like the multinomial logistic model, the stereotype model specifies a simple form for the log odds of one category against another, e.g. \begin{equation*} \log\left(\frac{p_{ic}}{p_{ik}}\right) = (\beta_{0c} - \beta_{0k}) + (\gamma_c - \gamma_k)\boldsymbol{\beta}^T\boldsymbol{x}_{i} \end{equation*} In order to model a multinomial response in the generalized nonlinear model framework, we must re-express the data as category counts $Y_i = (Y_{i1}, \ldots, Y_{iK})$ for each individual (or group). Then assuming a Poisson distribution for the counts $Y_{ic}$, the joint distribution of $Y_i$ is Multinomial$(N_i, p_{i1}, \ldots, p_{iK})$ conditional on the total count for each individual $N_i$. The expected counts are then $\mu_{ic} = N_ip_{ic}$ and the parameters of the stereotype model can be estimated through fitting the following model \begin{align*} \log \mu_{ic} &= \log(N_i) + \log(p_{ic}) \\ &= \alpha_i + \beta_{0c} + \gamma_c\sum_r \boldsymbol{\beta}_{r}\boldsymbol{x}_{ir} \\ \end{align*} where the ``nuisance'' parameters $\alpha_i$ ensure that the multinomial denominators are reproduced exactly, as required. The \Rpackage{gnm} package includes the utility function \Rfunction{expandCategorical} to re-express the categorical response as category counts. By default, individuals with common values across all covariates are grouped together, to avoid redundancy. For example, the \Robject{backPain} data set from \citet{Ande84} describes the progress of patients with back pain. The data set consists of an ordered factor quantifying the progress of each patient, and three prognostic variables. We re-express the data as follows: @ <>= set.seed(1) subset(backPain, x1 == 1 & x2 == 1 & x3 == 1) backPainLong <- expandCategorical(backPain, "pain") head(backPainLong) @ %def We can now fit the stereotype model to these data: @ <>= oneDimensional <- gnm(count ~ pain + Mult(pain, x1 + x2 + x3), eliminate = id, family = "poisson", data = backPainLong) oneDimensional @ %def specifying the \Robject{id} factor through \Rfunarg{eliminate} so that the 12 \Robject{id} effects are estimated more efficiently and are excluded from printed model summaries by default. This model is one dimensional since it involves only one function of $\mathbf{x} = (x1, x2, x3)$. We can compare this model to one with category-specific coefficients of the $x$ variables, as may be used for a qualitative categorical response: @ <>= threeDimensional <- gnm(count ~ pain + pain:(x1 + x2 + x3), eliminate = id, family = "poisson", data = backPainLong) threeDimensional @ %def This model has the maximum dimensionality of three (as determined by the number of covariates). The ungrouped multinomial log-likelihoods reported in \citet{Ande84} are given by \begin{equation*} \sum_{i,c} y_{ic}\log(p_{ic}) = \sum_{i,c} y_{ic}\log(\mu_{ic}/n_{ic}) \end{equation*} We write a simple function to compute this and the corresponding degrees of freedom, then compare the log-likelihoods of the one dimensional model and the three dimensional model: @ <>= logLikMultinom <- function(model, size){ object <- get(model) l <- sum(object$y * log(object$fitted/size)) c(nParameters = object$rank - nlevels(object$eliminate), logLikelihood = l) } size <- tapply(backPainLong$count, backPainLong$id, sum)[backPainLong$id] t(sapply(c("oneDimensional", "threeDimensional"), logLikMultinom, size)) @ %def showing that the \Robject{oneDimensional} model is adequate. To obtain estimates of the category-specific multipliers in the stereotype model, we need to constrain both the location and the scale of these parameters. The latter constraint can be imposed by fixing the slope of one of the covariates in the second multiplier to \Robject{1}, which may be achieved by specifying the covariate as an offset: @ <>= ## before constraint summary(oneDimensional) oneDimensional <- gnm(count ~ pain + Mult(pain, offset(x1) + x2 + x3), eliminate = id, family = "poisson", data = backPainLong) ## after constraint summary(oneDimensional) @ %def The location of the category-specific multipliers can constrained by setting one of the parameters to zero, either through the \Rfunarg{constrain} argument of \Rfunction{gnm} or with \Rfunction{getContrasts}: @ <>= getContrasts(oneDimensional, pickCoef(oneDimensional, "[.]pain")) @ %def giving the required estimates. \subsection{Lee-Carter model for trends in age-specific mortality} In the study and projection of population mortality rates, the model proposed by \cite{LeeCart92} forms the basis of many if not most current analyses. Here we consider the quasi-Poisson version of the model \citep{Wilm93, Alho00, BrouDenuVerm02, RensHabe03}, in which the death count $D_{ay}$ for individuals of age $a$ in year $y$ has mean $\mu_{ay}$ and variance $\phi\mu_{ay}$ (where $\phi$ is 1 for Poisson-distributed counts, and is respectively greater than or less than 1 in cases of over-dispersion or under-dispersion). In the Lee-Carter model, the expected counts follow the log-bilinear form \[ \log(\mu_{ay}/e_{ay}) = \alpha_a + \beta_a \gamma_y, \] where $e_{ay}$ is the `exposure' (number of lives at risk). This is a generalized nonlinear model with a single multiplicative term. The use of \Rpackage{gnm} to fit this model is straightforward. We will illustrate by using data downloaded on 2006-11-14 from the Human Mortality Database\footnote{Thanks to Iain Currie for helpful advice relating to this section} (HMD, made available by the University of California, Berkeley, and Max Planck Institute for Demographic Research, at \texttt{http://www.mortality.org}) on male deaths in Canada between 1921 and 2003. The data are not made available as part of \Rpackage{gnm} because of license restrictions; but they are readily available via the web simply by registering with the HMD. We assume that the data for Canadian males (both deaths and exposure-to-risk) have been downloaded from the HMD and organised into a data frame named \Robject{Canada} in \R, with columns \Robject{Year} (a factor, with levels \Rcode{1921} to \Rcode{2003}), \Robject{Age} (a factor, with levels \Rcode{20} to \Rcode{99}), \Robject{mDeaths} and \Robject{mExposure} (both quantitative). The Lee-Carter model may then be specified as \begin{Sinput} LCmodel.male <- gnm(mDeaths ~ Age + Mult(Exp(Age), Year), offset = log(mExposure), family = "quasipoisson", data = Canada) \end{Sinput} Here we have acknowledged the fact that the model only really makes sense if all of the $\beta_a$ parameters, which represent the `sensitivity' of age group $a$ to a change in the level of general mortality \citep[e.g.,][]{BrouDenuVerm02}, have the same sign. Without loss of generality we assume $\beta_a>0$ for all $a$, and we impose this constraint by using \Rcode{Exp(Age)} instead of just \Rcode{Age} in the multiplicative term. Convergence is to a fitted model with residual deviance 32419.83 on 6399 degrees of freedom --- representing clear evidence of substantial overdispersion relative to the Poisson distribution. In order to explore the lack of fit a little further, we plot the distribution of Pearson residuals in Figure \ref{fig:LCresplot}: \begin{Sinput} par(mfrow = c(2,2)) age <- as.numeric(as.character(Canada$Age)) with(Canada,{ res <- residuals(LCmodel.male, type = "pearson") plot(Age, res, xlab="Age", ylab="Pearson residual", main = "(a) Residuals by age") plot(Year, res, xlab="Year", ylab="Pearson residual", main = "(b) Residuals by year") plot(Year[(age>24) & (age<36)], res[(age>24) & (age<36)], xlab = "Year", ylab = "Pearson residual", main = "(c) Age group 25-35") plot(Year[(age>49) & (age<66)], res[(age>49) & (age<66)], xlab = "Year", ylab = "Pearson residual", main = "(d) Age group 50-65") }) \end{Sinput} %$ \begin{figure}[!tbph] \begin{center} \includegraphics[width=6in]{fig-LCall.pdf} \end{center} \caption{Canada, males: plots of residuals from the Lee-Carter model of mortality} \label{fig:LCresplot} \end{figure} Panel (a) of Figure \ref{fig:LCresplot} indicates that the overdispersion is not evenly spread through the data, but is largely concentrated in two age groups, roughly ages 25--35 and 50--65\null. Panels (c) and (d) focus on the residuals in each of these two age groups: there is a clear (and roughly cancelling) dependence on \Robject{Year}, indicating that the assumed bilinear interaction between \Robject{Age} and \Robject{Year} does not hold for the full range of ages and years considered here. A somewhat more satisfactory Lee-Carter model fit is obtained if only a subset of the data is used, namely only those males aged 45 or over: \begin{Sinput} LCmodel.maleOver45 <- gnm(mDeaths ~ Age + Mult(Exp(Age), Year), offset = log(mExposure), family = "quasipoisson", data = Canada[age>44,]) \end{Sinput} The residual deviance now is 12595.44 on 4375 degrees of freedom: still substantially overdispersed, but less severely so than before. Again we plot the distributions of Pearson residuals (Figure \ref{fig:LCresplot2}). \begin{figure}[!tbph] \begin{center} \includegraphics[width=6in]{fig-LCover45.pdf} \end{center} \caption{Canada, males over 45: plots of residuals from the Lee-Carter model of mortality} \label{fig:LCresplot2} \end{figure} Still clear departures from the assumed bilinear structure are evident, especially for age group 81--89; but they are less pronounced than in the previous model fit. The main purpose here is only to illustrate how straightforward it is to work with the Lee-Carter model using \Rfunction{gnm}, but we will take this example a little further by examining the estimated $\beta_a$ parameters from the last fitted model. We can use \Rfunction{getContrasts} to compute quasi standard errors for the logarithms of $\hat\beta_a$ --- the logarithms being the result of having used \Rcode{Exp(Age)} in the model specification --- and use these in a plot of the coefficients: \begin{Sinput} AgeContrasts <- getContrasts(LCmodel.maleOver45, 56:100) ## ages 45 to 89 only \end{Sinput} \begin{figure}[!tbph] \begin{center} \includegraphics{fig-LCqvplot.pdf} \end{center} \caption{Canada, males over 45, Lee-Carter model: relative sensitivity of different ages to change in total mortality.} \label{fig:LCqvplot} \end{figure} The plot shows that sensitivity to the general level of mortality is highest at younger ages, as expected. An \emph{unexpected} feature is the clear outlying positions occupied by the estimates for ages 51, 61, 71 and 81: for each of those ages, the estimated $\beta_a$ coefficient is substantially less than it is for the neighbouring age groups (and the error bars indicate clearly that the deviations are larger than could plausibly be due to chance variation). This is a curious finding. An explanation comes from a look back at the raw death-count data. In the years between 1921 and 1940, the death counts for ages 31, 41, 51, 61, 71 and 81 all stand out as being very substantially lower than those of neighbouring ages (Figure \ref{fig:deaths2140}: the ages concerned are highlighted in solid red). The same does \emph{not} hold for later years: after about 1940, the `1' ages fall in with the general pattern. This apparent `age heaping\footnote{Age heaping is common in mortality data: see \url{http://www.mortality.org/Public/Overview.php}}' explains our finding above regarding the $\beta_a$ coefficients: whilst all age groups have benefited from the general trend of reduced mortality, the `1' age groups appear to have benefited least because their starting point (in the 1920s and 1930s) was lower than would have been indicated by the general pattern --- hence $\hat\beta_a$ is smaller for ages $a=31$, $a=41$,\ldots, $a=81$. \begin{figure}[!tbph] \begin{center} \includegraphics{fig-deaths1921-1940.pdf} \end{center} \caption{Canada, males: Deaths 1921 to 1940 by age} \label{fig:deaths2140} \end{figure} \subsection{Exponential and sum-of-exponentials models for decay curves} A class of nonlinear functions which arise in various application contexts --- a notable one being pharmacokinetic studies -- involves one or more \emph{exponential decay} terms. For example, a simple decay model with additive error is \begin{equation} \label{eq:singleExp} y = \alpha + \exp(\beta + \gamma x) + e \end{equation} (with $\gamma<0$), while a more complex (`sum of exponentials') model might involve two decay terms: \begin{equation} \label{eq:twoExp} y = \alpha + \exp(\beta_1 + \gamma_1 x) + \exp(\beta_2+ \gamma_2 x) + e. \end{equation} Estimation and inference with such models are typically not straightforward, partly on account of multiple local maxima in the likelihood \citep[e.g.,][Ch.3]{Sebe89}. We illustrate the difficulties here, with a couple of artificial examples. These examples will make clear the value of making repeated calls to \Rfunction{gnm}, in order to use different, randomly-generated parameterizations and starting values and thus improve the chances of locating both the global maximum and all local maxima of the likelihood. \subsubsection{Example: single exponential decay term} Let us first construct some data from model (\ref{eq:singleExp}). For our illustrative purposes here, we will use \emph{noise-free} data, i.e., we fix the variance of $e$ to be zero; for the other parameters we will use $\alpha=0$, $\beta = 0$, $\gamma = -0.1$. @ <>= x <- 1:100 y <- exp(- x / 10) set.seed(1) saved.fits <- list() for (i in 1:100) saved.fits[[i]] <- gnm(y ~ Exp(1 + x), verbose = FALSE) table(zapsmall(sapply(saved.fits, deviance))) @ %def The \Robject{saved.fits} object thus contains the results of 100 calls to \Rfunction{gnm}, each using a different, randomly-generated starting value for the vector of parameters $(\alpha, \beta, \gamma)$. Out of 100 fits, 52 reproduce the data exactly, to machine accuracy. The remaining 48 fits are all identical to one another, but they are far from globally optimal, with residual sum of squares 3.61: they result from divergence of $\hat\gamma$ to $+\infty$, and correspondingly of $\hat\beta$ to $-\infty$, such that the fitted `curve' is in fact just a constant, with level equal to $\bar{y}=0.09508$. For example, the second of the 100 fits is of this kind: @ <>= saved.fits[[2]] @ %def The use of repeated calls to \Rfunction{gnm}, as here, allows the local and global maxima to be easily distinguished. \subsubsection{Example: sum of two exponentials} We can conduct a similar exercise based on the more complex model (\ref{eq:twoExp}): @ <>= x <- 1:100 y <- exp(- x / 10) + 2 * exp(- x / 50) set.seed(1) saved.fits <- list() for (i in 1:100) { saved.fits[[i]] <- suppressWarnings(gnm(y ~ Exp(1 + x, inst = 1) + Exp(1 + x, inst = 2), verbose = FALSE)) } table(round(unlist(sapply(saved.fits, deviance)), 4)) @ %def In this instance, only 27 of the 100 calls to \Rfunction{gnm} have successfully located a local maximum of the likelihood: in the remaining 73 cases the starting values generated were such that numerical problems resulted, and the fitting algorithm was abandoned (giving a \Robject{NULL} result). Among the 27 `successful' fits, it is evident that there are three distinct solutions (with respective residual sums of squares equal to 0.1589, 41.64, and essentially zero --- the last of these, the exact fit to the data, having been found 20 times out of the above 27). The two non-optimal local maxima here correspond to the best fit with a single exponential (which has residual sum of squares 0.1589) and to the fit with no dependence at all on $x$ (residual sum of squares 41.64), as we can see by comparing with: @ <>= singleExp <- gnm(y ~ Exp(1 + x), start = c(NA, NA, -0.1), verbose = FALSE) singleExp meanOnly <- gnm(y ~ 1, verbose = FALSE) meanOnly plot(x, y, main = "Two sub-optimal fits to a sum-of-exponentials curve") lines(x, fitted(singleExp)) lines(x, fitted(meanOnly), lty = "dashed") @ %def \begin{figure}[!tbph] \centering \includegraphics{gnmOverview-doubleExp2.pdf} \caption{Two sub-optimal fits to a sum-of-exponentials curve} \label{fig:doubleExp} \end{figure} In this example, it is clear that even a small amount of noise in the data would make it practically impossible to distinguish between competing models containing one and two exponential-decay terms. In summary: the default \Rfunction{gnm} setting of randomly-chosen starting values is useful for identifying multiple local maxima in the likelihood; and reasonably good starting values are needed if the global maximum is to be found. In the present example, knowing that $\gamma_1$ and $\gamma_2$ should both be small and negative, we might perhaps have tried @ <>= gnm(y ~ instances(Exp(1 + x), 2), start = c(NA, NA, -0.1, NA, -0.1), verbose = FALSE) @ %def which reliably yields the (globally optimal) perfect fit to the data. \newpage \appendix \section{User-level functions} We list here, for easy reference, all of the user-level functions in the \Rpackage{gnm} package. For full documentation see the package help pages. \begin{table}[!h] \begin{tabular*}{\textwidth}{@{}p{0.2in}p{1.3in}p{4.5in}@{}} \toprule \multicolumn{3}{l}{\textbf{Model Fitting}} \\ \midrule & \Rfunction{gnm} & fit generalized nonlinear models \\ \midrule \multicolumn{3}{l}{\textbf{Model Specification}} \\ \midrule & \Rfunction{Diag} & create factor differentiating diagonal elements \\ & \Rfunction{Symm} & create symmetric interaction of factors \\ & \Rfunction{Topo} & create `topological' interaction factors \\ & \Rfunction{Const} & specify a constant in a \Rclass{nonlin} function predictor \\ & \Rfunction{Dref} & specify a diagonal reference term in a \Rfunction{gnm} model formula \\ & \Rfunction{Mult} & specify a product of predictors in a \Rfunction{gnm} formula \\ & \Rfunction{MultHomog} & specify a multiplicative interaction with homogeneous effects in a \Rfunction{gnm} formula \\ & \Rfunction{Exp} & specify the exponential of a predictor in a \Rfunction{gnm} formula \\ % & \Rfunction{Log} & specify the natural logarithm of a predictor in a % \Rfunction{gnm} formula \\ % & \Rfunction{Logit} & specify the logit of a predictor in a % \Rfunction{gnm} formula \\ & \Rfunction{Inv} & specify the reciprocal of a predictor in a \Rfunction{gnm} formula \\ % & \Rfunction{Raise} & specify a predictor raised to a constant % power in a \Rfunction{gnm} formula \\ & \Rfunction{wedderburn} & specify the Wedderburn quasi-likelihood family \\ \midrule \multicolumn{3}{l}{\textbf{Methods and Accessor Functions}} \\ \midrule & \Rmethod{confint.gnm} & compute confidence intervals of \Rclass{gnm} parameters based on the profiled deviance \\ & \Rmethod{confint.profile.gnm} & compute confidence intervals of parameters from a \Rclass{profile.gnm} object \\ & \Rmethod{predict.gnm} & predict from a \Rclass{gnm} model \\ & \Rmethod{profile.gnm} & profile deviance for parameters in a \Rclass{gnm} model \\ & \Rmethod{plot.profile.gnm} & plot profile traces from a \Rclass{profile.gnm} object \\ & \Rmethod{summary.gnm} & summarize \Rclass{gnm} fits \\ & \Rfunction{residSVD} & multiplicative approximation of model residuals \\ & \Rfunction{exitInfo} & print numerical details of last iteration when \Rfunction{gnm} has not converged \\ & \Rfunction{ofInterest} & extract the \Robject{ofInterest} component of a \Rclass{gnm} object \\ & \Rfunction{ofInterest<-} & replace the \Robject{ofInterest} component of a \Rclass{gnm} object \\ & \Rfunction{parameters} & get model parameters from a \Rclass{gnm} object, including parameters that were constrained \\ & \Rfunction{pickCoef} & get indices of model parameters \\ & \Rfunction{getContrasts} & estimate contrasts and their standard errors for parameters in a \Rclass{gnm} model \\ & \Rfunction{checkEstimable} & check whether one or more parameter combinations in a \Rclass{gnm} model is identified \\ & \Rfunction{se} & get standard errors of linear parameter combinations in \Rclass{gnm} models \\ & \Rfunction{Dref} & estimate weights and corresponding standard errors for a diagonal reference term in a \Rclass{gnm} model \\ & \Rfunction{termPredictors} & (\emph{generic}) extract term contributions to predictor \\ \midrule \multicolumn{3}{l}{\textbf{Auxiliary Functions}} \\ \midrule & \Rfunction{asGnm} & coerce an object of class \Rclass{lm} or \Rclass{glm} to class \Rclass{gnm} \\ & \Rfunction{expandCategorical} & expand a data frame by re-expressing categorical data as counts \\ & \Rfunction{getModelFrame} & get the model frame in use by \Rfunction{gnm} \\ & \Rfunction{MPinv} & Moore-Penrose pseudoinverse of a real-valued matrix \\ & \Rfunction{qrSolve} & Minimum-length solution of a linear system\\ \end{tabular*} \end{table} \newpage \bibliography{gnm} \bibliographystyle{chicago} \end{document} gnm/vignettes/screenshot2.png0000644000176000001440000020065112515147523016051 0ustar ripleyusers‰PNG  IHDR¢1‘MI‰zTXtRaw profile type exifxÚUŽË Å0 ïª"%¬>–¬rBH ¼òŸŒLæ -‹™Îß}Ñ6`Y‹îéŽÂÒRö XÀcל<[¹’¬šTfðì[íé_šz÷+,›~HÙåTV©Y‡hX1¾‘K² ¦F¿½2>zúH,Hz+²Ô ìiTXtXML:com.adobe.xmp ÎLé)sBITÛáOà IDATxÚìy\•×ÿ¿çYî~á²Èª,Š€(*hÔ˜°‰ *·Ä¨i3é4i§“IÚiºLû³ÓšîM§M§N“IÚÄÅ%£ˆ¢¨HÜ‚+Š€ˆ²_.p÷û,ç÷ǽlÊrp!ß÷ëæ•ëóœ{¾ßsžÃó9ßóœç’œ”(ɲ^§€°°è }&Þ'îKÛoN×1½—Ä÷לÓ¾ïÙÁÚÚz0™Z†aX†“eª×éüü `·;AAÇ· »•át:­ÁàK)ŪAA‘Áàë‰é±.AdD"Içëëãe(O)•©Le @½ù ¥@¸SRJ(!0,ËB†¡”BðJ ‚ ÈÐâëëœw‚Me*SIA”$I¢nyfY±,ÏqwO£”JRY$Y’dY–)ea†aYžç–e†‚J ‚ Ãç•ÆKT§Ã™6#Ü?Ê Ö©,öªVó©Û-;¯ÖªTJŽãºÿ¡”Š‚èt¹¢¢¢ÂÃÃýü &S«ÑØ|ûvmUU•R¡ϲ,¸ÃüG *‹%,ËÞ‡_²hn5ÛdNgÐk9lc_&¨l·X­Q¢ §7è4uYíf›K¤ çã§cmí&Uêô¾ê>×´ŸÁ4`AFdaVæ/Ë.—0JÁü&ub˜NIe d™Ê2a`°µçŠ.59%^¡ÏÛ”RêžçÓÓÒüý ’'ž†cY†e™––ÖÃGŽ‚ R(XŽõ^æ)•œ6‡Ý)Š¥ Ë*”J­F1„·3*;Û,|@¨?g1Úào”²hƬZ÷Å~×?zóí*Tú/ÆKV›~ÆšŸK‰¥ÉT¶ñ—ÿ½¿I3aÉW¾š5eŒ–@[éÿþçÿ5fýà?æë.¼ÿ«·OÚ´*† CûTFda6[´—©,b¸’};=AM¡Õ|wš`…ê¯ÿµðb£$rGÈ2ˆ’¤T(çÏŸÇrÄb±P Tvr† Z:kÁ‚}ûö»DAAËz5J.³kÔÌœ% gO7J͵6\;÷ù¾-{Ëì ~Hng”ºÚŤŸ|ðr<4ä¾ñ“=•–»‡|©,H>qÉcx€¸'¢T{ÏÈ2‹Ï$¾"/X¥¨¾ÿ¯™¾ 9FǶ¶K~s_ýÑŠh.§ÀºLŽ€i3Âx¦=É]• f¨ÛÏ 0‚ # n ÑQpýdV gmq:í½'²¶ó*ÍO“Â_þü:Dz²,¸\i©OS*Úm"íé@žWÌž=ãàÁÃî§ô^Ê‚™$¼ò›WÓ»F#´Á1OÍ*ÝöéEªÙesXí‚HV¥Vë<㢒Ýl·»d‰R ,«T«uJ÷í“JN§ÅætI„0,¯Õ³@ÛaÒÖÚnãTþ%Ke‡Õaw‰¢ìÎDákPQ«Õì”<x^£U«8¥ÊTø?ÿP=í{³àŒS­`}š¾³x÷î|ç¯D‘R°JÍ=fÞK¡Ô|×Ï%»ÙndI¦„pœZID‡à’(V¡Vùh:äžò÷Îg*»ì«Ãý[° ¯RëÕ,é¯B¼w¸§B;V» HáÔZƒ†%}^Ž>ê$ O›á œ}û_o^©$¾)Ñ vëº7¶ßrV¥VÔm>ž¦-Ûv™zjúë«{ûQ>/qïõàžêÚK¾«©ë|ÔJ|ÕAF¼Ì÷ÍS ¢(, ÑQ›ÐÚ 6W³]HкTšl:ž Ó)%»9DgXª;`t±,'Hbxx˜J¥²Ùìî|Y–)u?²÷{Bˆ ˆjµ&,<¬©¡‘㚤N«rú·_N €†¢¿oØyú¦ITŠŠÇT˜¼ŠŠV³bü⯯͚>Ö—Z®}þY·8ëtøgþË‹ó&„‡ø* €ÐrãØ® ÿ(jT¨Ù)òs~°&cJˆ dgëÍ“ïþnÃÉ3N±ê­wWÀÍMÿúãÃEÀŒç¿–5m|Ô(- `¿uð—ëvðÏÿè_S£4 €£±üÄö¿o;ÑJ:nÙ ’¹væìé÷kF-ìÝtOa£’Ír¯Î³„H6‹".û•&Gê—ñÊÑO6n8Þ¤Tq^e®e§~ú…úÉ–:•Š# Ûþ™ßøJfÂèPƒŠÙ\}æÀ‰Ö¨”ÙIap4œÍÛø—]W-ÏÜ[þ›oˆú KòY–lvõ„Å_yn^Ò8ÔÙÖP[Q´áóëE2jnïB.ïîÙrXIxÚÚ²gN¥¡µbß_½½F’T½]ž>ê\ 2£òUŸôúÛ[^°~¾îõ}¾€°gßüøY¸ñ×oý—%#cês#oOU³Š#ö»¬ÿâeGû9óA½èröz‰‡³÷†Á+ºÚº5à#£¼³©o¼,©¨ô2¢a(…¾> ˆRÚ(µhm—EA…‡ô»Ó5ÛdQ(onÿMÉͧä>%ZÛÓG©Q¢$Q µÛí‚ ‚(‚(Š¢(J’ûÿbçq»Ý*JR?nx>’àÔO}&I Uþð?‡®µ‹À»ñÚécWJ\6óâÿ[·jÆX_(ðþãÒ_üÑÏ–;-¢‹úNš>6ÔWId—CÞ?:퟾ûR µº«ú‰×¾½pJˆJjk¬3 Z¥ÝÞÕïÌ­-¦Öúf›Ì‚‹>9wê¸QZ§j¦µ]”$M ï4·YPŧÿË÷Ÿ qr·ú• Óí*šLïÇyAvY`ì ?þñ³Ó#u²¥Ý©Ÿùò¼6MiqÊ^enwJbÀóCÙ%ûNšnP1² £œ¾lÕܤ0$¨‚“–¿ú/Sy«CvÞSþ¢2æ+ú,;í\Â?¯ûgŸg`ÛŒ¢Ò7$zÆäQT„>¯¦ä•ÃN*wµ+*Øäð?úÅK©F)A$Þàë²¹D¶¯9û¬sJ;Öš•-m-¦ÖæºVWg;p´·˜Z[êZìr~t¯ÖÅnË„Êö>Ì9¤¾F÷ßw5`†Úûhêÿéá?øyl?îAû>£yD–¤1*"[mT–`Rá{KÒ~÷ÙáQ†Um¯/JI$í²Ù²,ÖªeI"„PYÒiµ.—ZJ )}”èíÓÊô>ê¼ÈÄ{¬üí÷ß¹"†UpžM"Êßùñº6B%QLîjÓ¢]?ûNë:"¿®$†Yš»«®wÔC·¬Æ÷®¦®Sr—¿D/ñ =0„Rª¡‚[°@nmŠÓ +g$þ£øÜ ³'O”[Äö–ŽNQÉ‚û¼L)E!ÇŽ¼;ïÙ³gº+U¬g@¿™ïÖ î.C÷Ý($‘‹šŽ‹y'[•>\Ýç…W¿:>Ž˜Äœªíú©‚˜Ê+L0ÖO e)ëN—ç-ˆþúÛoÏ=U¸k{þ‰fò]éUzš R÷þR¯ÕpÔiWMZûíofÅè»N)´ ¦’ô4-Ë”2ƒr^’Iø¤€ðçó·ç;SÂý8©NöŒ{Ÿ¹R¯ÕpÝ:f=®a¬Õ¥õ0=”:KZ¯Tš`œŸÚWÃÒ7¼È_MN/}™Ž²‚R«JÅ‚wU*zåpWÓ¡¢ÈEM‰"Îó;ŽÜæô:†Ê¢ƒë»Dt}ù/»·FÃÊön³Y8ZσËL»?˜àÂï´N©ÔýÙžõÝë¡«ßÝÔòO0‰2ßÏ€>!„˜íN¥$Sêm,¿YûIyÛš''ï,)‹Šóà«è¸­2»ƒBB±Û”J@“žžê~$ßÓS v»]¦2°Û„é!¬ØRÓ ‡ ¤‰†­Úy5ËeI”ž®ˆ{Á½½‚nùÉ%¹3$„Q1Uüø—ÕKŸyfٯ͘ùãŸ}lô¸ÃB)•)¹s/" ²ËÅL~ùõ¬…tóà惗£Ÿyi^Dÿ3š»›î&£”ÞŸó”÷ÄEãÉü/:níÔr¥á:ó÷>óΞΎUO:,“îÿ`Ð{ȼô™P(ȲwUJ¼t¸[ÍwúC©ìY´¨Lú)QŸþ““H=•Ù³>)½«õf½Ç¼3wGà pWVÝ[SßÒH4ø¾‚Œìgóà¹M÷ö!”aØ+fY–Y*[íÿu¹õß2ŸXÂ/ëÉ?–·–™lîS”Ê×,N†e€a“©U’d÷óx›Íf·Ûív»Ãáp±Ù¬’$J¢(Irkkòý¹áþ0 gºp¤`ìW^{95JÏ€KäµÁfMõ§"#ܾxT“?apš¥Ð§2c@¬¹Ô(’~°J½XUðÁï_ûúú-u2kZ ã²Y%ð¦tY¬íVñ®Á*KÄ'°Þø #5^iUâ™”hý•}õ>úé˜ Xó@X¹¡²T‰Ù³F9Û,¦6—ÈqRŸ%’Ä>ý'½Yé÷HoÖÝõÛks=-ÝÝ€e*öÖÔAé€zøÁ~ãOÿ3íeYÁóGÖ‰£€Š"hÔš×æÆOp5 MÆqzà f)oPÑÀ*5‹ ^ÃÃq\ss“V«–$‰™ad÷"öîó”R÷Ò·²,s×ØØȳެ¹KÊ–ÝþäÏWŒçÂ3_þAæËg§ùêÿ^n;ùqaöº9†éßúݦo‚ûfY·wÇy' lZ@!hÕÞÌж77Û¸ k-àrœºF“bùißùã6Qa/úßÝ|¹# “MWoX TýÕ·ÞJ¹mSGª¼Ð—îYt™U¶ß»ó„aŹDzޘ­K|ñ§¼(2Ï3®S¿úö®ÒnkªxŸ9½#š÷N6…÷ù–óÆg†åš>ßvvÑkIšÿò«þ¥o¯îÇánådXÆøùÖS ¿3C3í¿úøeAbx¦òÝ—×÷Q"Ç9úðÿÊÝ——örÁ)é~{·þÛ®Õ)h³—æzü¥p­»ðÇ5c^¸»©S¸;%‚Œðh¾Ÿz¢Dy¨½éB*‰a )ÎV+¶4PI”Z›Ç™oŽVÈT Ãœ¶Ð#–ãx ÀqŠ–“Åj“eY—àr¹—Kp¹:¾.A(¥-¦ÖV“‰ãy¯f 2<׸ïgo¼½éheÅ=’+Û›«N)³òËK¼ù«ígk,20 µUþño×m½Å*z,4Öó–ÆðŒõV£õ 1@û­/¶ÿéÏ%v…ºýèŸÿšw¹E¢Ö°öv™cï´FA*7ÿzãÉm’.,:.&Dål¯-¿tÓA½SùnE¸ç)ðJçÙ¿ýìWÛJ®™žgdKÃz™ã(È÷‘yoS4½á^ò'^ùL ¯´üÓOßÚvìü•êkå¥gkEÞñ à>š|2…Ò^ò?o¾õé¹›í0TE¡—žÇ·þõz‡RËó<È2\.I–Ç‹rGð«ât­“Ã0²D¯]»ÊrË2käy,;‚ ÊõDD,Ǫ”œûq·à.÷â( ¯äÕn¥¡²Õâ)QhTj$§Óâ”UøhY›]’d÷ # §àÕJ† TrØ—è^x„Óhxzd⾟J.—]$Ùó$ša9¥šWtOÉöiúîÝ~îÕyÒñ+g‡„a*¥’ƒûËüîñœng©Ëê°KÀªT:éÕ oó÷Âg*SN¯ššÍ"%Šðìÿ\ÿâ8Ƹwýë›j•Jè³B¼w¸g9§àtI’»ðœVÍ1}•¨ÿï.,•§ƒ«Qé<ÃìT°:l0 ¥^Åôn]ÅØ»çã¹; xwfåÞ›:‚ #—K éO÷ŸˆW(Á©t9^öiŸ¦t‚,yÞøvï,Ç°§ÊwÛ}œ µR©„Ž‰Ú²LN§$Jaa!¦û¦[ïm6[]]=Ë2¯D ׃Ez<\¶kSß|ë•hÙaµƒV«¸¹é_î1ƒmAÄ+A`£"ÇôŸH’$¥B)2\‘…¯q’P-¡JÌ2S&(rÍÚm6_V¥Q(:¦Ü „–ePS‹Éét¡ ËPY–l6›ÑØb4¼’ãyQŸ"=Gl$ÐEÆŽ‹ ðñѪ8p¯–|ò?ÿ·§FTà´pAïo¦²LÒÓf{“”ã9†áÜË×ɲ$Ë”að<Ïóa‰BÁ+X¬A{@%ÎË@Zp‰„H,Ë©Õj†!³û1<ÃJ%£¤ž…¶) 2¥” ‚ <ï$ æ‘»[¯äù;ûŽX-‚ ÷çýë4”RYv ÂPGmxAd˜duAAFp4•€ ‚ #VæQçAddB***°AdäñÏ_û*®‚… ‚ #”yAA™GAeAA/óös?͘óóRGŸçÏÿ,#ãÍ Ž¡µê¼±ùµùS'ÄO}öƒjkç÷ÿýl]?¶†ÇAy„eÞUñ‡Ì¸¸¸ÅïÜèX÷N¼ùœ¸¸¸´_]zdÑ~áo.}ú½’Ë綾TÖùýå'—¼özVßûá û9Û§©³ëR±s€ ‚<ÎѼvt˧»n¸@¨ÞóIC˜Ï#½ ¨«¥Æ0ZMz~gg,[úD`;ŸôAAF¨Ì«bWdѼWœàºöé.ç¼g'¨Ý2Omå›°lVB\\ÂÌ¥ßßtÙJ@6_xï[ó¦ÄM˜šñO:Ñ"uÉomÁoÿ)sj\\ܔ̯ÿé˜QêÓ¤³f߯_š›—ðDöw?¹%öeë®^å>a=ý›W~}%ã¿‹/]*þKæµß¼ò›/,ÔRò›oý¹å¹J.–lùº¸³ At'v”ýùŸ¿4á?÷^(;÷ÙA;¿óùM½ ½½ôíþþáèïïlÝç>yÞ_vývš6ñg{OÛÿɾ|Ï÷Ï7½2NÑ_þô姳¶¨rêú-{~´²æO?ÝÓ îé_ÿù¹à°><úÅGÞ]âºð—7÷GþôàÅŠòÒâM?^Êõ‡6|¸ÿ–Ðó;‚ ‚< ¸Î+¢-÷}iKiÓò]…Ú¥ïE)/ºåòòöWæïÿù `ú×¾?ïãoo+{ý¹‡äyX›¨gAÿÔ7_¾å·žÄ¹;ZÒ~õ­ÔP%@ÄüW_y7gû9Ëœ ^Tþré¿ùÎüH5ÆOìËÖ·ù=½ä¹è‰þKÓkþÝÏöæ§ðó¾¶`´"ž~Úÿ£3·]+CÔ=s&¬’±V”_®N2Œõ‹˜œÀÍü¯’óžóšnßä"77+AF«V­òZæ=ÿÙ ¼óžñjÀŠoŽá¯»Šíµ6Ÿ'=A²"0Ú×~¼ÍÙ^kó}2Ð=•hÂ"}Ý»ESµ±µä»Y) @3‰m·Ë½È¼ØzËj˜¬êy°7[ŽÞó 0½æßu¶×<%?`´~÷°áŒäïÞ@5éßþðúïþëÏÿ© bæ|åû?û×Ô |Ø<¬[·+AF•••ÞË<°ÁsV^ÿÃâ)¿XÊ‚Gæ9Ÿ0M{•Ñãyp5ßhS‡ú*}B5¶úv€¡µÑ"Gp†ð€ ð?îÿÅ´î!°½—à–3ŒÖµ]kp¸³…~l©zÍ“¶Ðgé-ÿng{õ³¤ÏìtíûÃøN}áç¾ðsÁxáãï¾ô½ßÎ:ôÖ ¶5ä1¼) òø’››{GÇÝ‹÷æÙQY¿Þø¿]Üžª',Ëäüþƒ3­’Ôzöƒß°sV&øÅ/O§67K®ª]ï³ÊžÄ«s´û~ñç÷ì2HöæÊâÂs-½>›WOXõŒïá_ý©°Æ.ËÎæŠÓ—[¥>lyç€ùw;{/y²ºQ*sõ-‹»ŒÎªýy_Ô˜EÊé}KNÁCA6Þ,CÔ£“žL£î>[MûÄÿö½È}ßœðä7öE¼ñ·ÿxBÇè¦ÿàímo-›¿dùÚŸ]ž4Û3~êI¯¿ÿÛ§+µ4yBÜÄY9ÿþÁéÖ>ÔSøúÿýrvù›ÙS'LH^ôÝMåÚ»-â}žæßuö^òTD-{%õÖÿ{zB\ÒšÜZÁ|yë—Ïœ?5û–gùF²§à!‚ ܈A¾Ô¸‡øpÐAFä_4nD‹ ‚ #”yAA™GAeAAG„^Þ›‹‹ÃzAÇ‚õë×{“¬û’X‚|Ùe²³³±jäg÷îÝ””ÔýàÙ³gï8’éeþN¦M›†µƒ ,ééén™¿CÑÏž=‹•ƒ H'ølAF«V­B¥GeAFÏÚðI<‚ (ó2é|=‚ n8o²¾ç-œ¯<ñî¥Öâ–õÅ­ß;a’h• ‹]åؕߘ¯in(µ^9WÒ,bûC†<”ï èqÚ‚ ÷"óRÛ¥üNÍRÕÐØUÇåd ù»¯÷³e›ìh¼QmâÍ^–]¡áØ'[.X(¶9d¸Cù΀þŽ)÷‚| yXƒö„S)Ù:-gö쯴È#Â.‚<¸P¾3 ÇšA¤ÿhž÷OH››ÌI‚Óeo*=ÑÐã¬" !eÎÔ Ž†#Žš’ƒÅ­]ãÐŒ>&mQÚxhsH²ÃdSØOm+¬Ø€äEóGG†ÐUkgH’éBþ¾Ò®M_]læ¢'‚”êUùÇï7wÛ –õŸš½bŽß¹}×Ç$†òÀp¬íê¡ü“µÑ9뛑1Á à8N¡ ækÅûWÛd€AÚ`ÔaI©qB)Éxñð‘ÒF'õÊ+àFÍX¶v `©éì¾¥FÜ•¶PzA”yF¿pqDå®Üm"«I]œΟì8듸$'ª|Ç–ƒ&xÿĬœ,ºuO¥Í=,­ŒžŸ¡<¹á*€*rþÊXÏ~õ’ñÌg›Ê'®ZèÚ¹ýJ¯ƒç²¥²`K%ðás—GÞjŸÛwdÒkó"¸µÈ,£‹_²25lcA­G4¥¶²ƒ;˺•m¢ˆÎ=•žÑE§®\«!=äú>íRkåÑÚ‰‹fGhÏXǸ´EQWŽÞpxã—f|öšE“ÝFýtŠæü¹;gø!‚ ȃ‹æA¶–ïÝͤ-X=—EQšÏçíó_±˜ï#ÚÜÜÜuëÖá„|ÜñýÊË/á40A™°,‹2 ‚ (ó‚ ‚›GAŒæAÉ2ÏúNÌzþåo¿˜¨}PÞñaY¯¾<ËïwD†Ñ.«zò™Õkž_½víÊ”p%¶@Adxežó6­Ôv)‡S³T54¦Õq9™BþîëýìL+;oT[†xGö‡e¡³Wf–å}²©ÁI±ñ!ÈÈ#66+á¾Á—‡îd~ˆ!œJ9ÀÞ-RË™=ûGŠ]àÞžZº}Gi;nYƒ ¨UHOrssW­Z…õ0ôÑ<Çõ/ó¼BÚܤ`Nœ.{S鉆žái@BÊœ©A% G5%‹+Z»ö cô1i‹ÒÆëD›C’&›Â~j[a­À$/šŸ8:2„®Z;CLò÷•¶vnÌêÙ+N©P•ü~q÷[Yÿ©Ù+æøÛw}Lb( ÇÚ®Ê?YÛ³¾  Žã b¾V¼ÿxµ{Û¼AÙ`ÔaI©qB)Éxñð‘ÒFw\> W K]]\„ ‚<8™ï7šgtñ GTîÊ=Ð&°ú˜ÔÅiáüÉŽ³>‰Kr¢Êwl9h€÷OÌÊÉ¢[÷TÚÜÒÊèùÊ“Þ©r¨"篌õlü.Ï|¶©|⪅®Û¯ô:x.[* ¶T>wyä]¡ö¹}G&½6/òø‡[‹Ì0ºø%+SÃ6ÔzFÙ¥¶²ƒ;Ë<ƒâŒ!yEzD]^•k°vøàÙËSiÑ®Í56€ÕE§ä,¶w[I‹ä…WŒ6H]Ý9úÉœ§bt²ÃX~ôйzºGA†YæûžhÆJN².,osè’ùjqÑM‘ñh5<#ÙQ¸ÿ’ɾ -¥ÅÌŒä@ÖósÑTÝ4qRLX ŠGõþ nÕónguá¡ ³äVæªrk€ßyŽ¨‚Ò—>¿fÍšçŸ{vizü¨ÝP<˜ ÚØ”ðKyÇkÜ# YnͯŽM‰Ryç¯dõÓNvÜöÑÆÜ=å†ÌìI>ø–‚ 2œôûlžÕJ·¿pu;"´Ü4ºž³ZÃè”嫟즄¬ŠoV³îñ–’­›}B"¢cgΈÖ“ÖòÃ_Ô9‡ÈsÚSYîú8sÙýÙ¼OŽ]€ÕOZ¹xhzDº¶¡ÒÖ}Ì]2ךTÁjæšCÀ+ê°©GYüýt½lUÇNL]ës©¤ñAaŒæû–yÉÜÌNT”Õt*=ï ¨÷ë-ÕuÕõŸæW÷!ÜD¡×RK{ÝÕóuWÏFŸó\Ê­ êÄa-_|´ñжrcǾË)RÅPä,YŒrp¨–©o“»”?ÔÏi´{%Ô²íöÕf³óa?•DŠk ‚ Ã-ó}KØtú¬:#s‚ÁÝ`}bÓ$t¾4/µœ;æ˜5oJ’tÒú1““Çê<ªÇç¬I­ê<©5h©«Û”6ÉæÐ…uŒj…!"&\;xÙ“m¢ˆÎýà€ÑE§®\«!=äú>íRkåÑÚ‰‹fGhÏXǸ´EQWŽÞpxç™Øt¡2ð©¤Q<0iºîÚÕ6 åAá”y®¿)x²µ|ïn&mÁêé¼,Š’Ð|>oŸÿêŒå9ʽ{JšEç­¢]4)#çy_–€,¢Ð^uº¸cX› ŸµvŽ‘$Ê°Î[§viì åÕ…'br^z…·[ì.‰ ÍeÅõžŸzæ»óÀ¨ƒÃ5«£œ·î>|ÓD36ã™9ã¢`i¦kWáuã7eAzL4“5½~W‰Q¤æ²‚Òy‹Ö¼@(á¡þ|ñþªe)+žQæå6Šƒ° BýçÛ‹’3sÖ¤2”"6•îßyÁ(À@^¹k³íÂÞ3,[¥áxN¬û"ï°Qˆ ȃÄ~®`Ñ¿3ÊËLTyuyìð½y¡¥ì@nY·—ÿXÞ­`¿}:oËéÞ[Såέý¾>*›+ >ªèõŒ{¾{ïAµízáæÿ-ì ÎMçw½{~W·h½õRþÇ—ºýâü/ÝŽò–ìÞTr^yº6u§vm9…ÍAá‚ÞÚ°!óç-sÿúÍ?ÏQ¬„eq#ZA‘‚`ÊËmöbOl¬nÁ‡‚ÜÛb·‚ È#ëú¥­·B¿ýçÑï½zæHÓøåÁ È溿ÿ`÷Ÿ¶‘°ˆÕ«”è܉û:ŽŒ4™ÇhAdD Vl»Ô8qrÆ´ ‹ƒê7°H®’ßìøsË”Jþ½dË qç•Ï©¾Ž## ÜoAd„`«Û¼ÇžølT ÚoQ¶¾|Sy`oÜqˆÎû÷¤D=Êüæ«£=sìú:Ž`4 ‚¥a€\8!üÚ…]7$µ6ÕØ@÷žhÂü|Ý/÷ui2Ïá³yA òŽS¯µ»„eþñ? •2ýd{ó+¯+C5B}» À€Ðj±È®ãFó‚ È£†l¼±ásHÿÝ×|åÀW|eëwƒ?»PFƒ–§Óªš%—i×ûÕV÷ |UÇ‘'óFó‚ »Ê7ž;ã3ñƒ¹AÏ¡€å3ŸüëMçÓ~óƒg^üÁÞeó‹Gùk'L‹ ¬¢˜Þëqd„Áá u‚ =LèskJŸëy(0þ½³ñþò_¿þrç‰v$ðéý82¢yïežõ8oáìxå‰w?(µ>ïø°¬o,nÝøÞ ÓƒL»ªØ5ßÌ`Ú;–·å|‚Å¿nºâÄ6ˆ ‚<2/µ]ÊßáÔ,¢·.Ôq9™Bþîëýìh¼Qm†¶ÈÇ®Øzùè§O×{2åBç<oÆ%íAGF懩”ËRË™=ûG†]±ñl·¥ÿ9ÿ˜`ã•£¸‚ 2¬p½PÇû'¤ÍM æ$Áé²7•žhèqV2gjG ÃGMÉÁâŠÖ.íbô1i‹ÒÆëD›C’&›Â~j[a­À$/šŸ8:2„®Z;CLò÷•¶v†·ž½â”êUùÇï7w‹{Yÿ©Ù+æøÛw}Lb( ÇÚ®Ê?YÛ³¾  Žã b¾V¼ÿxµ{Û¼AÙ`ÔaI©qB)Éxñð‘ÒF'õÊ«+ßq£Û*N ØþA‡Í3ºø…‹#*wåhX}Lêâ´pþdÇYŸÄ%9Qå;¶4‰Àû'fådÑ­{*m”Ñó3”'7¼SåPEÎ_ëÙ.I2žùlSùÄU ];·÷þtÚ³W>wyä]¡ö¹}G&½6/òø‡[‹Ì0ºø%+SÃ6ÔzdSj+;¸³Œzœ4$¯H¨Ë«r Ö.<{y*-Úµ¹Æ&°ºè”œ¥Óön+i‘¼ðª{µê£#-åg\ØüAá–ù~^¨ãF%'Ùo/osè’ùjqQtü©5h©«Û”6ÉæÐEøñWê Ø­ñúíÁ.·(;ÚDÿ{Ë$0ºè§s–Ä*öï…ÁÛ¥ÖÊ£µ«Í6~zì¦M`õãÒE]Ù[â¸G•ã¯±Rlx2©¬¬\µjÖò˜È<ÈÖò½»™´«§ó²(JBóù¼}þ«3–ç(÷î)i·ŠvѤŒœç}Y² ˆB{ÕéâŽam*XLt|ÖÚ9>D’(Ã:oÚy¤±+”wVžˆÉyéÞn±»$*4—×{~ê™ïΣ×D¬ŽvRpÞ>ºûðMÍØŒg挋‚¥™®]…×mŒß”é1ÑLÖôú]%F‘šË Jç-Zó¡„c„úóÅû«–¥¬xF™—wÚ(Â.õŸo/JÎÌY“ÊPJˆØTºç£yÕ­¶ýBåÊR3® ‚<HEEŇâââ²³³§M›†µƒ ,éééëׯOJJê?evvveee_gsss×­[×OA#îø‹¾x®w¨CA Ê<‚ ‚ Ì#‚ ‚2 ‚ Ê<‚ ‚ (ó‚ ‚ Ì#‚ ‚2 ‚ (ó‚ ‚<ÎpÞ&d}'Î[8;^yâÝJ­Ä5>,ë‹[7¾wÂô`‡»¬~üÓ fŽQQJÀz­xÿ‰j®y‹ #‹ØØX¬„û—b|Ø2/µ]ÊßáÔ,U ]u\N¦¿ûz?;ÓÊŽÆÕahËûì²OdOµä~Ô,(Þ^±p’iÇ…vzA­B 77wý&Ö =áTJv€ŽEË™=û+-òH°« —/7»ûÎúKW‘¶>Aä¡Fó¼BÚܤ`Nœ.{S鉆žÚ2gjG ÃGMÉÁâŠÖ®íØ}LÚ¢´ñ:Ñæd‡É¦°ŸÚVX+°É‹æ'ŽŽ ¡«ÖÎ$Ó…ü}¥­ûÔzöŠSªTå¿_ÜÜm[ÖjöŠ9~çö]“Êñ¶«‡òOÖvFç¬oBFƃ‚ã8…‚˜¯ï?î”]F–”‘g ”‘Œ)õlÂ; WÎÚ‹æ¹I×ߴɬÏؤˆ–Ël}‚ ÈÓyF¿pqDå®Üm"«I]œΟì8듸$'ª|Ç–ƒ&xÿĬœ,ºuO¥Í½™º2z~†òä†wªªÈù+c‰'Z6žùlSùÄU ];·_éuð\¶Tl©>|îòÈ»BísûŽLzm^äñ·™%`tñKV¦†m,¨õ¨¦ÔVvpg™gCwƼ"=¢.¯Ê5X»|ðìå©´h×æ› Àê¢Sr–NÛ»­¤EòÂ+¹ýҾϟ~ñÕo3ÍvŸËá÷>¾jÃMçA‡'óܨä$ûáíåmî]2_-.ŠŽŸçÑj.xF²£pó%“'|ZJ Š£W&^+n’DSu[ÐÄI1®ªÆS»£zÿGÕCæ´³ºðP…Yr+sU¹uª‚JTA ³f'†iY\.~”þà¬I¢M ¿”·µÆ3qN²Ü8š¼*%êâ§×{Åèâç§+ÿõí«f‰óK˜» sì§×í¨ô‚ ÈC’yV(Ýþ¢»> -7.…ç¬Öß0:eùê'»)!«â›Õ,€{@¼¥dëfŸˆèØ™3"‚õ¤µüpÁuÎ!r›v $•å®p3—-ПÍûäˆÑEXý¤•‹‡Ä « `*{LŽ—̵&U°š¹æð ß Sä;®š%MeGÎÄfÇho”ZPçA‡#ó’¹™¨(«éTzÞ?"@Qïþ.¶T×Uך_݇p…^K-íuWÏ×]= m|Îs)·6¨‡³4¬_|´ñжrcǾË)RÅPä,YŒrp¨–©o“»”?ÔÏi´{3U° "I]š.‹«c Ê<‚ 2œô=Ó^l:}V‘9Áàî °>±é ´²×rî˜cÖ¼)AJÒÑaЙœgMêhUçI­AK]ݦ´I6‡.Ìïè"bµƒŸõ/;ÚDÿ{*=£‹N]¹$VCzÈõ}Ú¥ÖÊ£µÍŽÐ0ž±Žqi‹¢®½áðª“`*¿á?;)X@T£gÎR\¾fÆ×éA‡̓l-ß»›I[°z:/‹¢$4ŸÏÛç¿:cyŽrïž’fÑy«hMÊÈyÞ—% ‚(´W.îÖ¦‚ÅDÇg­ãC$‰2¬óÖ©G»Byguቘœ—^áí»K¢BsYq½ç§žùî<0êàpMÄêh'çí£»ßtÑŒÍxfθ(XšéÚUxÝÆøMYÍdM¯ßUb©¹¬ tÞ¢5/J8F¨?_¼¿jYÊŠg”yy§â ì€Pÿùö¢äÌœ5© ¥„ˆM¥ûw^0J0W ¿Ø]òÔüç^ÐP `¾R´·´ UA¯ öºË žUF¡ ê« X1ˆ7ŠŠŠ;ÅÅÅeggO›6 kAYÒÓÓ322Ö¯_Ÿ””ÔÊììì~ÖlÉÍÍ]·n.ê2$ÄÆÆKMR{ÝåJKèÄ?Vv™oÜhâ#ÆFÒÚ¸<ÎÖd÷¿è‹çJpAÇ$,Â*ô~ʆf‡H#@…¶º›·›Í.™QèG‰ ÑsÔróòMep55YFDªnµØDF=6DÇíÍ·jêM6Xµ!,bL€ÊU_qÅ?ÖÀ-7/W‘È„1:r· Gx¬À­kA¨,X[LNV«á €l«¿ZÝ®?yÊä¸0ÞTu³U¤ »ZÛ¹ÐøÉ“ÂXSÕ•ëmšˆ S&ÇËM·šœdKíõZ»>jâ”)£}·¯ß¶ÈJ?ÞÖlqÏ–mÆv0øk˜¾L (ó‚ È ¼µúâ¹sç/\ºÚÁcGû°²½¥Eô öQ0„Qú†+¬-Vç¨e £ð Pâä§bÑøú+«Km-­Ô7,XÇÂé‚Â|i›ÑN¾ »±]¤’Õhf þj¦OÈãÚ#‚<úmäÄñF°6U]oj¶j|8*9ÉR}ùC(•@%Ê”Âòž­;a˳ % C)*»dFÇ{ÆÞ §de‹D‰Î7@Qolý öf3c«b Àâ¸=Ê<‚ 2ÔZÏ𺠈ÖòÛMAºP%Ëó¼_Ô„ˆoË/²Q0²S¤¨è”ž%¼O ª®¥ÝƶYX¿5 ÷jy¬ÀK‡ ò8I½Â?ÄGln0‹DèÇ´Ý®owɨ,:ÚÛ­^=8g4þ¾¤­¶É*HÖ¦º6ð P3„÷ PÙëkë-œŸŸŠs¿&ŒæAûÒyNÀVÖ·>A¡ã"ëoÝ®(uJ”áU:¿°½Wñ.<:¬¦æúÅ: Xµ!ll¸{U1Âé457-ÊP?ÏÊgŒæ>M (ó‚ ˆw®M˜Ò]¥ÕaÿVø„Žõ í‘Z1qBÇwÖw\¢oç0@@ìä€Î€>02.ð®­8 ï3ÅÿŽÑƒ»M 8h wÜèXÿ,A™÷À‡e½úò,¿á»/ò!³—e„óCšå |öòö ë;1ëù—¿ýb¢öAY|Xå»|hÆs£ûØˈQ?=s隯ÿÛמ `ñÖ€ ÈÈ ÏA{6 yÉÊùq¤ö–Y ²$¶U<[ï¼cö…ìh¼Qm†M×ü¦Î‹kÊ?9¤ãó0—w¤¶Kù;œš¥ª¡ÉN—“)äï¾î|Àå}Xv…ÆÒÛ 'ût¢Eº»`ñ±UÚY>mI4Þñ2/Ïì+ÏKŸn­t0êÑO.]1ëÀ–ãM=ö’•ZÎìÙ?|1]èÌépæ£fihÅr>oy0„S)å‡PÞ‡e$ãùÓÒWŸ ?³ç¦ëÎsÎúÒÓõ>ƒDäK!ówEWö['ò+WΉ+ýä’…tîè¦T¨Ê?~¿ø)f}22&Ç)Ä|­xÿñj[×­]÷ïU?¿eý§f¯˜ãwnßõ1‰¡<0k»z(ÿd­shú:þ is“‚9IpºìM¥'zœU$¤Ì™ÄQÂpÄQSr°¸¢µ«}LÚ¢´ñ:Ñæd‡É¦°ŸÚVX+tVTd]µv† ™.äï+mí,Ö`ÊÛ÷Õ”Ýþ¯¾7WZ®Ÿ3gL Sܬrá_?‚ (ó=C¡¶ÊJ.mÉâ-•[*Ÿ»<²·Äew–yÄ—1$¯H¨ËóÜZý¤ÅóÎíøøºUFñÔÊåQm…ùô‚õç ÑáÒÍŽ*ÏÏ^žJ‹vm®±É¬.:%gé´½ÛJÜð’ñÌg›Ê'®}qÖÊgÛ/æmßØäb”zìé`ôçó@^õó[©åܾ#“^›yüíEf ]ü’•©a j=êÌèâ.Ž¨Ü•{ M`õ1©‹ÓÂù“g}—äD•ïØrÐ$•“E·î©´¹ëL=?CyrÃ;UUäü•±¤s´æ³MåW-tíÜ~¥×îÈ`ÊÛ÷Õ”Ýþ¯¾wWÁQ_íJ‰6pU"þù#‚2ßó,8@Å{7+Š¨‚fÍN Ó0²(¸\ü(ýÀÀMKl;´íº{edÙ~óøþ Y½Iœ6Pí¼jïÎmlJø¥¼­5ž‘Érãh~𪔨‹Ÿ^stK&Öm9Xm§ ;ÍojÂ[¯úÂY]x¨Â,¹•ªªÜ:Õ‡AË<7*9É~x{y›[“$óÕâ¢èøy­æ‚g$; 7_2yKh)-(Ž^™x­¸IMÕmA'Ÿª[LíŽêýUYÓ鯼ý\ýÁàÅÕð*Hv“S¯c ÿúA™¿Cvy8Z½Ù¶€ œ¹lþlÞ'GŒ. Àê'­\Üy’ÕùKµ-Ýn½bÛ­—²—l¨D CíúmÛPiëîƒd®5©‚ÕÌ5G×Q¡¥ºÞ~oK5yïUŸÐ.‹T–‡d¥(V(Ýþ¢»> -7.ÏdqVëo²|õ“Ý”UñÍjÀ=¸ÑR²u³OHDtìÌÁzÒZ~¸à‹:ç5ž¾ÊÛïÕTmxsõ¸ „0„J¸÷2Läææb% ­Ì³¾±±âõ=^¨¬_|´ñжrcÇ®Ë)RE×ÙÈ&(.ßêP/Îw´¿¢ùîl¤ö›*PË@{ç]Y²åàP-Sß&wÝûCýœFû`oÜ^{u¿‘¨zôôYñ>Rý¹c¢÷^5³“e5JÏûG(ê==‘–êºêúOó«ûn¢Ðk©¥½îêùº«ç£Ïy.åÖ†uÃ:^ÝÿÕÔ5‚«ÏêU¶³„ûÈÐSYY¹jÕ*¬ä‘Âë÷’mtê⸚¢ ‹QªìhýCÜ«'£‹N]¹$VÓ¹£‘Øtæ‚>=uœû<£‰š=g¼ŠÒÞdþÆ 1F×ÍIj­±é :_š—ZÎsÌš7%HÙQ¹œ~Ìää±5¦Ÿ³&u´ªó¤Ö ¥®n'Ùº0?¾£O`ˆˆ ‚ *ú¿úƒ±;WŸÑ†G’ªm(ó‚|É£y6 yAjôXòÜÚ'$"·]>¸ýLçÛtžùÐ<0êàpMÄêh'çí£»ßtPsYAé¼Ek^ ”pŒP¾xÕ²”Ï(óòNEÍ—öìçç.Y3›]NGËå#Gn¦ð½ŠÜ¹KÌ’xÃénï9 õŸo/JÎÌY“ÊPJˆØTºç£Ôáõ´Åó'øFwLäî1»_Ÿðª¿ßÍas8± IDATØŒg挋‚¥™®]…×mŒß”é1ÑLÖôú]%žÈ]:)‚s–çÕÜÛ¹l-ß»›I[°z:/‹¢$4ŸÏÛç¿:cyŽrïž’fÑy«hMÊÈyÞ—% ‚(´W.îÖ¦‚ÅDÇg­ãC$‰2¬óÖ©GºM=sVžˆÉyéÞn±»$*4—×˃.ïWÿþíöõ½¹ ÀøÄMæ.íéeú£Ÿ¹hFϨƒÂµQk"²ëvñg‡«x“@äq†TTTÜq(...;;{Ú´iÎ uüs9ìîÍ—lwû§›øìê1Ç?ȯyðï?õíÕ½£ˆx浃?ÿŸ÷1Ž|xðáó¾:»6wû%Ëãÿp>===##cýúõIIIý§ÌÎή¬¬ìëlnnîºuëúI€ ÈcÄÑÏ•< ‹ws~ñéÕÞ_2§–ò‚/hDÿHyuÏ0Úƒ±xÏiÔø‡ÛÒüÃœ圀‡ È—çÆ÷° 3ºØ9 §â#ÕÚ[dêC%Ó¹}Ÿ?r^Ý#²éÔ‡ÇÖö°NjÀj@eþ [*l}äÆ M¯Aä>ÃW¬AA™GAeAA”yAAPæA¹8¬AîæðáÃX òˆ“žž>ü2χe}cqëÆ÷N˜†iÍ>dvö„ê݇n C—å |öòö ë;qÞÂÙñÊï~Pj} Vy‡Å.š±,þÊÎC·z_R‘óŸ4wn¢Ç°Œ«ö‹‚¢ŠÖ/ù–ô¯¼ò ÞFä‘¥®®Îl6߿̳ÉKVÎ#µ·Ì•%±­²èàÙz盹ȎÆÕa¸ŠÁúM×”rH Æça.ïHm—òw85KUC“:.'SÈß}Ýù€Ëû°ì ¥·,Nöÿ¨Û 0º ‹_Ø‘{Ã&ç7yñâ§Í[× x+A䱦O™—Œgöç¥O·V:õè'—®˜u`Ëñ¦!ŽÔrfÏþá‹éBgN‡35í±ƒñyxËû€!œJ)?„ò>,» ÏŸ–¾údø™=7ï èwãÀÑî]DÓ¥cUÏN âko£Î#22eþ®èÊ~ëD~åÊ9q¥Ÿ\rïEëÙKL©P•ü~ñRÌú&ddL0(8ŽS(ˆùZñþãÕ¶®[»2x꜌IþŒ(Ér{Åñã—§+«Nåí9Ýc+vEXòxÛùâ;ö¾eÔaI©qB)Éxñð‘ÒÆŽA6 yÑüÄÑÁÎÛåœ:况z•NÑxtÛÞ²vy@Ÿû÷ªŸß²þS³WÌñ;·ïú˜ÄPŽµ]=”rˆÄçýÒæ&s’àtÙ›JOô\®U2gjG ÃGMÉÁâîcÍŒ>&mQÚxhsH²ÃdSØOm+¬:+*²c+¿;6óƒÁ•·ï«?(»ý_}o®µ\?gΘ¦¸Yu‡Î‹Í¥ºW«!€˜/ãâ÷‚|id¤¶ÊJ.mÉâ-•[*Ÿ»<²·Äew–yÄ—1$¯H¨ËóÜZý¤ÅóÎíøøºUFñÔÊåQm…ùô›9Ct¸t󄣇ÊóÁ³—§Ò¢]›kl2«‹NÉY:mï¶÷0¬d<óÙ¦ò‰k_œµòÙö‹yÛ76¹¥^#{î×ýùî$€!^–ýó™sfô1úÈøŽ>stŸ×u£4BQãjý–Ȩk߈n|¬.»¬^±®ã¬¾N¼ï.xI/ÐSãA„G#‹BH"3û´U Z`|àó’yœä#ó;W29>ûOdˆŸ—üV­µ²â¨ü¬ÙqT$¥‡ts†^J?¨³òø–f1ÃbgŸõ&ÞªLsm  Cã|_þÚ<û©M§ÖL±Ëj ×­Zv¶D–aØ•ð!ö¡ß<›«6]¿Öʵ_J%Iyæ(!Á'ÇBÓ‹ºÆëW<·Êäaû÷É|ÅØ;eUù³á•’¯ÅêëÔûnµ†+Þ_ †c,Í,ÇÇdeú©î”uO‚=Ÿ›Ì›Â¨žbTÂ+B®}xC©u¬áZ-Ë™µD´7÷åÌkMœMRî؇_CO¼5ñ}„8š˜ri£–ñõâ=3;öûyY´SîŽÊ.[õ±‘¨ poB„'­iyüBK¹nÕ±Û‡Û90£ô¤TæÍÕØg":õ°Zs»T½ˆpc\±5N w·w·>F¸0"÷TÒàOëú¢˜sï»å£ð>!òá›ÞZŽçúîË9èÙrëþkÐx~/¸| .”'g…Ôt]ˆR³ž’nÓOŠäÉù9aØL˜<ÚÜ&NMÞ1}÷NL å³ìB2ßÛËʶ‰æÉNªj‡"&Ê ÉýÃrÊ8e¥YÛXg†q»¾KxÿãËuî}W¼€pÏðÝœŽâ·ßñ‚Rl%‡“òÎØeŸðÜl«xñ|„‚A€O¬««kÞGáááÙÙÙñññkg… âT.q÷J‡éCûD‘_l«ÿgé€uÍÛfq«–Wvìûó¾þúC½vo¯d@ú·‰CW ;ŒŸþª|jjªB¡¸téRll¬ó;³³³U*ÕbW¯^½zñâÅo¨ªª‚Sð`#³à)xó~Ñ/Z7BêŽWD$Û½ðKæ¬QYþŒ•y“ʪeƒ ·J´uÅM ñëÛÓ¤ÔÓr¥þóÀç3ð­WÁ¸(,íHüfBN7Ü«_Déñ–û6œUË„oøñÐÛÖêmÃ÷Р€Ì¯ŒQUq]µÑšccZ¾BÈ< ó€Ì2È<+)ó¤æ¿KðZ½é¹5ñ„"`EÇqÇæU¯¯SˆM‘™§¿ûóùháZ•¸^õ]•rI?Å)EàbYtpm_dþ—9wîÜ·ßæ§ìÀOggÚçädž 4B,CSzU̓ç˼d.Œy¤Wm´­–}„WLzøhéÓ-À›W¹¾K@ë;JoZ<ŽóWæëá¹m¥w{,k\ßõ*×6Òþ&#+NzyNŽ„ÙŸ‚WØv¦õöσS BÜ­‰'Äho4¿ƒóø}Ê<­m¾_JÒ·¯«Ì!„ ?™Pq­~ô½d´®¹¸lõb:¿ý{Qó展=#Ö›W·¾k Æáó˜u¨ïz•‹hmkýí€æâþr$PÚ–j­ãëHç+)U姅·ÿÖ1‹ÖžKŒ'ðæ+ù¡nž›v);%\‡Ãåb†×ueõjÓìÐÎóISDIqŠf˜‰®úÞмT^_CIqÓ{©Ø¹þq¡¦Öºy¹oq¬"9\‚±,Âh틪êöÇ"áwôpt ¯¥âoE–˜ô/·‹ù"îHí{Ì’6;·Êɳ„4&ûdšWËýžmÑ~$Â9„©ûaéÓ:Ÿ”îJ9ëË¡mëÔhû“÷kåzïJJ‹ÙÂa1œƒ™Ôu½›mC\’r4%TD™Ì4c7q§nTÙf*È‘Êo^2?ä^}÷¾[å:÷¾+^`=-E¼?·¿Ï™~âûÂ]6!ø\d!DëU*NÊf²ÃhE!ƨ*¿¦BdÀ¡¼ …nî|p«Ó.¾¸$îdªl¸Ä>´â⨬tï–›¿ôL2ȾÌÏ ÖWþåvÛä<ã$òºÿ‰ù=•'}ó’Ùš¢+&!B$OÊ=ïFãô2,­m¾ó«2òìù„ü¯'^”þÁpfóRV9y–ֵܯŽú>=¨þÇë5ᢈœüdÿŸË‡ÜÖ \q$K¦*ºZ¡§"Ä!ÉY)äSÇUÏèœÜ`åÍkÆ)„)ÎÌÍd¯«LÓmÆ“Vðžþô÷>3Bñƒç‡a3«5w~UF~sÄz«ðÕ‚Ów껸÷Ý*×¹÷]ó‚Y£¶&É%œ¾…SÌÒØcù£|Æ«þñ“£>™GŒÍŒø¤k“0þ–] ‰Ñþ8CÙ¬Vr³¸ƒ!„8[â£õoôL2ÓËýõem»2’8¡ÀÒ=57œÃ„aI%×ì+´±·¶Ô÷›¤à·_›çÜF ×\{ žbBŒÅ`t¥%\µj1,êʇ]zZ©ú”“1^$r[æ9›ãb§ª •úiM¢ Ýu5òˆt»Vs|÷Å™+¯tŒÛ˦k/¯“çÇù¼®¥BˆWë·DF…XûFtãfuÙeõŠugõuâ}wpÁûKzž·"D8Yx.«{~óï­¥~{Ò‡#ï–tLÀÿæø|d'ùÈìÒž$ŽÏþâç%¿Uk­,B„8*?k6`Ié!Ýœ¡—Òꬼ¾†¥Y Ç0„ØÙg½‰·*Ó\hÃÐ8ßW€¿6Ï~jÓ©5SË Å\·jQØÙY†Y‘HûÐožÍÕG›®_kµo'„RI`R^Á9JHðÉ1Ðô↮ñúÏ­2yØþ}2_1öNYUþlزBg±ú:õ¾[­áŠ÷—ð†áK;ïÄÌÔðóûÄ_ïÛªª‚…{>™'6……Q=Å. {„W„\ûð†RëXõZ(–;;2k‰hoîËA‡zq6J¹c DVoM|!ŽfC*Ú¨e|ý„¸FÏÌŽý~^í”»Q—ËV}l$*Ü›áIkZ¿ÐR®[5FìöávÌ(=)•ys5ö™ˆN=¬ÖÜ.U/"ÜW,dÃÝ­ÃÝ­.ŒÈ=•4øSÅ0µš}ʹ÷ÝòÑ xŸùðMo Knêd,&$àa0Dðiãò«Á¸Pžœ>PÓet!JeÌzJºUDL?)’'çç„yÌŒ˜Ôhs›85yÇôuÜ#81-”ϲ É|o/+Û&šc$;©ªŠ<š(óÀíÑ£Á¯j{ÍËV}$‚ ”Œ”¯öˆ&ŒÔ²¬jz.PÜ)™žža©»f^š§u-Í é{¶ÌÈG¼mwÜvG‹ BsÏ$òg. %BÖ:Gàh“YäïE:æYH€Ðý—Å{ßrWÀû¸0 ëëÕ ó¼€}_Fùò§¿ãùÆÚ©i~ûìø½Fó„w\F²|;vêì4BaŒþåƒÂæ™·éìû¡I„ | ÉýÃrÊ8e¥YÛXg†q»¾KxÿãËuî}W¼€pÏðÝœŽâ¶ßYÞªF‚¾:~VJ²4ƒ©·þV ¨<Ÿ-\«׫¾«R.é§8¥\"‹.Úyü¿ÿ_ÿ²“ß>ÎδÏÉ?Ž h„X†¦ôªšÏ5–yÉ\óH¯Ú¸jÉ: ¯˜ôðÑÒ§+Z€;6¯r}—€Öw”Þ´x_!„ç´•Þí±¬q}׫\ÛHû›Œ¬8éå99æÃJŒ¹÷€›È~*¿áììlÈX¶ÌÓÚæû5¡$}ûºÊŒB¸ ðÀñ“ ×êGßKûAëš‹ËV/¦óÛ¿5_[Ù3bݱyuë»Æ`>Y‡ú®W¹ˆÖ¶6Ñßh.î_8G×ÿ^Ûãۣ€OêÿYKY#€Ì/ÅN >)U姅·ÿÖ1‹ÖžKŒ'ðæ+ù¡nž›v);%\‡Ãåb†×ueõjÓìÐÎóISDIqŠf˜‰®úÞмT^_CIqÓ{©Ø¹þq¡¦Öºy¹oq¬"9\‚±,Âh틪êöÇ"áwôpt ¯¥âoE–˜ô/·‹ù"îHí{Ì’6;·Êɳ„4&ûdšWËýžmÑ~$Â9„©ûaéÓ:Ÿ”îJ9ëË¡mëÔhû“÷kåzïJJ‹ÙÂa1œƒ™Ôu½›mC\’r4%TD™Ì4c7q§nTÙf*È‘Êo^2?ä^}÷¾[å:÷¾+^`=-E¼?·¿oçlŽKä7¿±¢0øÌd!DëU*NÊf²ÃhE!ƨ*¿¦BdÀ¡¼ …nî|p«Ó.¾¸$îdªl¸Ä>´â⨬tï–›¿ôL2ȾÌÏ ÖWþåvÛä<ã$òºÿ‰ù=•'}ó’Ùš¢+&!B$OÊ=ïFãô2,­m¾ó«2òìù„ü¯'^”þÁpfóRV9y–ֵܯŽú>=¨þÇë5ᢈœüdÿŸË‡Ü^uÆEG²dª¢«z !B’œ•@>u\õŒÎÉ VÞ¼ö`œB‘ÒèÌÜLöz±Ê4Ýf<ùaïéOï3#„?èp~6³ZsçWeä7G¬· _-8q§¾‹{ß­r{ß5/˜5jk’\Âéû -.‰úÒ§£¢a !ø·<Ÿ£Ì#ÆfF|ÒµÅLŒ¿eWBb´¿ÎP6«•Ü,nã dE!ΖøhýÃ=“Ìô2A}Yۮ̅$Nè#°tOÍ ç0aXR@GÉõûÊmì­-õý&)øÅí×æ9·QÃ5ר§X„c1]i W­Z ‹ºòa—žVª>ådŒ‰Ü–yÎæ¸Ø©ªB¥~Z“hCw]<"Ý®Õß}qæÊ+ãvŲéÚËëäùq>¯ëFi„¢ÆÕú-‘Q!Ö¾Ýø„Y]vY½b]ÇY}xß\ðþ’^ §Æ-‚ŽFæ}·(ìËྺ"H_Àç+ó8ÉGæw® ƒŸý'2ÄÏK~«ÖZY„qT~ÖÌEB$¥‡ts†^J?¨³òø–f1ÃbgŸõ&ÞªLsm  Cã|_þÚ<û©M§ÖL-o•ëV- ;["Ë0+²…‹ûÐožÍÕG›®_kµo'„RI`R^Á9jEðÉ1Ðô↮ñúÏ­2yØþ}2_1öNYUþlزBg±ú:õ¾[­áŠ÷—ð†áK؉Iï¾É‘ovN7»,€Íˆ¹û\©ƒøldžØFõ» ^ríÃJ­c ×j¡XîìȬ%¢½¹/êÅÙ(åŽ}ø5ôÄ[ßGˆ£‰™Q™6j_?!®Ñ3³c¿Ÿ—E;ånæ²U‰ ÷&DxÒš–Ç/´”ëV»}¸3JOJeÞ\}&¢S«5·KÕ‹7Æ YãÄpwëpwëc„ #rO% þT1L­fŸrî}·|´Þ'D>|Ó[ÃêmUÿðÿ9þà‡}‹ßàÓÇåíĸPžœ>PÓet!JeÌzJºUDL?)’'çç„y`3aòhs›85yÇôuÜ#81-”ϲ É|o/+Û&šc$;©ªŠ<š(óÀíÑ£Á¯j{ÍËV}$‚ ”Œ”¯öˆ&ŒÔ²¬jz.PÜ)™žža©»f^š§u-Í é{¶ðËoÛ·ÝÑb‚ÐÜ3Éü™‹B‰µÎ‘.Údù{‘Ž9D tƒ¹sï»Sî xa}½zp>óhžðŽËH–oÇNý‚F!ŒÑ¿|PØ<ó6}?4‰po€‡¬@na‘åMíݪ~3B¬¡³¼=ýè™s‹qp›¦µ®¬ïDÒÉc¼’’&-…CGqy(çL"AY-fÝËêêþ$rA‘kéÀs"$MsÞs¶iÖÄÌ=“Œ³,†Q£íe·Ú´´Ãêø¬Ã;½$ŽÜïmãvjóV9{óØ®8–¶#?h-ªì1á^{2RCäxæ^MQ£=rçùEÉ8eiÍÀò–Ì™Iå½»xJFÁ^’¡(Ú6ÖZr_Z ÈËåÝ+n£,ƒ5El¬"÷ô&CŒÍFÙ&úšêËÚ¬Í8ΆfžMóÄhšÅ Ë`íê9[Ï,êÊ'!¹¸@N§¬4kë¬Ó0n×w ï|¹Î½ïŠî¾›ÓQ<²øT‹ÆÉØäÏdÆÞiÖÂ|€O¬««kÞGáááÙÙÙñññkg… âT.q÷J‡éCûD‘_l«ÿgé€uÍÛfq«–Wvìûó¾þúC=èÆ:B¤›8tµ°ãw°Ñ.55U¡P\ºtéâÅ‹K¾7ÿ7TUU]¸pú lX†‡‡ ƒó_ô‹–ÆpÇ+"’í^ø%sÖ¨,ÆÊ¼É eÕ²Á…[%Úºâ&ÐøõíiÒêi¹6Óð |ëU0. K;¿™ƒ†Óà ÷jÆQ@z¼åþ£ gÕ2aÆ~üô¶õ†zÛðð-4 ókcTU\Wm´æؘVÀG†¯Ð2È< ó€Ì2ÀJÊ<éŸùÇï¼Voº@nM<¡XÑãqܱyÕëëbSdæéïþ|>Z¸V%®W}W¥\ÒOqJ¸XÞöüïÿÏoÏLsúØ^ |ê8;Ó>'ÿp864h bšÒ«j<×Xæ%saÌ#½j£mµì#¼bÒÃGKŸ®hîؼÊõ]ZßQzÓâqœ¿2_'Ï=h+½ÛcYãú®W¹¶‘ö7YqÒËOÎ=g{ûø·kmFÆ~ÿ2Ok›ïׄ’ôíë*3Bá‚ÀÇO&T\«}/í­k..[½˜Îoÿ^Ô|yleψuÇæÕ­ïƒqøk™ÿ ºš|RªÊO oÿ­c:ܱçã ¼ùÊ_~¨›'ÅĦ] ÅN —Ãáp¹˜áu]Y½Ú4;´ó|cÒQRœ¢f¢«¾74/•××PRÜô^*v®\¨©µn^x… ücÉáŒeFk_TU·8︣‡£}-+²Ä¤¹]ÌqGjoÜëœ`–´Ù¹UNž%¤1Ù'Ó¼Zî÷l‹ö#Î!LÝKŸ®Ðø¤tWÊ¡X_m³X§V¯s6 IDATFÛŸ¼\+×{WRZÌ‹áÌ<Ðø ®ëÝlâ┣)¡"Êd¦ó¸‰;Õp£rÈ6ÓPAŽT~ó’ù!÷껸÷Ý*×¹÷]ñkìi1(âý¹ý} èK÷ÔÜp†%t”\°¯ ÐÆÞÚRßo’‚_Ü~mžs5\sízŠE1ƒÑ•–pժۨ+vèi¥êSNÆx‘Èm™çlŽ‹ª*Tê§5‰6t×ÕÈ#ÒíZÍñÝg®¼Ò1nW,›®½¼Nžçóºn”F!j\­ßbíÑO˜Õe—Õ+ÖuœÕ׉÷ÝÁï/ézjÜ"ˆáhäƒîmî3ÍÌR_Ö½þú‹­äðè<ŸÌã$™ß¹’Å“ã³ÿD†øyÉoÕZ+‹!ŽÊÏš¹Hˆ¤ônÎèIéuVÞ_ÃÒ,†cBìì³ÞÄ[•i® ´ahœï+À_›g?µéÔš©åí¤rݪEagKdfE6rbúͳ¹úhÓõk­öÍâ„P* LÊ+80G >9& š^ÜÐ5^¿â¹U&Û¿Oæ+ÆÞ)«ÊŸ [V¨ó,V_§Þw«5\ñþ^À0cé¥;1m6°¼¡tük“¶Z(–;;2k‰hoîËA‡zq6J¹c Œ´oM|!Ž&fFeÚ¨e|ý„¸FÏÌŽý~^í”»9Ä]¶êc#QAàÞ„OZÓòø…–rݪ1b··s`FéI©Ì›«±ÏDtêaµæv©záƸb!kœînîn}ŒpaDÁŸ*†©ÕìSνï–VÀû„ȇozkXrS'oó6ÎhCŸ6.G+¸Pžœ>PÓåÊëFŒYOI·Šˆé'Eòäüœ0l&Lmn§&Ž{'¦…òYv!™ïíeeÛDsŒd'UµC‘Ge¸=ÚÝ‘r4øUm¯ÙÝ–pÙªD”’‘òÕÑ„‘Z–UMÏŠƒ;%Óó1Â3,5c×ÌKó´®å±9!}Ïž£q9âm»ã¶;ZLš{&9?sQ(²Ö9G›Ì"/Ò1'ÈB„îǯνïN¹+à}\„õõê?yÐì3Gwûp½ý«$Ö–~Å`ãDó„w\F²|;vêì4BaŒþåƒÂæ™·éìû¡I„ | ÉýÃrÊ8e¥YÛXg†q»¾KxÿãËuî}W¼€pÏðÝœŽâ‘§Z¦W÷Kñ©ßü71¢ŒÑwVw!øÄÁºººæ}ž¿vV"Nåw¯t˜>´OùuÁ¶ú–¬ý{΋[µ|¸²cߟ÷}ô×êµ ®d@ú·‰CW ;ŒŸ¾€§¦¦*ŠK—.]¼xQ¥R-vÛÕ«W?ªª .@Ÿ€ Ëðð°Á`pþ‹~ÑÒ¸¶q¼""Ùî…_2gÊòg¬Ì›ÜPV-\¸U¢­+n_ßž&  ž–+¤ðù |ëU0. K;¿™ƒ†Óà ÷jÆQ@z¼åþ£ gÕ2aÆ~üô¶õ†zÛðð-4 ókcTU\Wm´æؘVÀG†¯Ð2È< ó€Ì2ÀJÊ<éŸùÇï¼Voº@nM<¡XÑãqܱyÕëëbSdæéïþ|>Z¸V%®W}W¥\ÒOqJè<‹!>p¬àÌ邳gó“x0DðiãìLûœüÃáØРFˆehJ¯ªyð\c™—Ì…1ôª«–“›ðŠI-}º¢¸có*×w h}GéM‹ÇqþÊ| <÷ ­ôneë»^åÚFÚßddÅI/ÏÉ‘ð\¿ÄüLŸÎ’ß~}kaatà÷,ó´¶ù~M(Iß¾®2#„.]¡ñIé®”C±¾Úf±N¶?yÿ¸V®÷®¤´˜-Ã9˜y ñA]×»Ù6ÄÅ!)GSBE”ÉL3æqwªáFåm¦¡‚©üæ%óCîÕwqï»U®sï»âÖØÓbPÄûsûû>Ðy®\ º®4€ÏRæB´^¥â¤l&;ŒV„bŒªòk*DÊ Zèæη:íâ‹KâN¦Ê†KìC+.ŽÊJ÷n¹ùKÏ$ƒpìËü¼`}å_n·MÎ3N" ûŸ˜ßSyÒ71/™­)º2`b"Dò¤Üãñ÷n4N/ÃÒÚæ;¿*#ÏžOÈÿzâEIáÏ£Vœ'ö`ìC·3›—²Êɳ´®å~uÔ÷éAõ?^¯1ÐEää'ûÿ\>äöª3.Š8’%S]­ÐSâ䬔ò©ãªgtNn°òæµãBˆ”Fgæf²×‹U¦é6ãÉ+xOú{Ÿ!„øA‡óð™Õš;¿*#¿9b½UøjÁéˆ;õ]Üûn•ëÜû®yÁ¬Q[“äNßüd´¸p‹@Óc <ûeˆˆ1k•µ[4°tÀg$󈱙ŸtmWÆß²+!1Úßg(›ÕJn·q²"„gK|´þážIfz™ ¿¾¬mWæB'ôXº§æ†s˜0,) £äú€}e€6öÖ–ú~“üâökóœÛ¨áškÔS,Bˆ±Œ®´„«V-†E]ù°Ë@O+UŸr2Æ‹DnË'™ÇI>2¿seØãøì?‘!~^ò[µÖÊ"Dˆ£ò³f.")=¤›3ôRúAu¡=Í,Íb8†!ÄÎ>ëM¼U™æÚ@†Æù¾üµyöS›N­™Z^ æºU‹ÂΖÈ2̊Ą؇~ól®>ÚtýZ«}³8!”J“ò Ì+‚OŽ „¦7tׯxn•ÉÃöï“ùŠ±wʪògÖê<‹Õש÷Ýj W¼¿„0 ÇXzNÌšM‚Í“ÿhÒXB¦¾ÇObN†yv4¾à3‘ybSXÕSìB€JxEȵo(µŽ5\«…b¹³#³–ˆöæ¾t¨gS ”;öá×ÐoM|!Žf#*Ú¨e|ý„¸FÏÌŽý~^í”»ƒ±ËV}l$*Ü›áIkZ¿ÐR®[5FìöávÌ(=)•ys5ö™ˆN=¬ÖÜ.U/"ÜW,dÃÝ­ÃÝ­.ŒÈ=•4øSÅ0µš}ʹ÷ÝòÑ xŸùðMo nêdLoºÇ<,ôì“b 8W€O—‡1\(OÎ ¨é2º¥2f=%Ý*"¦ŸÉ“ósÂ<°™0y´¹Mœš¼cú:ÊgÙ…d¾·—•mÍ1’TÕEM”yàöhwGÊÑàWµ½fw[Âe«>APJFÊW{DFjYV5=(î”LÏÇÏ°ÔŒ]3/ÍÓº–Çæ„ô=[xŽÆ刷íŽÛîh1Ahî™ä@þÌE¡DÈZçm2‹ü½HÇœ@" º/lνïN¹+à}\„õõêé…ÚºMåóeìf!„8ÞQ{E¯»õÊð;æ ︌dùvìÔÙ/h„ÂýË…Í3oÓÙ÷C“øxÈ äYÞÔÞ­ê7#Ä:ËÛÓž9‡±·iZëÊúN$<Æ+)iÒRˆ1t—‘‡rÎ$”ÕbÖ½¬®îO"¹–<'BÒ4ç=g›æQaMÜÁÜ3É8Ëb5Ú^v«MK;¬ŽÏ:¼ÓKèØÈýÞ6n§6/a•³g1íŠci;‚ÑñƒÖ¢Êîµ'#5DŽgîÕ5Ú#wž_”ŒcQ–Ö ,oÉœ™TÞ»‹§dì%Š¢mc­%÷¥Š¼\Þ½âÆ1Ê2XSÄÆ*rOo"0ÄØl”m¢¯©Î±¬ÍÚŒãlhæÙ4OŒ¦Yœ° 6ܪž³õÌ¢®|’û‡ ä”qÊJ³¶±Î: ãv}—ðþÇ—ëÜû®xážá»9Å# Nµ}Û½'û2N|ãÁ!9Ôð³’*- CŸ6XWW×¼ÂÃó³³ããã×Î AÄ©\âî•Ó‡ö‰"¿.ØVÿÏÒëš·ÍâV-®ìØ÷ç}ýõ‡zPŽu„ Hÿ6qèja‡ñÓÒSSS Å¥K—.^¼¨R©»íêÕ«wCUUÕ…  ÏÀ†exxØ`08ÿE¿hiÜÿ{äxED²Ý ¿dΕåÏX™7¹¡¬Z6¸p«D[WÜ¿¾=M@=-Wa%€Ïgà[¯‚qQXÚ‘øÍ„0œn¸W3¾ˆÒã-÷m8«– 3Þðã? ·­7ÔÛ†‡o¡™_£ªâºj£5ÇÆ´ >2|…&y@æ™dyVRæIÿÌ?~—àµzÓrkâ EÀŠãŽÍ«^_§›"3O÷çóѵ*q½ê»*å’~ŠSŠÀ…³èðÃÎüùœ+˜á_þÇŸ By0Bð‰ãìLûœüÃáØРFˆehJ¯ªyð\c™—Ì…1ôª¶Õ²ðŠI-}º¢¸có*×w h}GéM‹ÇqþÊ| <÷ ­ôneë»^åÚFÚßddÅI/ÏÉ‘à€z÷²öö‹&½HŽ_Ú×8´€ß­ÌÓÚæû5¡$}ûºÊŒB¸ ðÀñ“ ×êGßKûAëš‹ËV/¦óÛ¿5_[ÙáÖ›W·¾k Æáó˜u¨ïz•‹hmkýí€æâþù9¨‘çMs~Ò_í«Z F~¯2ÿAt55ø¤T•ŸÞþ[Çt.Z{.1žÀ›¯ü出yRLlÚ¥Pì”p9—‹^וիM³C;Ï7&M%Å)ša&ºê{CóRy} %ÅMï¥bçúÇ…šZëæå¾Åþ±Šäp Ʋ£µ/ªªÛG‹ „wÜÑÃѾ–Š¿YbÒ¿Ü.拸#µ7îuN0KÚìÜ*'ÏÒ˜ì“i^-÷{¶Eû‘ç¦î‡¥OWè@|Rº+åP¬/‡¶Y¬S£íOÞ?®•ë½+)-f ‡Åpfh|P×õn¶ qqHÊÑ”Pe2ÓŒyÜÄj¸Q9d›i¨ G*¿yÉü{õ]Üûn•ëÜû®x5ö´ñþÜþ>g¹ˆM;õ]Om0@ðÙÈZÿðFÏ$3½LÐ__Ö¶+s¡ÑVè#°tOÍ ç0aXR@GÉõûÊmì­-õý&)øÅí×æ9·QÃ5ר§X„c1]i W­Z ‹ºòa×ôscŸr2Æ‹DnË™¯{§¬*6¼RúµX}zß­ÖpÅûKxÃpŒ¥ubÌCBwÝ›‚Ñ€ÏLæ‰MaaTO± *á!×>¼¡Ô:Öp­ŠåÎŽÌZ"Ú›ûrС^œMRî؇_CO¼5ñ}„8š˜•i£–ñõâ=3;öûyY´SîæwÙªD{"ÜÎ¥'¥2o®Æ>Ñ©‡ÕšÛ¥êE„㊅¬qb¸»u¸»õ1Â…¹§’ª^Õ­eνï–VÀû„ȇozël=? {Un‚Á€ß.¿—Œ åÉYá5]F¢TƬ§¤[EÄô“"yr~N˜6&6·‰S“wL_Ç=‚ÓBù,»Ì÷ö²²m¢9F²“ªÚ¡È£‰2ÜíîH9üª¶×ìnK¸lÕG"JÉHùjhÂH-˪¦çÅÁ’éùá–š±kæ¥yZ×òØœ¾g ÏѸñ¶ÝqÛ-&Í=“ÈŸ¹(”Yë£Mf‘¿é˜Hd!B÷ßTwî}wÊ]ï〠¬¯WO;Qùpòµz’…Á€ßu4OxÇe$Ë·c§Î~A#„Æè_>(lžy›Î¾šD¸À7ÀCV ·°Èò¦önU¿!ÖÐYÞž~ôÌ9ŒÅ8¸MÓZWÖw"éä1^II“–BŒ¡£¸Œ<”s&‘ ¬³îeuu¹ Èµtà9’¦9ï9Û4 kâæžIÆYèÑö²[mZÚau|Öá^’@ÇFî÷¶q;µy «œ=‹ylWKÛŒŽ´Uö˜p¯=©!rZútE pÇæU®ïÐúŽÒ›ãü•ù:AxîA[éÝË×w½Êµ´¿ÉÈŠ“^ž“#aNÿ";ÆX~õò˜ !žÿW'Dßl›€óø}Ê<­m¾_JÒ·¯«Ì!„ ?™Pq­~ô½´´®¹¸lõb:¿ý{Qó展=#Ö›W·¾k Æáó˜u¨ïz•‹hmkýí€æâþr$p7¡—wƦgMG7÷+oNÛ„ ~—2ÿAt55ø¤T•ŸÞþ[Çt.Z{.1žÀ›¯ü出yRLlÚ¥Pì”p9—‹^וիM³C;Ï7&M%Å)ša&ºê{CóRy} %ÅMï¥bçúÇ…šZëæå¾Åþ±Šäp Ʋ£µ/ªªÛG‹ „wÜÑÃѾ–Š¿YbÒ¿Ü.拸#µ7îuN0KÚìÜ*'ÏÒ˜ì“i^-÷{¶Eû‘ç¦î‡¥OWè@|Rº+åP¬/‡¶Y¬S£íOÞ?®•ë½+)-f ‡Åpfh|P×õn¶ qqHÊÑ”Pe2ÓŒyÜÄj¸Q9d›i¨ G*¿yÉü{õ]Üûn•ëÜû®x5ö´ñþÜþ¾ùúmza8+ë©ï71„çöX™îå Œ|.2¢õ*'e3Ùa´"„cT•_S!2àP^ÐB7w>¸Õi_\w2U6\bZqqTVºwËÍ_z&„ d_æçë+ÿr»mržqyÝÿÄüžÊ“¾‰yÉlMÑ•ƒ!’'å¿w£qz–Ö6ßùUyö|Bþ×/J µâ<±cŸ`8³y)«œ`_ ½µ¥¾ß$¿¸ýÚ<ç6j¸æÚõ‹b,£+-áªU‹aQW>ì2ÐÓJÕ§œŒñ"‘Û2ÏÙ;UU¨ÔOkm讫‘G¤Ûµšã»/Î\y¥cÜ®X6]{yÚtýZ«}³8!”J“ò ÌQB‚OŽ „¦7tׯxn•ÉÃöï“ùŠ±wʪògÖê<‹Õש÷Ýj W¼¿„0 ÇXúÃNŒoÚ¹‡yr³Û@#„¨ñÎêæ°ìao»t€ÏEæ‰MaaTO± *á!×>¼¡Ô:Öp­ŠåÎŽÌZ"Ú›ûrС^œMRî؇_CO¼5ñ}„8šÝîLµŒ¯Ÿ×è™Ù±ßÏË¢rwG´ËV}l$*Ü›áIkZ¿ÐR®[5FìöávÌ(=)•ys5ö™ˆN=¬ÖÜ.U/"ÜW,dÃÝ­ÃÝ­.ŒÈ=•4øSÅ0µš}ʹ÷ÝòÑ xŸùðMo lêÄ.FÓ³šÎP4!"æN1>A\~5Ê“³Âjº\‰n³ž’nÓOŠäÉù9aØL˜<ÚÜ&NMÞ1}÷NL å³ìB2ßÛËʶ‰æÉNª‚L_ IDATj‡"&Êì/5â)Gƒ_ÕöšÝm —­úHA))_íM©eYÕô\ 8¸S2=#<ÃR3vͼ4OëZ›Ò÷lá9—#Þ¶;n»£Å¡¹g’ù3…!k#p´É,ò÷"s‰,$@èþËâνïN¹+à}\„õõê?yz\Ù+MŒõå"„ÆÜŸÀ}ùÚ¯Óð{æ ︌dùvìÔÙ/h„ÂýË…Í3oÓÙ÷C“øxÈ äYÞÔÞ­ê7#Ä:ËÛÓž9‡±·iZëÊúN$<Æ+)iÒRˆ1t—‘‡rÎ$”ÕbÖ½¬®îO"¹–<'BÒ4ç=g›æQaMÜÁÜ3É8Ëb5Ú^v«MK;¬ŽÏ:¼ÓKèØÈýÞ6n§6/a•³g1íŠci;‚ÑñƒÖ¢Êîµ'#5DŽgîÕ5Ú#wž_”ŒcQ–Ö ,oÉœ™TÞ»‹§dì%Š¢mc­%÷¥Š¼\Þ½âÆ1Ê2XSÄÆ*rOo"0ÄØl”m¢¯©Î±¬ÍÚŒãlhæÙ4OŒ¦Yœ° 6ܪž³õÌ¢®|’û‡ ä”qÊJ³¶±Î: ãv}—ðþÇ—ëÜû®xážá»9Å# Lµhí³»_>u΃E2¼ª¹×®•àSëêêš÷Qxxxvvv||üÚY!ˆ8•KܽÒaúÐ>Qä×ÛêÿY:°öï//nÕòáÊŽ}Þ÷Ñ_¨×ÒÐëÖ 2 ýÛÄ¡«…ÆO_ÀSSS Å¥K—.^¼¨R©»íêÕ«wCUUÕ…  ÏÀ†exxØ`08ÿE¿hiÜçyr¼""Ùî…_2gÊòg¬Ì›ÜPV-\¸U¢­+n_ßž&  ž–+¤ðù |ëU0. K;¿™ƒ†Óà ÷jÆQ@z¼åþ£ gÕ2aÆ~üô¶õ†zÛðð-4 ókcTU\Wm´æؘVÀG†¯Ð2È< ó€Ì2¿W0‚€+HæqᎴü‚‚³ÿýß¿ûʇøݵ¹5ñ„"`Eí!ý3ÿø]‚¾ÖϺ±)2óôw>-\³æ_§ú®J¹¤Ÿâ”"p‘Ü>8sèÞƒÇÏüÿþ¯_zÿþ~G¬ŸÌoN8ý§ÿúÏ?eϦbåÉýé?ÿëO§÷ovám{fòuå_½Vúrb¡3Çá¹ÙÛÌòŠyÈýëÿú;—+%!Í9÷oÿñ?ÿõüýì'ë á“>úBc[É/eÌ#½jmÍŸuZßQzóa¿y…òü,Þ7V·¾ëU®m¤ýMÐÁ8é‚"NJ¶zšºÞº÷bÎîà#XT°G«Û‚'ÑÞpq›A‡Ç ;[zèçM£îç1Å8|ÞÂÃkê­«ëñÆïÜQšBºçw -Y¶Ú^Ó†ÉJúíß‹š/­ìɵ´®¹¸lžÝp,Þ7V·¾ëU.¢µ­Mô·š‹û?ÈÜ`Ñ´7iâøÁXÀÊÊbaæ±ã`~ÖN›z”+ᓸöù³wÁ±r1yTXúÒžRt‘gë¾1iŠ()NÑ 3ÑUßš—Êëk()nz/A<×?.ÔÔZ7/#¯³o¶WÙ×Rñ·"KLú—ÛÅ|w¤öƽΠÍdeã ¼ùÊ_~¨›?}pj•“g iLöÉ4¯–û=Û¢ýH„sS÷ÃÒ§+tL?)Ý•r(Ö—CÛ,Ö©Ñö'ï"ËõÞ•”³…Ãb834>¨ëz7Û†¸8$åhJ¨ˆ2™iÆ«»å:÷¾k^0kÔÖ$¹„Ó7BÁ ÀJ²Ôf"ZÛÖÆßé±×ãE˵Vëæ\ÿ”üùI~3{Ý0aXR@GIý€=¥:mì­-U‡%óC}Wç[ B”ql¸ëå[ B´i‚æ“Ø’Ïr¶ÄGëÖôL2!ÄLõ×—µéª/.ôXƧæ†sKY帮¹Vô|ÔŠb,£ jë²U‹NÔÔ•» ô´Rõ)'½½V`ß gs\ìTU¥R?­I´¡»®¦ŸÂíZÍñÝg®,ë·+–M×^^‡ï‹›Ù‰I«õ["£Bü}<ù2«Ë.—¿Y©ÿw;«/Æß²+õøé3gÎœ>õõñÔˆÍÞ¢•Èéà‚÷—ô=5nøˆ`/=kÍ#Ä_=3ü·ÞÜý‡Aü52Ë:T}çªÒ<ý‡`—cÙ!BäM¼U™æþ •6 ó}økóÿXuþ,!’ÒCswWQúAu¡Y,Íb8†!Ä.Ó*›N­™ZÞLÉu«÷àl‰,ìÈDûÐožÍ]_¶éúµVûb!”J“ò ÌQB‚OŽ „¦7tׯxn•ÉÃöï“ùŠ±wʪògÖê;‹Õ—ã³ÿD†øyÉoÕZ+‹!ŽÊÏZ‘]òþ^À0ciØdÀÚËØ ¼æ›ßi£–ñõâ=3;ÊúyY´SŒ›ÏÒ-íÍ}9èP/Φ@)wl¯™xkâûq4»õÙ«–°ÙU«>6îMˆð¤5-_h)×­#vûp;f”ž”ʼ¹ûLD§Vkn—ªnŒ+²Æ‰áîÖáîÖÇFäžJü©bxU׫ ¯¹öá ¥Ö±‚oµP,w½ûäì>|Ó[ #+Œ+«„ŒIÓ74ùÁE ‚@ïéÅGî–=é·/#Ö§Mf‘¿cåãJd!BWla'UµC‘Ge¸=®Ü‘r4øUm¯ÙÝg©Ñæ6qjòB¸GpbZ(Ÿe’ùÞ^V¶mî«;V9Çe«>APJFÊW{DFjYV5=(î”LÏ Ï°ÔŒ]3/ÍÓº–Çæ„ô=[xŽY G¼mwÜvG‹ BsÏ$òg. %BÖJ»ß7–èÅf=%Ý:ÝŽÉ“ósÂÞGs=ú¤ãg( Âúzõ ó¬Y4¿=5}ÜkûißWUwj-ñ“sRC=·§öÝxÐcbmo5Eåžû–¥F«šž…~•¹WSÔ¨¥û’I„ |‹ºòIHî.SÆ)+ÍÚÆ:ë4 Barű´ÁX~Ž ¤ä¹~SìÑ£ òæDuûaï‹lšG…5qsÏ$ã,‹aÔh{Ù­6-æ±#5}Ïv‰ÿWý7ª5ˆðÚž¤¾ñp›"Ó§m^ôY„b Åeä¡œ3‰eµ˜u/««û“ÈE®¥ω4=ÑÍ ËN¿™ðŽÏ:¼ÓKèØÈýÞ6nçmåÜ*gÏbÛÇÒv£ã­E•=&ÜkOFjˆwø!„x~Q2ŽEYZ3°¼%sfRyï.ž’Q°—d(Š¶µ–Ü—(òry÷ŠÇ(Ë`M«È=½‰Àc³Q¶‰¾¦:Dz6k3Ž³¡™gÓ<1šfqÂ2Øp«zÎÖ³Eû†[õe åíéGÏœÃXŒƒÛ4­ue}'’Nã•”Ø_¤øÈr{ß/ Ü3|7§£xíw¸(ôàÑ}[H\°%@|&ÈÂXßÔÝ©R›aèÀ5°®®®y…‡‡gggÇÇÇîm#ˆ8•Kܽ²À+‚˜(òë‚mõÿ,°n «–Wvìûó¾þúC½âÈõƒ Hÿ6qèja‡q9ÿãIMMU(—.]ºxñ¢J¥Z춫W¯~Ü UUU.\ïÀ†exxØ`08ÿE¿hi„­½‹®sxED²Ý ¿dΕåÏX™7¹¡¬Z6¸p«D[WÜ¿¾=M@=-Wa«1Ä@¼'|¢°´#ñ›9a8=Üp¯f|¤Ç[î?ÚpV-f¼áÇ€Ó×êmÃ÷Р€Ì¯ŒQUq]V¿“ðš@æ™dy@æX ™Ç…;Òò Îþ÷ÿî+âw×näÖÄŠ€=‡ôÏüãw ^øZ?ë>ĦÈÌÓßýù|´pÍšê»*å’~ŠSŠÀE³èp¤Q™§ Îœ=wîë´p ¼ ÀÊÈüæ„Óú¯ÿüSvðlÒSžüØŸþó¿þtzÿf†fòuå_½Vúrb¡Ó½á¹ÙÛ̧ŠyÈýëÿú;—+%!Í9÷oÿñ?ÿõÜÛ0­FxŤ‡¾ÐØVòKóH¯Zg[óg݇Öw”Þ|Øo^¡Œ:‹÷Õ­ïz•kit0NºÐdíÌ:âÛu÷꯿ü|ùJÅXDÖWþ$ [¸Î¢‚=ÚXÝÜ“5õÖÕõxãwî(Í!Ýó;…,[m¯‰Ý(Fúíß‹š/­ì±´®¹¸lžÝp,Þ7V·¾ëU.¢µ­Mô·š‹ûççHÀùœÞŠÚÞé¬?ÔxÇ㾯c¶ColÀ=™G!ÖÐÕÄ&íñ鬡ï=QSµq¡Mç\;²o]ñ×__Yáµ;3#f›?ªýß?.R…ðŽ;z8:0È‘®m^Â6V þ±Šäp Ʋ£µ/ªªÛG,,Âö.8V.æá# K_gÏ:Ö+|cÒQRœ¢f¢«¾74/•÷ÿ³wæqQegÞî½µ/ìŠlÅ®€¬íÎR ‚"₶ ±;É´Éd’Nf²ÏŒ“Ž3™$“u2“žîžéîÄî¶] *"(»¢² *û*UÔz—÷ªŠÂ*·s¾øù`Õ¹ç9Kßyž{–®;ùyµÏ\ÅÎqðS5T*s¬š]dgíµ÷s´a‰[½Å<çqEÖ•–Ifî?ãòy­§?ª|~ú`Ò*i ‡°Ýâíë¯vx„¸°gª‡%·­t >Ûa]ìöpg¥×êÔ#M·ž=®•ã¸.:>l5‹Áp¦é­¹^ÙöäiâbßØ]±~"R¥¡h͸Š£¾“U< 7§oXR^ÂvL¶ÖŽÃb±8Lñ¨²°ª{Z@-Ê×të›Ó Œ²£^!‹tåôt=§óähSãìjµsÄ÷Ñá÷ÂJ2 í­îzmƒïÊÀš×$½ÕÕZÿƒÿQ›“­~=ÅàŒ7æió;°—cžïr÷òç­A¯ïÔ]Ì~°xÉa;oÙÔçœéUÑ„H¶7òJVÍ¥zTZÔ⓬+?}ióÞw<È>uªŸ‘$íqh¿9J™J ¸88%ѱþÂéŽ)p¾dkú~¯‰â?^jœz®Êì¤nTÏ­gCÔ&Ÿl,ò‘76¥œ¼—Ÿýو犴a¼¦•íEçÚí¶}¿ç sš¬2‘–«¿Züv¢gÕ©óå pQ`jzŒëgEû‚¸(pgŠ¤=çìµ €ûƤĺ±o?µ IMój½pîú8 l‡ä´dæ|^»! ÕîqoòA—€ç¹#Ý3·oXR^j¢åúÅC»ávâ$ƒù]:Kó5Ýú浂f¨[-µcu=ž?RÆrŒØù¤âºi@,b¼^P“y‘k]£÷êGÉ¥Š›s\cÓ3Œ¤G»Ì¼ŽÄ„þÑnÍùU½†ËË)egEA·´Ï8Ô·µ kHåè`Ûýa-¥š¤xllÁ´¬Õ‘!%åS4­î©*lœ˜«¼¸Ð‰¯WÏk²Êø5r°ü\N݈h­Bi†ÚšmÕ¼µîâ’65­T]­SŽöVxµËZ®.-n˜Ö$Jñ°²¼‡Ä ZÍrÞ¡).l7(–~¬©¨ß1³“ïžXìëêdÃ#@Ó]øi‘բЦʋñV¯‹Û{833óð¡ƒ{ãW9Š¬± ÍŒÖ_°(õ¸–ï$šÿçHØ%ípj¼\Ù§E£°ž7À(T+ÞÜ·¹?÷c ¼%2K7Pvùl«fúþ:cØ€9ÃíªÙaKJ10Îsæã4 3M§%DÔÀìÕUäDߘn®Y Å`8†0‹´J?Ö=¤^ÜLÉ|«æoÁ§924m•‰!v¢ú«gÇ—õc=r!˜CìÜ£÷glž¥„=Ê'¦ƒc5çÏج‘Hý7n8‹±'­¥EÕƒÖ’¯ùÊËrÚ¸/I\—ÿ—2¹Ž ÄÁé)VÉЬÖ_ 0 ÇŠž×KIvi¿\øp ì„•e@×_‘“Û¤èѼ°xÉ¿SJ9íì"ć&解¬‹½V®¦-LK)äDˆ#ç~ŸQ½X¶îœÑ939¬â9 qxº…À«°Ù\«^Öå»Gm ´¡†êoÞ““æ[5J¬wâ´ôÎ(=ÛAâÈ2ÌDƺ»‡.tÏ#ÜG,d”“ƒ6Ü\˜v(ºï“kƒä«ì9„} T^’Õ*7FðuZ’á,wŸ|šÀ‰§VÌŽç8oHM°©¿xõÒx±xÌÙL«†º^bH¥‚ïî8|ä¬MLð^„¯O©4"WcäãØI|Ý„æØÂLµW íÚ"à¿Ò'v—׃ŠN¥iÉ‘»â¸€ ¼¶Äûñf.™ïìd$³¬–Xe³­zIøž±I±ÛBE“JrQVÕÖñe k ›¸ ÿ¸¤u3›æ©±ú›šM‰¡«¹ÆY Kì±>ÂÛXc|¿´ÌwÞ̇B;!££,ï ôbÍé°fºIcÒSýŸÝ£¹}Òø3ºyb]ÏË<Æs߶7ŽW}áºAãÙk"ÃW£­ó ޼w\b¨ÔÞû°óƒÒËå}Z®{Ljœ¯›»w\WÖõ£ï¿QœvôCF·×ÞðÛ–5”S#'ë’Ù€óÝ’ ©–mEniaèÓvßòMûòq¶Z©ÖQŒ~´¥rˆÀRÙžx/,=•ŸŸ_7a¾k×&©½/ž¼TÒ©f@?t#»<"!-3g #Gš /6Ê)Là—êm纭'«lûõ‰ÑÝY%}@Ø%NÛß×Ξ=ûr_(--=~ü8jbÅ288¨P(Lÿ¢ïÕ× 3íçsØ1çÞdÎ([‹ª‰#{EYµhpá;ye^-Òøåíinäí¢V%zóŽ@ ^ŃªàáùÇïŒ\ÅÀpjðΕòñy¯¿zcÅYµHèñ;§>F¾ÜÃwJ†Q5 $óK­l¿v¾Y…@ ˆ/ˆûŠª@ $ó@ Ì#@2@ É<@ $ó@ –Bæq¡O|zFÆ‘¯~ë­mNÄ®ÞØk¶ì“¹Yõx¶kò7ßÚd/uZË!lƒ’¿õ7B„KVýËTÞW’/ÛEvHæ>ß-:¸Àãµ”Ã_úÒÑ£GKõ£™9X óî›_µépf‚74œ~/·ËpæWºçë!ÐqítÖí‘…Õ¦§g=¶Ûöýžs|ÌHKÐäv¼xœ&&>Åío¾5s¦}«¶æü™ë*feÔa–0Rp[o͇ҚÇÝJý’§µj¢¹à‚V°×J×Ïß7^my—+_ýã¦þ¤”‡OgÝ‘ðôjïïM7\ú¬OMpÖl9°3Lžu÷ :1@X*ó#5e^“„(*@ÜÓ¨ pq@¤°¥¾ƒª«±üâ ŒÅãÎ=T1ªÎÊÊGüòåV ŒÕ]ÎÖ Rô+FãØ.£àÖ=#–»›W¸ iWó÷W[ÞåÊ(yC-ul³ÛݼžîH åõerãºÇ-ðh;ÖÝ':@ Ëd€Q´Õ2Ñ¡N-•IÂ14X]S1á0}çÚÎ >Ôµw? Â~}rR˜‡+T¼wjá+ULj];BÜ=×µ=waÛÂà|×pYL€Æ0€Qò{¥eMµ `Ÿ„ô”µúîŽË몟x…KÅ\üñì‚û ÚTZ\ç°xY°NR4=ÙVÕé·?ŽÛu'?¯ö™«Ø9®~ª†J%cŽU³‹ì¬½ö~Ž6,q«·˜'â<®ÈºÒ2IÃÌýg\¾#¯õôG•ÏOLZe"-á¶û@¼}ýÕ6à,Bõ°¤à¶•Äg;¬‹ÝîÌ¢ôZz¤éֳǵr×ELJ­f1ÎÂ4½5×+Ûž<­C\ì»+ÖODª4­WqÔw²Šôæô KÊKØ®“ÉÖÚqX,‡ƒ)UVuO_›gQ¾¦[ßœV`”õ Y¤+§§Ë”~bŸ ʶ=·„ud@Û[ÝõÚÞ•5¯Iz««µþÿ£6'[ýzŠÁoÌ;Óæw`/Ç<ßåîåÏ[ƒ^ß©»˜ý`ñ’ÃvÞ²?†)Ï9Ó«¢‘4:mo䕬š1Jõ¨´¨Å'YW~úÒæ½ïx‘}êT?#IÚâÐ~s”2•pqpJ¢cý…ÓS4à|ÉÖôý^żÔ8õ\•ÙIݨž[Æ<«fùÈ›ÒNÞËÏþlD‡sÅÚà;ÒÊö¢sís¿àXÈ*i©±ú«eÁo'zV:_® ¦¦Ç¸~V4`±Nà¢À)’öœ³×&HBì“ëƾmüÔ&$5Í«õ¹ëã$°B’Ó’™óy톀 WºCƽýÉ]žçŽtÌܾaIy©‰–ë[ í†ÛEˆ“ æwé,Í×të›× š¡n]´ÔŽÕ5÷³„Cøžô„`§ñÒ?i_9a-ñÿ ­ç¡ä¼Èµ®Q‚{õ£äR 0×Øô #éÑ.3kÝ0¡´[s~U¯áòrJÙYQÐííÅ3õm-ÃZR9:ØvX @©&)[0-kudÈDIyÇ @«{ª 'æ*/.tâkÇճݹ…¬2~,?—S7¢Z«0ç¯ÙVÍ;Që..iSPÓJÕÕ:åho…uƒ¬UáêÒâÖ‰iM¢+Ë{HÜ Õ,ç šâÂæqƒbéÇšŠ*ñ 3+1Éñî‰ÕAÁ¾®N6<4Ý…Ÿõ[ËC5U^Œ·z]ÜÞÙ™™‡Ü¸ÊQd;Ìhý[RkùN¢ù~ŽÔXÝ…~ûëïÚÈvÙ ExÂjÞ<£|P­xsßæþÜ4ð–È,Ý@Ùå³­šé?øëŒaBäH ·«f¿B¥ãþH³À‹UÓi ‘506Kmȉ¾1w®*¡ Ç0f‘VéǺ‡Ô‹›)™oÕü-ø4G†¦­2Q#ÄNTõìø²~¬G®3s¡ƒ{ôþŒÍ³”à±GùÀtpc¬æü›5©ÿÆ g1ö¤µ´¨zPk¥¾3_yYN÷%‰ëòÿR&×1„88=Å*šÕú ´†áC™îÀ´z°îêuñÁ kÚ¯  À=°’Ìèú+rr›=Z€6ÆaKm/¥”ÓÎ.B|h‚~:ʺØkåjÚ´”BN„8rî÷Õ‹eëîÀã1“Ã*ž“‡IÚV-`³¹V½¬'ÊwÚhC Õß¼''Í·j”XïÄiéQz¶ƒÄ‘3d˜‰Œuv]*èžG¸1ŽXÈ('6 >l¸ ¸00íPtß'×ÉWÙsû@©¼$«UnŒàë´$ÃYî>ù4O5¬Xpy ­UŸ‹¡ @˜‹9ñ?Z5Ô50õÂE*|wè¤u IDATÇéà#guhb‚÷"|}J¥¹#—ÇNâë&4Çfª½b h׉7ø•>±»¼Ttj,MKŽÜmÇÅøˆ\àµ%ÞÇ0sÉ|g'#ñ˜`µÄ*Ó˜mÕKÂ÷ŒMŠÝ*šT’‹²ª¶Ž/KXk7=K$lüã’ÖÍlš§Æêoj6%†®ž‘#–Øc}„·±Æø~i™1î¼™…vBFGYÞ7èÅš ÒaÍt=.’Ƥ§ú °åî“ÆŸ¡ÐÍëêœxAæ¹n¶;ó¦Œq#¶¯ºÛÖÙ#˽yï¸ÄP©½÷a祗Ëû´\÷˜Ô8_7w︮¬ë*Fߣ68íè1†Œo¯½=à·-9j(§FN×%³ç;» $R-ÚþŠÜÒÃЧí.¾å›öåãlµR­£ýhKå € ¤²=ñ>^Xz*fß¼ÔÞO^*éT3 º‘]‘–ƒ3 †‘#M…å&ð‰K õ¶sÝÖ“U6„ýúÄè> lƒ§mž7-­hÎ+doOÍÜB:­fì~YYO4{N‘«oÆSíjgís6ùdÂ12eÇZ{;wãBîg–q›®+ÓV™J‹ ¼e{â}¼`o‚.§¸C…Û‡&ÅùJqcp]‚%,mkAyïâBæôTë•\<6)#ŠM“$¥mÈ¿ê!۟ƽ’W3JjûÊs˜pYÚa[Z¯'õ“]µ•Æ°6£WŽ3~ÉGâm0ŠbpBÛwçbÙ¬¥góö ‹ÊË(ZŠšweÅŒ…ë‡* »öEØÃÍÏ7l¤xÉ|M·¾9­¸MÀzVsÞËï´Ãí=·í=âÀf(@ÕYu±©<XX[[Ûsÿ°{÷îÈÈÈ¿öºáJ#rÏ̱EÌð¨úsA¯nYµx8’=o¿á|ãݪäú-,l·Äc[Îf7+óŽ'..N&“0ˆy8÷&sFÙZTÍHÙ+ʪEƒ ×ØÉ+ój‘Æ/oOsp#oµ*ÑÉvâU 1¨ ž>‘üÎÈU, §ï\)ŸG©ñú«7VœU‹„¿sêcÔèË 9|§dU@2¿ÐÊökçÛ‘Uø‚¸¯¨ @2@ É<@ $ó@ Ì#@2@ b)dúħgdùê·ÞÚæD|áê½fË>™›UÇa»&ó­MöøR§µÂ6(ùð[ßy#D¸dÕ¿Lå}%ù²]d‡dî Ü¢ƒ‹Öî=þ­ï~i-Â2¿jÓáoÿÓ?~{·×ÓKO¹Ò=ßþÇúöá«ÌØmOO=*Îúüós÷'ç:Ý‹¶Û{ÎûT1tûW~ø£¿=šî@ážzôoôƒ¯l— Vν\„}XbÀȽ!«ÞJkwvé—<­åPÍJz4VºQgþ¾ñjË»\ùê7õ{&D8˜š ó<·?¾r½WÇ@,†y{¤¦¬Ñk’Eˆ{4.ˆ¶ÔwPuµ#–ߊ±xܹ÷dT••ŽøåË­«»œ­¤è+:U+f„c»lŒ‚»ŸŽZ÷ŒXjìn^á2¤]qÌß7^my—+_ ä µÔ±Ínwózæ¾#ãº1Jó҈Р½eC Ö’yFÑVËD‡:µT>& ÇÐ`uMÅd„Àôk;7øP×ÞýüûõÉIa®PñÞ©…¯T!#víq÷4^×öÜ…mfD ø®á²˜;Œa£ä÷JËškÀ> é)kõÝ#;—×U?ñ —Š¹øãÙ÷´©´Æx…sX¼,Ø')šžl«êôÛÇíº“ŸWûÌUì×?UC¥’1ǪÙEvÖ^{?G–¸Õ[ÌqWd]i™¤aæþ3.ß‘×zú£Êç§&­2‘–pÛ} Þ¾þj‡Gˆ p¡zXRpÛJâ³ÖÅnwfQz­N=ÒtëÙãZ9Žë¢ãÃV³ gašÞšë•mOžÖ!.öÝë'"UŠÖŒ«8ê;YÅzsú†%å%l×Édkí8,‹ÃÁ* «º§¯Í³(_Ó­oN+0ÊŽz…,Ò•ÓÓ5‡Î³VEláÕæõëÀ Xª2 í­îzmƒïÊÀš×$½ÕÕZÿƒÿQ›“­~=ÅàŒ7æió;°—cžïr÷òç­A¯ïÔ]Ì~°xÉa;oÙÔçœéUÑ„H¶7òJVÍ¥zTZÔ⓬+?}ióÞw<È>uªŸ‘$íqh¿9J™J ¸88%ѱþÂéŽ)p¾dkú~¯‰â?^jœz®Êì¤nTÏ­gCÔ&Ÿl,ò‘76¥œ¼—Ÿýو犴Áw¤•íEçÚí¶}¿ç sš¬2‘–«¿Züv¢gÕ©óå pQ`jzŒëgEGqQàÎI{ÎÙk$!öI‰ucß6~j’šæÕzáÜõqØ!ÉiÉÌù¼vC@†+Ý!ãÞþäƒ. ÏsGº?fnß°¤¼ÔDËõ‹-†vÃí"ÄIó»t–ækºõÍkÍP·.ZjÇêzá2ZÜ.x«Sóµ;jôZ@¼Äx½ &76ò"׺F îÕ’K7ç¸Æ¦gIv™Yë† ý£Ýšó«z ——SÊΊ‚nÿh/žq¨okÖÊÑÁ¶ûÃZJ5IñØØ‚iY«#C&JÊ;¦hZÝSUØ81Wyq¡_;®žíÎ-d•ñkä`ù¹œºÐZ…Ò µ5Ûªy'jÝÅ%m jZ©ºZ§í­°nµ*"\]ZÜ:1­I”âaey‰´šå¼!BS\بV¼¹osîÇ zÉü Ý@Ùå³­šé?øëŒaBäH ·«fz”b`œçÌÇi M§%DÔÀìÕUäDߘn®Y Å`8†0‹´J?Ö=¤^ÜLÉ|«æoÁ§924m•‰!v¢ú«gÇ—õc=r!˜CìÜ£÷glž¥V=Ê'¦ƒc5çÏج‘Hý7n8‹±'­¥EÕƒZ+õùÊËrÚ¸/I\—ÿ—2¹Ž ÄÁé)VÉЬÖ_ 0 ÇêÅÌvôq¶³ÛùúÚéj—¸1IásëÆÐÕÁÂZ2 ë¯ÈÉmRôh^X ¼ä‹ß)¥œvvâCôÓQÖÅ^+WÓ¦¥r"Ä‘s¿Ï¨^,[wÎè™Vñœ„8<ÝB`‰U Øl®U/ë‰òÝ£6ÚPCõ7ïÉIó­%Ö;qZzg”ží qä f"c݃ÝC— ºçnŒ#2ÊÉÁ‡ ƒn. L;Ý÷ɵAòUöÂ>P*/Éj•#ø:-Ép–»O>MàÄS +^Po]wÞG2þÁó?˜†_E@ ƒ9 wiÕP×ÀÔ C©TðݧƒœÕ¡‰ Þ‹ðõ)•FäjŒ\b;‰¯›Ð[˜©öŠ ][$ÜàWúÄîòzPÑ©±4-9r·Qã#"p×–x?ÃÌ%óŒÄcv€Õ«Lc¶U/ ß36)v[¨hRI.ʪÚ:¾,a­Ýô,‘°ñKZ7³iž«¿©Ù”ºškœ²Äë#¼5Æ÷KËŒqçÍ|(´2:Êò¾±@/ÖLk¦ëp‘4&=ÕÿÙ=šËÑ'?C¡›'ÖÕ9@,™7ï—*µ÷>ìü ôryŸ€ë“çëænãוu½CÅèûoÔ§=Æ€Qãíµ·ü¶%G åÔÈIãºd6à|g7$Cªe@Û_‘[Úcú´ÝÅ·|Ó¾|œ­Vªu£m©¢0T¶'ÞÇ KOåçç×M؆ïÚµIêFï‹'/•tªÐÝÈ.HHËŒÁÃÈ‘¦Â‹r øÄ%†zÛ¹nëÉ*Â~}btwVI¶A‰Ó6Ï›€V4ç²·§fn!HV3v¿¬¬'š=§ÈÕ7ã©vµ·žúU&ŸL8F¦ìXkoçn\ÈýÌ2nÓueÚ*Si1·lO¼ìMÐåw¨pûФ8_)nl#®K°„¥m-(ï]\Èœžj½’‹Ç&eD±i’¤ô£ ùW2dûÓ¸WòjFIm_y.K;lK`@ëõ¤~²«¶ÒÖfôÊqÆ/ùH¼ FQ Nhûî\,›µôlÞ¾aQyEKQSâ®Ì£ƒ±pýPCea×¾è{¸ùù†/™¯éÖ7§· XÏjÎ{<ÿT‹pÛ™´ÞÝÓ•NŽ¹|WŽæÂL°¶¶¶çþ+ `÷îÝ‘‘‘íuÃ<”Fäž™c‹ & :˜áQõç‚^Ý ²jñp${Þ~ÃùÆ»U!ÝXFØn‰Ç¶ œÍn^ÔB»¸¸8™LvòäÉ'N´··Ï÷µ³gϾÜJKK?ŽZX± * Ó¿è{õ5è´yãöAÌù7™3ÊÖ¢jFâÈ^QV-\¸ÆN^™W‹4~y{šƒy»¨-¦G ¯dˆAUðŒð‰üãwF®b`85xçJùø< H×_½±â¬Z$ôøS£F_nÈá;%è’ù%€V¶_;ߎ¬B ÄÄ}EU€@ ’y@ Hæ@ ™G d@ ’y@ K!ó¸Ð'>=#ãÈW¿õÖ6'â Woì5[öÉܬz<Û5ù›om²Ç—:­å¶AɇßúÎ!Â%«þe*ï+É—í";$sŸï®wúÛ_?–i¸”ùðž('´ @,‚y‡ŒU›g&xCÃé÷r» g®q¥{¾ž×NgÝYè¦zêQqÖ#`»mßï9ÇÇü€´}AnNjǹaiÂáCQÜþæ[3gÚ¹jkΟ¹Þ©bVF­öa‰#·õÖ|(­yÜÙ­Ô/yZË¡&š .h{­tMñü}ãÕ–w¹òÕ?nêOJ‰pøôÖÜwÏé‡oþå\£’A£°¦ÌÔ”5zM¢¨qO£‚ÀÅ‘–úª®vÄòC1;÷ឌª³²²Ã¿|¹Ucu—³µ‚}ÅŠÑx¶ËÆ(¸ûé¨uψ¥Æîæ.CÚÇü}ãÕ–w¹òJÞPKÛìv7¯G±42À(Új™èP§–ÊÇ$ᬮ©˜Œð˜¾smçêÚ»Ÿ?Ða¿>9)ÌÃ*Þ;µð•*„cÄ®!îžÆëÚž»°map¾k¸,&ÀcÀ(ù½Ò²¦ÇZ0OBzÊZ}÷ÇŽÇÆåuÕO¼Â¥b.þøFvÁ}m*­®sX¼,Ø')šžl«êôÛÇíº“ŸWûÌUì×?UCåsî•©'Šì¬½ö~Ž6,q«·˜'â<®ÈºÒ2IÃÌýg\¾#¯õôG•ÏOLZe"-á¶û@¼}ýÕ6à,Bõ°¤à¶•Äg;¬‹ÝîÌ¢ôZz¤éֳǵr×ELJ­f1ÎÂ4½5×+Ûž<­C\ì»+ÖODª4­WqÔw²Šôæô KÊKØ®“ÉÖÚqX,‡ƒ)UVuO_›gQ¾¦[ßœV`”õ Y¤+§§kg­Ú°ïH(€ãÔHÝÕkMr=¸„Ud@Û[ÝõÚÞ•5¯Iz««µþÿ£6'[ýzŠÁoÌ;Óæw`/Ç<ßåîåÏ[ƒ^ß©»˜ý`ñ’ÃvÞ²?†)Ï9Ó«¢‘4:mo䕬š1Jõ¨´¨Å'YW~úÒæ½ïx‘}êT?#IÚâÐ~s”2•pqpJ¢cý…ÓS4à|ÉÖôý^żÔ8õ\•ÙIݨž[Æ<«fùÈ›ÒNÞËÏþlD‡sÅÚà;ÒÊö¢sís¿àXÈ*i©±ú«eÁo'zV:_® ¦¦Ç¸~V4`±Là¢À)’öœ³×&HBì“ëƾmüÔ&$5Í«õ¹ëã$°B’Ó’™óy톀 WºCƽýÉ]žçŽtÌܾaIy©‰–ë[ í†ÛEˆ“ æwé,Í×të›× š¡n]´ÔŽÕõÂe´ú¡›¹¹“½CSa¼;uëèÙÒA$ôÂÜñzAMnläE®u ŒÜ«%—*nÎqMÏ0’í2³Ö úG»5çWõ./§”ÝþÑ^<ãPßÖ2¬ •£ƒm÷‡µ”j’â±±Ó²VG†L””wLÑ´º§ª°qb®òâB'¾v\=Û[È*ã×ÈÁòs9u#: µ sÞñšmÕ¼µîâ’65­T]­SŽöVX7ÈZ®.-n˜Ö$Jñ°²¼‡Ä ZÍrÞ¡).l7(–~¬©¨ß1³“ïžXìëêdÃ#@Ó]øiQ¿µtËTy1Þêuq{gff>tpo\à*G‘5–³™Ñú ¶¥×òDsüiÕ`×Ðe˜§Ü¯|´&t  \ÂJÞ<£|P­xsßæþÜ4ð–È,Ý@Ùå³­šé?øëŒaBäH ·«f¿B¥ãþH³À‹UÓi ‘506Kmȉ¾1w®*¡ Ç0f‘VéǺ‡Ô‹›)™oÕü-ø4G†¦­2Q#ÄNTõìø²~¬G®3s¡ƒ{ôþŒÍ³”à±GùÀtpc¬æü›5©ÿÆ g1ö¤µ´¨zPk¥¾3_yYN÷%‰ëòÿR&×1„88=Å*šÕú ´†áC-|-¥Q0.Ú‹@ ¬'óºþŠœÜ&Eà…qØRÛK)å´³‹š ŸŽ².öZ¹š¶0-¥!Žœû}FõbÙº;pFçxÌä°Šç$Äa’¶†U Øl®U/ë‰òÝ£6ÚPCõ7ïÉIó­%Ö;qZzg”ží qä f"c݃ÝC— ºçnŒ#2ÊÉÁ‡ ƒn. L;Ý÷ɵAòUöÂ>P*/Éj5¾Ø¦tZ’á,wŸ|šÀ‰§V,¸<…»Êƒ5RK¢ @˜‹9~­ê˜zaÈ"• ¾»ãtü³:41Á{¾>¥Òˆ\‘KŒc'ñušc 3Õ^1´k‹D€üJŸØ]^*:5–¦%Gî6Šãb|D.ðÚïÇc˜¹d¾³“‘xÌ°Zb•i̶ê%á{Æ&Ån M*ÉEYU[Ç—%¬µ›ž%6þqIëf6ÍScõ75›CWs³@–Øc}„·±Æø~i™1î¼™…vBFGYÞ7èÅš ÒaÍt=.’Ƥ§ú °åî“ÆŸ¡ÐÍëêœxAæ~»3w­wâ¾%Ý-h¨·òñ×éÍ{Ç%†Jí½;?(½\Þ§àºÇ¤Æùº¹ÛxÇue]ïP1úþµÁiG1$`Ôx{íí¿mÉQC95rÒ¸.™ 8ßÙM ÉjÐöWä–ö†>mwñ-ß´/g«•jÅèG[*‡hL •í‰÷ñÂÒSù3ûæ¥nô¾xòRI§šýÐìòˆ„´Ìœa0Œi*¼Ø(§0O\b¨·ë¶ž¬²! ì×'Fwg•ôa”8mó¼ihEs^!{{jæ‚Ôi5c÷ËÊz¢ÙsŠ\}3žhW;kŸ³É'Ž‘);ÖÚÛ¹r?³ŒÛt]™¶ÊTZLà-Ûïã{t9Å*Ü>4)ÎWŠÛ€ë,ai[ Ê{2§§Z¯äâ±IQlš$)ýhCþU‡ Ùþ4šQRÛWžÃ„ËÒÛÐz=©Ÿìª­4†µ½rœñK>oƒQƒÚ¾;Ëf-=›·oXT^FÑRÔ”¸+ó(Æ`,\?ÔPYص/úÀn~¾a#ÅKækºõÍiÀmÖ³šó¿8ÕR=¸Z€oŽ{ýM1P4FO´\ËkVÐhÜB fƒµµµ=÷_»wŒük¯~à¡4"÷Ì[1QÐÁ ª?ôêVU‹‡#ÙóöÎ7Þý¨JŽÄåƒí–xlËÀÙìfåb<..N&“0ˆy8÷&sFÙZTÍHÙ+ʪEƒ ×ØÉ+ój‘Æ/oOsp#oµ*‘“Ž@ ^ŃªàáùÇïŒ\ÅÀpjðΕòñy¯¿zcÅYµHèñ;§>F¾ÜÃwJ†Q5 $óK­l¿v¾Y…@ ˆ/ˆûŠª@ $ó@ Ì#@2@ É<@ $ó@ –Bæq¡O|zFÆ‘¯~ë­mNÄ®ÞØk¶ì“¹Yõx¶kò7ßÚd/uZË!lƒ’¿õ7B„KVýËTÞW’/ÛEvHænúBìµyOFæáŒ#Gңݸ€@ f2ï¾ùU›g&xCÃé÷r» g®q¥{¾ž×NgÝYè¦zêQqÖ#`»mßï9ÇÇü€´}AnNjǹaiÂáCQÜþæ[3gÚ¹jkΟ¹Þ©bVF­öa‰#·õÖ|(­yÜÙ­Ô/yZË¡&š .h{­tMñü}ãÕ–w¹òÕ?nêOJ‰pøtÖ ÏÀqÙ’žìÔ’ÿ—χµ ³„ud~¤¦¬Ñk’Eˆ{4.ˆ¶ÔwPuµ#–_„‰±xܹ÷dT••ŽøåË­«»œ­¤è+VŒÆ°]6FÁÝO­|Q5v7¯pÒ®8æﯶ¼Ë•/Pò†ZêØf·»y=º¹"Ûv¸4e_hšD§á"+Ê<£h«e¢CZ*“„ch°º¦b2Â`úε|¨kï~þ@ „ýúä¤0W¨xïÔÂWªŽ»v„¸{¯k{Áù®á²˜;Œa£ä÷JËškÀ> é)kõÝ#;—×U?ñ —Š¹øãÙ÷ ×zÍ“Ö×9,^쀓MO¶UuúíãvÝÉÏ«}æ*vŽk„Ÿª¡RɘcÕì";k¯½Ÿ£ KÜê-æ‰8+²®´LÒ0sÿ—ïÈk=ýQåóÓ“V™HK8„í>o_µÃ#Ä… 8‹P=,)¸m¥ñÙëb·‡;³(½V§iºõìq­ÇuÑña«Y †³0MoÍõʶ'OëûÆîŠõ‘* EkÆUõ¬â½9}Ã’ò¶ëd²µv‹Åá`ŠG•…UÝÓ×æY”¯éÖ7§eG½BéÊéézAç9®îÝ•åHãÄ+ymou×k¯ÝáJwȸ·?ù KÀóÜ‘î™Û7,)/5Ñrýb‹¡Ýp»ˆq’Áü.¥ùšn}óZA3Ô­‹–Ú±ºž¿Œ®æuhÝ7§mõÑykEIý Ý#óÇë5¹±‘¹Ö50Jp¯~”\ªá…㛞a$=Úef­&ôvkίê5\^N);+ ºý£½xÆ¡¾­eX @*GÛîk(Õ$Åcc ¦e­Ž ™()huOUaãÄ\åÅ…N|í¸z¶;·UƯ‘ƒåçrêFt@kæ¼ã5Ûªy'jÝÅ%m jZ©ºZ§í­°nµ*"\]ZÜ:1­I”âaey‰´šå¼!BS\بV¼¹osîÇ xKd–n ìòÙVÍôüuÆ°!r$†ÛU³#˜”b`œçÌÇiˆkšNKˆ¨±YjCNôéæZÓÌP †c³H«ôcÝCêÅ̷͔jþ|š#CÓV™¨b'ª¿zv|Y?Ö#ׂ9„ÐÁÎ=zÆæYbEðØ£|`:¸1VsþŒÍ‰Ô㉳{ÒZZT=¨µRß™¯¼,§û’Äuù)“ëBœžb• ÍjýZÃpŒ¡æèÀŒFÅ_5uíãÚ!-¨ºnÞ ;àoÓ\óñ„ud@×_‘“Û¤èѼ°1[j{)¥œvvâCôÓQÖÅ^+WÓ¦¥r"Ä‘s¿Ï¨^,[wÎè™Vñœ„8<}]j‰U Øl®U/ë‰òÝ£6ÚPCõ7ïÉIó­%Ö;qZzg”ží qä f"c݃ÝC— ºçnŒ#2ÊÉÁ‡ ƒn. L;Ý÷ɵAÂú°ÿ IDATòUöÂ>P*/Éj•#ø:-Ép–»O>MàÄS +^\žB«úŽ ´ÔÓ9&ÉÈ›G fc΀A«†º¦^²H¥‚ïî8|ä¬MLð^„¯O©4"WcäãØI|Ý„æØÂLµW íÚ"à¿Ò'v—׃ŠN¥iÉ‘»â¸€ ¼¶Äûñf.™ïìd$³¬–Xe³­zIøž±I±ÛBE“JrQVÕÖñe kí¦g‰„\Òº™MóÔXýMͦÄÐÕ\ã,%öXám¬1¾_ZfŒ;oæC¡ÑQ–÷z±f‚tX3]€‹¤1é©þl¹û¤ñg(tóĺ:'¨¹êº±Ýikø*6Ë18Jôèárå„ÅÞ¼w\b¨ÔÞû°óƒÒËå}Z®{Ljœ¯›»w\WÖõ£ï¿QœvôCF·×ÞðÛ–5”S#'ë’Ù€óÝ’ ©–mEniaèÓvßòMûòq¶Z©ÖQŒ~´¥rˆÀRÙžx/,=•?³o^êFï‹'/•tªÐÝÈ.HHËŒÁÃÈ‘¦Â‹r øÄ%†zÛ¹nëÉ*Â~}btwVI¶A‰Ó6Ï›€V4ç²·§fn!HV3v¿¬¬'š=§ÈÕ7ã©vµ³ö9›|2ᙲc­½»q!÷3˸Mוi«L¥ÅÞ²=ñ>^°7A—SÜ¡ÂíC“â|¥¸±¸.Á–¶µ ¼wq!szªõJ.›”ŦI’Ò6ä_uÈíOã^É«%µ}å9L¸,í°-­×“úÉ®ÚJcX›Ñ+Ç¿ä#ñ6E18¡í»s±lÖÒ³yû†Eåe-EM‰»2b ÆÂõC •…]û¢ìáæç6R¼d¾¦[ßœVÜ&`=«9ïñœS-z¢ñÊ­ Iû^°Ø,r°:¿TN¡ @˜ ÖÖÖöÜìÞ½;22ò¯½nø‡ÒˆÜ3slÄDA3<ªþ\Ы[AV-ŽdÏÛo8ßx÷£*¤ËÛ-ñØ–³ÙÍÊÅxéqqq2™ìäÉ“'NœhooŸïkgÏž}¹/”––?~µ±bT(¦Ñ÷êkÐ[¾yãöAÌù7™3ÊÖ¢jFâÈ^QV-\¸ÆN^™W‹4~y{šƒy»¨U‰"ñâU 1¨ ž>‘üÎÈU, §ï\)ŸG©ñú«7VœU‹„¿sêcÔèË 9|§dU@2¿ÐÊökçÛ‘Uø‚¸¯¨ @2@ É<@ $ó@ Ì#@2@ b¯lC.ô‰Û¹Á™+p´}þQåèìöš-»×vç–XíU¶kò×Rž|öá­qzIÓZa”¸sK ÷Öÿþ¹ijiª™ÊûJòe»Èö>¸XÒ7Ç‘Š<ÿ̯ËðáIãχeãLÿÏç´€@ –ÉüªM‡3¼¡áô{¹]†1…+ÝóõŒè¸v:ëöÈB7ÐSŠ³Ûmû~Ï9>æ¤%è r;^®04áð¡(nó­™3íƒ\µ5çÏ\ïT1+£Öû°Ä€‘‚Ûzk>”Ö<îìVê—<­åPÍ´‚½Vº¦xþ¾ñjË»\ùê7õ'¥D8|:ëŽ#ä“û—îÕ²d¹Ä T C „d~¤¦¬Ñk’Eˆ{4.ˆ¶ÔwPuµ#–ߊ±xܹ"FÕYYÙáˆ_¾Üª€±ºËÙZAŠ¾bÅh<ÛecÜýÔÊ jìn^á2¤]qÌß7^my—+_ ä µÔ±ÍnwózžwèÉÇuµ³~­¾Îò$ „Å2À(Új™èP§–ÊÇ$ᬮ©˜Œð˜¾smçêÚ»Ÿ?Ða¿>9)ÌÃ*Þ;µð•*„cÄ®!îžÆëÚž»°map¾k¸,&ÀcÀ(ù½Ò²¦ÇZ0OBzÊZ}÷ÇŽÇÆåuÕO¼Â¥b.þøFvÁ}m*­®sX¼,Ø')šžl«êôÛÇíº“ŸWûÌUì×?UC¥’1ǪÙEvÖ^{?G–¸Õ[ÌqWd]i™¤aæþ3.ß‘×zúÅ&­2‘–pÛ} Þ¾þj‡Gˆ p¡zXRpÛJâ³ÖÅnwfQz­N=ÒtëÙãZ9Žë¢ãÃV³ gašÞšë•mOžÖ!.öÝë'"UŠÖŒ«8ê;YÅzsú†%å%l×Édkí8,‹ÃÁ* «º§¯Í³(_Ó­oN+0ÊŽz…,Ò•ÓÓeê.$ÂÖÇ}¢í¶ [ÂJ2 í­îzmƒïÊÀš×$½ÕÕZÿƒÿQ›“­~=ÅàŒ7æió;°—cžïr÷òç­A¯ïÔ]Ì~‰wŒlç-ûc˜òœ3½*€I£ÓöF^ɪ£TJ‹Z|’uå§/ `ÞûŽ'Ù§Nõ3’¤ý!í7G)Si§$:Ö_8Ý1EΗlMßï5QüÇKϽnfÙIݨž[Æ<«fùÈ›ÒNÞËÏþlD‡sÅÚà;ÒÊö¢sís¿àXÈ*i©±ú«eÁo'zV:_® ¦¦Ç¸~V4`±Pà¢À)’öœ³×&HBì“ëƾmüÔ&$5Í«õ¹ëã$°B’Ó’™óy톀 WºCƽýÉ]žçŽtÌܾaIy©‰–ë[ í†ÛEˆ“ æwé,Í×të›× š¡n]´ÔŽÕõx~WK=•­wuhÔB ‹¯ÔäÆF^äZ×À(Á½úQr©âæ×Øô #éÑ.3WÁaBÿh·æüª^Ãå唲³¢ Û?Ú‹gêÛZ†µ¤rt°íþ°€RMR<6¶`ZÖêȉ’òŽ)€V÷T6NÌU^\èÄ׎«g»s Yeü9X~.§nD´VaÎ;^³­šw¢Ö]\Ò6ý6—VvµN9Ú[áZ=ÖªˆpuiqëÄ´&QŠ‡•å=$nÐj–ó†Mqaó¸A±ôcME•ø†'ÂœïžXìëêdÃ#@Ó]øi‘Õ–2š*/Æ[½.nïáÌÌÌÇî \å(²ÆT3ZÁV ÔãZ¾“79µòòÖ¶ ¢ÅwšÞ<£|P­xsßæþÜ4ð–È,Ý@Ùå³­šé?øëŒaBäH ·«f¿B¥ãþH³À‹UÓi ‘506Kmȉ¾1w®*¡ Ç0f‘VéǺ‡Ô‹›)™oÕü-ø4G†¦­2Q#ÄNTõl·R?Ö#ׂ9„ÐÁÎ=zÆæYJHðØ£|`:¸1VsþŒÍ‰Ô㉳{ÒZZTm5ýš¯¼,§û’Äuù)“ëBœžb• ÍjýZÃpŒ¡Lu`L ñ¥Ú®¨Ñ˜…@ ¬+óºþŠœÜ&E€xaìYj{)¥œvvâCôÓQÖÅ^+WÓ¦¥r"Ä‘sf[ËÖÝ3:Çc&‡U<'!“´5¬ZÀfs­zYO”ïµ)Іª¿yONšoÕ(±Þ‰ÓÒ;£ôl‰#gÈ0ëìºTÐ=pc±QN>l|Øppa`Ú¡è¾O® ¾Ò¥e„} T^’Õ*7FðuZ’á,wŸ|šÀ‰§6µ†žçæ=(R¡! @,sŽÇ¡UC]S/ Y¤RÁwwœ>rV‡&&x/ÂקT‘«1r‰qì$¾nBsla¦Ú+‚vm‘pƒ_é»ËëAE§ÆÒ´äÈÝFq\ŒˆÀ^[âýx 3—Ìwv2ÙVK¬2ÙV½$|ÏؤØm¡¢I%¹(«jëø²„µvÓ³DÂÆ?.ip¦‚Æêoj6%†®æg,±ÇúocñýÒ2cÜy3 í„ŒŽ²¼o,Ћ5¤Ãšéz\$IOõ`ËÝ'?C¡›'ÖÕ9A™Pùö£î) YÂJÞ¼w\b¨ÔÞû°óƒÒËå}Z®{Ljœ¯›»w\WÖõ£ï¿QœvôCF·×ÞðÛ–5”S#'ë’Ù€óÝ’ ©–mEniaèÓvßòMûòq¶Z©ÖQŒ~´¥rˆÀRÙžx/,=•?³o^êFï‹'/•tªÐÝÈ.HHËŒÁÃÈ‘¦Â‹r øÄ%†zÛ¹nëÉ*Â~}btwVI¶A‰Ó6Ï›€V4ç²·§fn!HV3v¿¬¬'š=§ÈÕ7ã©vµ³ö9›|2ᙲc­½»q!÷3˸Mוi«L¥ÅÞ²=ñ>^°7A—SÜ¡ÂíC“â|¥¸±¸.Á–¶µ ¼wq!szªõJ.›”ŦI’Ò6ä_uÈíOã^É«%µ}å9L¸,í°-­×“úÉ®ÚJcX›Ñ+Ç¿ä#ñ6E18¡í»s±lÖÒ³yû†Eåe-EM‰»2b ÆÂõC •…]û¢ìáæç6R¼d¾¦[ßœVÜ&`=«9ÏÄò;–½ ÝÞ¤ Ñˆ…@  ÖÖÖöÜìÞ½;22ò¯½nø‡ÒˆÜ3slÄDA3<ªþ\Ы[AV-ŽdÏÛo8ßx÷£*9:reù`»%Û2p6»Y¹‹‹“Éd'OžûÙ¼êrss§±ÒˆŽŽ.))Aꈿf,ÝP‡4X± Ÿ'0Ü„»€jX±Þ<ªa‘Ì[ê.ÐOîüï;?þÑþ6ãøg=ú/\½i~þoÞ×Xõ‘­8rüü µÔi-‡.þ¯?øíkO–êÈõå*ï+ÉWÛþá?Ø2ïÕ…äè­?½ó½ï~ïûßÿî÷ÿýLÃôý8È›G Vðæ%îxý×õOUºûËC;ý8«Û ÙÆí6¼õÎÏñ¯oÇ®&æøXqãç¿©ó¼Vzâî{·kç›ßûU^Ÿô}ù¿úÞ›;wÿÝ{w'VÎ)aä@þÿÜðÚîkÕ{yq‘wd˜[ê´–ÃrŽÿÖ?ÿÍz±•²Ÿ¿o¼Úò.W¾\ibPýû¹ýsžh¯ï½ð›+_ùÙ¯ý«_ýú§¨Ó¿¿6B#o@˜?BÏëÍ_ÿèßäêÇ.VŽ®ßá„У7ržÈvE±wíñ´øW†ÖM)‰9ýÜ6âKG¢ú¨|?Zÿ/‰ßr¾a»b6ÿiÚ³.Aêo<¬{Ö-Û}÷?|cÒ®4LôWZÞåʀힴ‡ý÷gï§~7ä…‰£¦«b~àÁÅøÆoTҫݱª¢¢‚$I@ ˆ—–ù脯¼{|Ûì“‚nÙQ)[ßSpÍ&í˜Sî­iãòoŸ]ÍþÚŸ±Yä`áþpå^sôãßÅÛ.äêèûr÷ßEÍ ‰þ~6 €å–ôíon_cî‰|´¢-ïÿþ\1Dã80,„/y»·z¢úƒw~SÎ óR *´”Ç®}.uyµ£*Ê;ó'oÇ8¦ÒzUGÁÿ{½Ÿf³ bÕ¶Ã7O~¤Š8ð÷ßÝóŒ¢kÚ.WÙ&µÇͱjV‘[ ¾öá„ÿsºfD)×H½óØUÐcUïýö/êÉ^EÌüñˆäÙéƒi«L¤Õ÷_ùÍOÿw`ç·¢îµkÔêm7ýÍÛû„ÖpFm_ùŸÞÏ{¤#¸Økû!Ÿg>Õô–ú¿üN“zF´÷ø—¶®yj%¿ý§ßýé–œc+bábW;µÍþwÞ äšÓ7,)/9\úá‡eƒj½V«Ö€ãkGþîpèôäÑ¢|M·¾9­€ÛGítü(ç¾:$œÿ\= œÞË¿u(Ä–ÐÔä6ºÇ%p+íÑø…@ ,óæOÞï »ð—¦©o´_h Ù·_x3wÚÿðHýáOlþù·†‡¸ìø‡Ÿo½õÓŸ«Íô<ðóè’¿çÿø'›„‹¶Xûèóý3~쇿¶Åȱ»§~ñï9oÿd¯;Ûöµ¯|#®ú?oüǘڟ}õýO~÷»µxÓ,êÛœáÉ6•èÑâß¼Û»ëŸ~iO=ÙôÙ;'ë×¼õÉwØ?+‹úÁÚÎúƒbÜ<«fùogýô_V%üÃ;¿ôäÑSò –á¸ÃæoüÛfÐÞï_^˜Ó,`•‰´l·ßzóÚëïÖ¿þû“ÇœX@Uþò'nûõ7¹–vzìÆï׸õ?ûš3@?zçÔo>¾¯K7ªøȵ_þ¢>æŸÿí-Œ¶ïúþü?ñ“ÿ°ÙQÕþ÷‡Sé¿û Lˆ0S ÿó/7sû†%åe9Ǿõã8|ºâ¨¡Ü“7ú'Œoi¾Ï¶>Ãè嵧~ù³Kß~gŸ;ÛÌVÀD¾aüSwõáÞÏ͈Uñß<òé·$=mFz¿ò?¿Ø`‹<çÍŸ={eâe¼yŒí‘´Cù‡²¶Mw&¶Ó“­¼¹4fiZÿô“åF<ýãGœ·Œú2~ó“û²¿?lP –Cı·þã§u‰?Ú ¶ÛÖXoz‰ÿ¶­¾B ôv«ÙSZfÁ´úŽËEÎ_}'ÒžÀmB^ÿ»¤’?Ìa9Þ£l´½â`!« ..Íõ;öoÇCÅ8BG3*Âl«æCzüo¶:±¦•*<Æ!P˼¾ër®Í—²Íyº…ØNŽ¾Y[ñ®añ„îÑ_rEós™ gZ¸îÛÿîhí¿äöF™~ÝÃq sî¼~­š&uwu…~ã7¡Vë9¦ÊKOu–;Stÿ Åâðøu§Ÿˆöy‚ê7Ù8nN+04Åà¬û5Tv•8ôOØÀq‰ysÏ_Ý~ž8Û›Ÿ>³@ ^Æ›à¯;öÃïíX"xn…½…Û§^&9aïNµdŽ*sò±Uv“°Šc «Lc¶U/ë‰*ZrΕ?fûí<œàÁ1ß*‰¾ G<£ôš¾Æ^µ¯áS×p¿pßüV¨`ž<5ò'`¿Êoc²ßÆä  Ç+þýÄ'ë~û5ŽÕºÖ\ˆþŠ:ÉWí>]“‹/`ãjë÷Iš¦ ¦}’K_Tïþ:!™3±£Æz¶>«^XÔIë5›=3À‹ÔêŸ}7Ÿ››ëïï?çc§ƒù'Nœ@#ñW n›ŸùŽ­_X€Ý ‚ÅqX5ÙÜ«`ÔWßû Vaö ‰±m…cmƒZã¸?Ôtçþs6Ìáö[¿Xü»Ï›&¦÷.ëG«ÿôûºÍÇ"D–¦e{¥&Ê?üÓ1€š¨?óUSø&kudÖÔ4FYÇ*Ó˜mÕKδõú¯?}V0¶Ê‘½(«öìV|øAùnº¸#UýWé¸Ñ;e¹íÌ{· SehQF?z¯0·ÖPcŒâæ¿ÿðTË”ñCÝøÐŒ?KÊ^¶o, Æbgv߃ézjìî©Ÿþªj‚y%}*oT)Ûª>ýôÓ©©©iá7=Ûo©gÂ#W¿0ëf»n‹ê;סf€V6gSÇlpŸß7?ŸÌŠç#È›Ÿ×›_#r©ùñ£Í_þþ—‚„£jùä?>ºÕÒþ³ÿãþè+ÿ¯½û ‹êhû~oe—*ET@Œ%, •.J±¡ˆ#j,±¡€ˆú1Ć&*b|’X{0QSP£bCb!P|M,ˆb¤# Ë–÷Ã"Rv¢!ÿß®‹³çÌ9Ì°ç>3;;·&“Tú…~ôþ_­ŽÞ%a2d2n¯a³— 3âIË.ïøü§ŒÇ·öYÅM6!qÁ‰-»­WL5'qá™­òkVz,S×3bnÍ ÑÉb._MÝÐ-,ÌrO¢¾lo¿!âuiý'ÖWßKK®}ðãg[Î<ºñ|"7ÇÐg~ý4î—×Õ˯êeÇJË2·¯Ýv9K¶òK~ÌZ’‚Ÿ’v\Ê'þÐ/Æ¿î«Uy§¯ŠT]æ¿oѺ!s¦¶ó‚HÉΤ%ßW³¸\¶J/߈y‹¿Š_[9Ƙ£j>y1#eûªeOÄRbóT8*ú¶£CëžLž®ã×MÑ_=‘q8$®U³\6¹÷‹Ç ¥ÿ¯õ÷2u=fýï†è()C*’ðÞõ cûÉžž-ˆ¨û"Åß»<æzPâ< ^;ºªÖ«¾¶6(üNèîÏ'±ñ^x3„B¡··wmm-›Íþý÷ßÝÝ݃ƒƒÇŽËçó™LŃgÕ¹[¬Œ¢Óš‰b±XžºFÞ_9rd“ô3ÉÉÉÊ^j²R×¼^7uM§_f«¶àÂY†“â/™3u\>˸ö ¦]]U«IJn?î11þ‹Å2™ÌÚÚúäÉ“û÷ï?zôhuuµL¦ðS-QA7p¶[«b|ó·gJJ ºòÐœò™örÑliñ/_núánŒ!“²MÂ'(© ŽßÜvwU­Äê°! ï‚D"‘H$l6ÛÑÑñرcÇçr¹ÞÞÞ|>¿Ñ¤{""nŸqÓú´²|…«à™˜˜4ì”ãSyxY˜wuuí„‘ž©3hÖŠA¸*xýÞ<›Íf³ÙDäâârøðက.—ëééÉãñšEúVS¸ ^ý'ôèÊÀ‹‚Þ<À?Ñ›‹Åõî¡C‡8p`Ë–-éééÊGï[×›W¸½~"ºòðŠ0/ÿð?ñ?[ûSÞ›—Û׿¡FŒ±mÛ¶7z{{‹D¢7r–æRRRЕ€†”δŒŒDí´–P(>|xAAÊsò¡{¹={öÌŸ??::ÚÞÞžÃá¼æè}Ùöõ²²²šlÁL{€©…3íÙÊúM&î@ËI`2™–––wîÜ‘¿$‰–-[öúãö £øK~€NNA˜oÞ¿€ªªª²µµ‹Å,«¨¨èçŸ ›1cÆæÍ›'Nœ¨¡¡Áb±8NÏž=ûöí«¢¢òúg|IOߘøGzó> (++ó÷÷Ÿ>}úŠ+ØlöÀUUU ›Íæp8¨(@˜xûäç燄„˜ššoÞ¼ù£>š?~JJŠ§§§ššš²åðÚ›G¡)æ ©   kkëàààêêê¥K—†‡‡/X°`Æ 7nÜŸ;ÒÉL ˜Ù¹ùÏÚt©LÚÒ¸›ëözq·U%p ýæ‡û o Ð~aо}‘¥.™ó•pÆÖÓ“ìtjÿú¿Ì ¹- £±´¯•¥ç8v š ½yh©ÚÇÙ·kÍCƒ¨qy] -¼‚ÇÛj1‰ˆdU7¿^H¼X$©Li\‚¢£týŸÚ 3b\=#7ÅMãçåìä=w÷-!¾0€Þ<4Ã5r²áý/á£ÍUîö–¦ÆÚܺrŸe®±ævàWéßØRöÖéÌXÛ÷ø K¢š‚³×û9ôqo‘0»Ô'wQË€IDAT‚›t<Öªéê+«[âOöŠ;½kHWFÅýwU9T“4-ú‚ç¦ã_8w)<¶pIß”­k’&d½(A˜Ûô(ej ÒòlŽŒ3’ÝþrBHÜQï½ÝYhNôæ¡!†–ûªoù—·ÅLôh98øã”û5D$Ì=tJäµhÚ6[Çajô0ñ©osª‰ˆ­=tÆèÞ<ÑË&v1X*ÌgnæÞ+©ej[Yê±…¹ÉGŠÝÏvë¡Ââ{Ïa˜u(»òUG)=GoØT#.‘Š±‹‹ÎÃ+EhK„yh‘ùï ›½v÷±‹×®¦í™Â=½àë{b—TiöÖãÖõùõzk •ÕK£›Æ«»Í<‹yÂ{ýº2x…ãÈÓžHÄ%÷ŠJS#}]]]]]]=&lüƒ!.J_q”òÿ#5mUùÿƒÃeJªÅµ@˜‡—Å{•nŽ!SíÙ÷³‰ˆ­i Z~·¨®,zúg¿‡‡ˆˆÑ¨Ï Åë¨2µlÞûd÷‰Ë×.n/ܵ(!SÔÅPWßÿóSäÒ;´{\wfãšU…f@˜‡¿­öþ÷Ÿnúö×ÛËEµ~Þ³-CÚ×±§ ñÍÆzqR?Ýu¥T")ÍÚõé)Ö@ó¦ŸÀ³Ô»ò*îåW6Ÿr_s÷ä±Ë*Ä2¶†®ž&—Íb¨š…ø«X™t._(%‰ði^ú™ìbIãš…o| ÌÃk´‡ªÿþ¡ÿ„x °´pð‰<¦–¸1ÐE¤6 æ‹E½NÌln>hæ ã¨/– Pot¹ïŒá–ÿ3íÄ䂆#ìÒŠÜo–8õ75µ¹¡rüª(;>ß"|G‚KÞê1vf‚þý#ve–Jš”PÛü(´À[…qëÖ-Ô@§•œœ›——‡ªèxïèëÙèÍtXøÞ<Prr2*a:¦ØØXT@‡„Ïæ:&|6ßv’hÅaÂŒWïu¹5oæ"Þlioèš²ã<‡|r élþ ´o‹OEããã‰(((HQÔËŠU².=Â|{ÔäƒOù¯ô+L±”Åf¶ý±î™ÿxW¾>Æ3 FƒuäLLLšíÞ,,‘¤äÒ–ižVåÐ9/r¼6K#Û$"¥%Lñv´Vžï'œ)|þºøéùuï¹Ù˜™Úú-Ø{³J^Zó,·µw>î0åØÓºõª2–¹º,½TõÊó¾:aîGUÒŠ«Ûg³˜Ùx†%þZü¢…7=ÿSóíO}€g¼lÅXu—5Iº¼·ûÂåËç¿ÛED¢ücÇUf~“yãâÝs?[~´PBT—FÖ|ùñ«9Ù?Fé·pÉO5 ¸|Aðg)—sr2}¨q8:îL‰ŒˆHôðèIðï3¯ÿ¶}øýõ³7^Òó,·ž›ÓoÜHßâug팵—kŒ‡í~íë_Š¥DDÏ®<Çðo©úÊó*(ªRFu s?>”švb¥£4cíì¤â {3®gü@üÝ©B±üXe…7ˈ$%÷ŠJ3"}]¹ ""Ymä\(!­úœ´Òò«ûV¯Ùs.¯¨–Á”T–³|kdDD¬.Æ]Uä»pu{jTÞ-“ȳÜh”åö—²Z¶ý°@ã-Ïÿåï“—ü3ßw»€O¢WWqQ¤Ý0a®¸¼ JkGþ7©ôª;Z¬¸pí&Û„ù¶/ÿx¾I€W:OY2Ù†ÍÖÅPWßpãÉ•öJ’ÉÔÜú**ñá¤]'ö˜k²jn¬ö›öDþ‚¤ôþÓ""=¨Pë¡Åz‘åö]5ÈrËê1dBŸßœ½¥•ö«æˆÝïªI_u^%EQĹlͪUËÅD"ª-}R)5VþG 3š&Û€àý?®uë‹)O&Û_aÙ;H«Ëkx}4X$yzq_ÊCÑó*/$îºR*‘”gíÙ¦:lT?žÒ,·¬nÞý¿/×lüMgÔ¨¾Ü–œ·% sy¦²Ô=éO%D¢»?ì¸øLÚ²? æÛ¼¼¼&÷æ[^PžL¶qœW”F¶áë–3£Ýó¢ý|ü'.:¥íd ò¼|ÃáCK?mo1`Ê÷†I ¬ø¤<Ë-Kß-¨áowºñ§EçmIÂ\†ºÃâM“«Öõº"×bpÝøý« €VÂb·»èÈææaæám¤,o,òÉ ÌCûQ““à;bc^qú‡Î¡GK^¬Ç#Éß ØÌ9S&{Ã%Â<ü»jïK¾­óç×}‹¥¨„yx“*ÎNuðß/_Fxe™Ûе7jHžÕ#|ÝÒ÷Fûz¸Ž|Yٺ䭑›â¦Žñórvòžû"}í«ˆþøî›|«±u3÷žž{NYÞX¥ùdaZ©æÑù›¶Ÿ$ÿðÓ©Sžmž«4‡l]â×´<›?}foàƒÄ¸£…Mƒ0ÛÀwáüaÝÕú½3ÝJµn¡ºê[ß~÷¤ÿxOû‘#ô¯íK•ç»¯T’7VÙv…%Â<¼ KËcÚpcÇÈï×*å9d«‰ˆ£7lª—HÅØÅEGž¾¶qi]¬||úkªtwåѳn)ܪk_-±ï¬Çï5|d÷›Ž?)Í«d»Â’a^æ5ºkÕedivW­,xžC¶w£Ä¯Êj‰ˆ©¦­*o\‡Ë”T‹_=j/+¿²ÿd…U ³.“¸½üFÞ9øß"yÞØ> òÆ*Ù󠃭”ˆ$ò ,©*¾˜'.¹[(OO5Oþ,kœC–ˆ'~m5YÙ¥}gËE——zÙZ[[·õžôÞáC·«_ä%ªËKD óÉ6Þ󠯧}÷Âs¿Š‰D÷ŽíL/;eÏ.$îÈ(‘HJ3wnHWiÙÖ“]Øó3Ãcݱs§SSSSSOŸù&ÒôÉsdJòÆ*Ë' íUÐÎÄxBÜ”‹‘C¶ëëõttqÔ»Wÿ×p¤oÅFû+…dìÙ0‡ìò%³o*“i™Ž'~ý‹ÔH Ï컢é¿kh_}ÕºMº3mýøÀï‹Ö.Þ4yqÌXïÄ®:zföƒõ äãê ·@ûD´o aFŒ÷RÝ/\d†ImÐ"HD Б!ÌtX´è˜0hБa¦=´TÂß–——‡¶CÛ!ÌîwPrrrPPÚm×Z´è°ææaæaæ;,áï+<=ã¯V£&pkE˜›‰nmð<çà3}cÚ_Ô À&«º}ä“)Þö@`å>!ú%’ço@Ë£%2""É_gâ|¬=¢~̯Eµ]Óÿ8ºnÖ¨Á@`ã¼èËôBÔ?Â|õ]ÿóœk'Üó¿Xw¾ UðFŸ¦ÿÜ3cüŠlÓˆý³¯]NY5´ô‹°‰ŸeUÊì#~t"6xaº}Â×kFqPgm×4ӟן”xì·ììK)Ÿ…\ÿîR1:;ó ƒÉb±¸½ÜBÇõ©¾ûWý£¬¤(-aŠ·£…@ °ò|?áL¡üŸ¿æÁ‰5S†Ú ó##狉ˆD§¼lµ×‰‹ð>wKÓ×oºn»âó(_.Ÿ«iì<ýÓcËw­ú¡àù»D”ÿcLpô× â}ºc­Ð6n›ømË'8wáóµŒlG-L\?²)ºõ 3b\=#7ÅMãçåìä=w÷-a݃n‰óoé³¼ÔÔ{œžýtt%ø‚àÏR.çädúPãptÜ™ ¯mš}®wôwÙ¹¹WŽ¯1UcUç$M‹¾`¾üøÕœì£ô¿[¸ä'ŒþUßIͬ±çÒ•Uÿ`­aåï®~ëÄõ ‰ ¾ Ž½å»eìû@›5M [·ÆÕÎP|ë#¢š‚´<›?}foàƒÄ¸£òð["Â|{'ºéfcmin7v‹pôÇË=uê_aéð÷±Ðç³Xê5J3÷ÂÝjanò÷¥ÿYèÝ‹ÏdòôL̵Y$ÌM>Rì¾x¶[ßØ{î ìCÙ•¨Z"IÅ“*]]>£Á6–V7 IyQ•”ˆ¤/\*éí7ÚN1¾4Ò[GoØT#.‘Š±‹‹ÎÃ+ED¸%6á¨vˆc±üðÖaÚ*ªÚšª, ë;øåW÷­^³ç\^Q-ƒ)©,gùÖÈÄUùϺØwã5*B\r¯¨4#Ò×•Ë "’ÕV0LÊ…ÒÂ} :=–†¾jÍýâj©Ö‡IYaSSW•ID\ÛØõöÉQ“çªØ2éÝæ!Ú´i”Þúˆˆ˜jÚªòÎ*ƒÃeJªÅ2"Ii‹o‰R}„yøW0¸šz]õ´›ß]jn}•øpÒ®{Ì5Y57VûM{BÄîb¤^v§°šÞm0¶Ïîb¨«o¸ñäJ{>ê ^_/;•Èo/üåë¯/î•U^ûþ|¥É‡ý5ψˆ¡åñ¿ÿֆΠç&'÷æ¡ÎÚ¬i†Ú«D|›öÄwLÃq{™Â[ŸÒ¨Öò[¢ð÷ÎP«´›H«Ëkx}4X$yzq_ÊCñÍ‚Fk[xæP*­yz+3·TB|³µ+“Îå ¥$>ÍK?“ɪDDŒ.®‘sL¯,Ÿ½þ§[ÅÕ¢Š·E„V ]âoXZ˜Ú΋w' ¿?1êÐ}ê¬ÍšÆ%*¼ÖG3â]É/ Ëf§l Jy\¥àÖ§ n‰óo1¾åÌh÷¼h?ÿÀ‰‹Ni;¨È·†o[5øfüH33»á‘nVʈøá;\òV±3ôè±+³Q@ŽÛ7l[r¬UΧ!ƒ¬-ü¢Oh~°ó@´ƒF£4–žGìÞž9MŒù±_Ün«¦é=iÛ¡•ÎvÎö`c3xÜâä"‡  ­Ýú”ß(qKlôðtëÖ-üg@Û011ÉËËC=ü ÉÉÉAAAh;´]«\ÏÎ@o ÃB˜@˜„y@˜ï\’h£S¾É‹Âì8Ï!Ÿ\«néöVÿ,Óžˆbccë·ÄÇÇÑ¿;iæáÍhãëÅWSaþ­ïÊ7‰ñõ U˜㽨6нäìo÷þ*fÙÎÙ¼n’€Ï "QÁ©K×ìÿ%_È3r¿vÎ@É÷G|r|¯¿žèú*Ÿq{û$þ¼ÃG»:+Ö/ŠŸtti_y‰µ|ôˬã;Fê1‰¨êòRïƆ+¨ÖŸTR”öiÔ'ß^¾WV«bàô^ìºÈ!ÝX$­¸ºsqdâé|†SH¦„ê–V¶½ùÖe)Ý:ó›•€¶„ùN¬¦ -ÏæÈÁ8#Ùí/'„ÄõÞØ%Ï¢è¹éøÎ] OƇ-\Ò7%ÑÚ]Ýé›Ïü•ddTwS»}æ¶ÐÇöÑ/WDKß©_†›Ókø„žÛ¤ýåÐEY_ŸãxoµPm|N¾ ø³”ÏÍt%Z<):ÎöÔçÃ8kg'OÞ›1Õ¬ú—µ“¦ŠC‰ˆ¨RÉvEW¸ÕýOÅ;Cg•——‡«ÐvЖ0¯ýQ’Z±yŃöÜëçî Ën¤•9Î ézãçÂ’ì´'½½ Â8Ë`XH¿¼ä³…’•eHWhÚ8¡Âü¶7œ•‹µÔ`±»:ÏšëP÷Ü t»¢+|¢dg@o¾Ó>z)J­¨0‹"×ÉÃìÙ†ßþðÌ»oäíîye×Ú+÷­Óïvó°êÒðùÕÍs¢Åú©ãFæìÿMgLx¿& B+Îo[P¥5HOžâ‰¡jЫ.‡­¸\ÉvEWXY¢xg@˜ï‚‚‚âãã~<ߪ)xJËJhˆñ£ƒG 5]çtë§%x²-%åÀŒîÛ•©ëj·*ñX–ú/Ù=Æ-ëÍi\¸âü¶š=T«—‹‰8DT[ú¤RjLô’튮P˜qYáÎÐF=GTÁ?MÙ¼VQ’E‘¥c뢛³óÈ3;Çn5SÏîW·-5óh–›¡å:ðéŽåŸfOðíÙôÙNa~[ži€‡,uOúS ‘èî;.>“½d»Â+ä*ÙЛï ä½v“¿×ç-Âw$0—¯c7³BÆÓëç8&j•;×ÈÙNãóB ÷Þ¯ ”À÷æÛ–žãØ1DDBÔ 7ßÁÔ¢?M™{°°`ïû®¸O?òXB$*8•æe#¬½>H¼X$!fĸY²míd7 À!àãÔü{ÇâÆ2œB6\*“I§® u³2˜š[;­¿¦äñAR”–0ÅÛÑB Xy¾Ÿp¦P¢àXWèÍCk©»¬IšÁM:kÅ#"ªÎÙÔ<Íë"ÑÃç´·ÿmVvdθ…/{ÏýôÔ5ÃûÛ'…Æíýýì¾â«[âOöŠ;½kHWFÅýwU9ÊNÙ<­ËÝ&Çv±lrU€Þ<¼N¾¢4¯•DÄÒñšÿžm6Ïxh  ƒíþ¿‰:KMà;ÊèIæƒj"K…ùìÁÍÜ{%µL c+K=%vŠrѶôX@o^‡Â4¯B‰6±4ô5XDD ¶ —£ÙMSÞ¢>G*ª•ñ,æm_÷ÙÊàÄ»ÔoȤèsÜô%ƒU”‹–çØìX 4Â<¼ ’Éž·“ÂD´ÂŒWÂÔ²yï“Ýï}R[tuä”E Ï®wTm¶—Â\´ Žç4¼*hç0hß^±Ô»ò*îåW¾$ÍkKfÀÕÜ=yìòƒ ±Œ­¡«§Ée³ —»U˜‹VÁ±¯ æáoá¾3v†[þ\̶“ $|‹ð .y«ÇØ™ úôØ•YÚ¢yîÒŠÜo–8õ75µ¹¡rüª(;¾¢Ýø–3£Ýó¢ý|ü'.:¥íd ¢øØ&W…fhß°¦=@Ç„5í:2„y„y@˜„y@˜„yhˆ}=;µÐ!ý?sÿ²F­P/ÔIEND®B`‚gnm/vignettes/screenshot3.png0000644000176000001440000020165712515147523016061 0ustar ripleyusers‰PNG  IHDR£2pÇTÙsRGB®Îé pHYs  šœtIMEÝ 3cø IDATxÚìw|UÇ™÷Ÿ9ívuTAH $!ª¦«QEl 8qœ vv×±7mw³dc;=Î&v²›ÄÞ8qQ ÆÑAÈ4K4!!D“P¿º’n¿§Ìûǽ’®P» Ñäçû¹þørιó<3gt~óÌÌ™!“R&( U¨¢(Š$IÁA€ ‚ Èc‹Ñh"„0 CÂ0 GBR?DF†wû3=&¾_Ü~‚øtq7çÈÝ\Ü~€ÜÍÅ= ¾eÐ}†ø–».Ùòíâ^‹¾ŸŸË«·‹{ø¡¯{.õùb ^ÿù|ñÝT@âS»©Ðäê‡oד»/î¨×>]Üþêãõ´Û$|º¸×_Ñ>= wsq§Kéý+?êCÖ¼OPŸ „öUºÝ\E}.Ÿ¯ï¸„Ò»¸øsÛkJ}—1½›‹ï­:÷§~߃ƒ&S3¡ å(E¡½>00ìv¶†AäñÅ-èn±B8EQôz]@€?¥KAAþn±gEÁâ@AA‰,ËœÁ ÷÷÷ó1 §”*T¡ P ¾ü„R ÜWRJ(!0,˸ç PJ !x3Ad`ñ÷÷®Mƒ}ÓxYEQ’eY¦n…fY±,Ïq]gËQJ ¥@©"Š²"ËŠ¢(”2„0 ð,Ïs ˲ CA±GAûçk(/SQgv\䔨 ˜C˜^Sg±—7›ÏÜnÚq­Z­Vqçý „R*‰’Ó劉‰‰ŠŠ  0™šÆÆÛ·«ËËËU‚ϲ,¸ƒýG ªH2%,ËÞƒ_Šdn6ÛN`ÐqX;JPÅn±Z’LΠ×rÔeµ›m.‰0œ_ žµµšT¥7økz¨\RúSDY3§ ú>d^Q\.qˆÀü:mL¤^E…* a`°ÕçòŠœ2/ày¥‚RJ]¢Èó|FzzPP€ì‰ê%†áX–aY¦©©ùè±c¢(ªåXß•žRÙisØ’$S À°¬ Ré´Â>Ѩâl±ðÁAœÅhƒ»|V*’E¶zÃwÞøè?hU£Øed^¶Ú SÖ>÷tjüȦ’ñÇý ÚÑK¿öõ¬ ÃtZŠþò_ÿWŸõ£Ÿ¯¿øÞ/ß:mÓ©rêO¿*0‚ ƒ³ÙÒwï½BQ”¢Tì[I*ŠÍæ®×„ ê?gŽù××êe‰ã8B@Q@’e• š?Ë‹ÅBRÅmÈEB€èôš¬ öíÛï’D–e|{–ºÌ®!S³—.œ1fä ÔZwýüû6ï-± ü€<Ñ(uµJÉ?y}"Ôå|ÿ'»-jwéRE”ýR†ñ OĨ÷žU'¾:/Zå˜gø¯sü@vŠŒžmn•ç¾ô+c ¸œ"ë29‚'M‰ä˜4k8›wQQ Ì@ן~V`A} TItýdZgmr:íÝ_dmåÕÚŸ&G­ÿâDzŠ"º\éi³(•ì6‰zAÚ$‘ç…3¦:tÔ=bïC¸,šIÒ ¿~)=¤£[B7sZÑÖÏ.Q—Íaµ‹`XµF£÷tÊv³ÝîRdJ€°¬J£Ñ«ÜOP*;›Ó%°¼ÎÀ!l›I[s«S¨Xª8¬»K’w"‚€šZ­f§ì9ÀóZFÍ©Ô¦Ãÿûõ,ÿ[Î:5ëèÑôÙ»{çÛ%I”†Uiï2ñn2¥á;~.ÛÍv»¨È  „ã4*"9D—L°‚Fí§m×»Jß7Ÿ©â²;¬÷o Ã2¼Zcа¤·ñÝáÎ"ípXí¢(S „Óè´,éñvôPæ ;hTúñÜ[ßÿåÉVàU*⟙Kª·løþ¶*aÕ¡fÓ‘¨t]ÉÖËÔSÒwX×j]ÞõG =ÞâîËÁ=óµ› Ü¥ªëý4*ƒòPú^bzJA’Ä…áºPj›› Þæj´‹IÁ:÷e&›žg#õ*Ùn×,ŒÐ4ºX–e)**R­VÛlvwO¾¢(”º‡ïÁ³J!¢(i4ÚȨȆºzŽï³ÙAVÕäWק‡@]Þß?ÜQxË$©‡Ä$Œd®˜^M%«Yµø[ë²&ðgĦë_|žóÁá*NÏ:Asþù¹y££ÂýU@lºybç‡ÿÈ«ÔŒâ”ø1Ù?Z›9!\ Š³ùÖéwûáYÙÓ[±úÍwWÀ­ÿúã£!xÊ3ßÌš4*fˆŽ°Wúņíü3ÿñ¯i1ÁZÀQ_zjÛß·žj¦"ùä³ R¸Væ\á{•Cvoº³¶QÙf¹[çYBd›EHXò³ S†ë—ñêñO?þðdƒJÍù”¸Ž†Éwdê'›kÔjŽ€bwÍyñks’†F¨PÌgžjŽI‘©Gݹ=ÿÏÎkŒŽgî.ýM7%Ãè¥}ù¬È6»fôâ¯==/ydÔÙRW}%ïÿäÖJdÈÜî „8\¾9ܹ ä°’¨ôuÏ.™:zˆ Äæ+ûþü«m•²¬îîvð ôPæ*Pµ¿øäWÞÚü €õ‹ ¯ìó×D>õÆ'OÀÍ?ÿË[23'Ž†a7÷ì.oTsÄÞÅúÏ?UµÕŸ³ï×J.g··˜q8»¯¼ÐÑÝæU9FugUÿø²¬PìdPÀgz|÷%9}ˆF²¶*’¨HbCþmaå¥úEK[]p«Î)»OIÖÖŒ!Q’)YR""#ìv»(Š¢(‰¢(I’$I²ìþ¿Ô~Ün·GFFH²Ü‹ž,: —%« üÃßÿï‘ë­ðÄn¼^xâªCÅËFâžûÏ «§Œðg€42ã¹ÿxmy˜Ó"¹¨ÿØÉ#"üUDq9àƒbÓ¿ñ½çã¨Õ%Z5O¼üê áj¹¥¾Æ(ê‚Tv{GÓG477™škm .2}îÄ‘Ct,ˆN 4Ls«$ËÚ@Þin±Š MÌøç. wˆÞkÈЃi‘vdM¡÷⼨¸,0âÙÿø©ÉÃõŠ¥Õ)š³þß_ž¤²8Ÿ·JŽ®™’ðüPq)þc§ÄE¨E1 Ÿüäê¹É‘ZYP‡%¯xéŸ'òV‡â¼«ô%UÜ×úôYqÚ¹¤ÚðïO=12€m©«3J*ÿðØ)ã‡PQ{¼›²O;©ÒQ¯¨hS¢VþÇÏŸO=DŠ(óþ.›Kb{Ê‘³Ç2§´m1ZÅÒÒdjn¬ivµ×Gk“©¹©¦É®tjJwk]òZGT±÷`Î!÷T1¼ßQj÷ý§‡üàç±ý´÷Þ÷Ó Š,SÅj£ŠcC~°4ý·Ÿ]°½¼å•E©ãH«b6€¢ÈCuE– !T‘õ:Ëå¼£ÏÀÝ{ßvD&„Ȳ¤Ój©¢ô½L³,³á±¡õE&^#° Žç8PD‡aÖº9Aö‚?ÿäwy-á ðæ×G [²bÜÞ· $w­~úƒ¿Þ\üú¯ž¡›”M._g†aàæßò‹Üz™áT]{ïuݧ¯mØz[fxÁOdz-‘Ÿüà2‘Qküôló»ß}毼Îß  zâ»?{zdð˜1[Ê[»ºßÅô•[´}^½'çK¯ÛýÓ¾6/àö¦½¾­BûoüV\Ê’Iþ§òEU[äêcâí™Òò,¡]·nÜß~i² ®ýõÿýG;ÿõß~k¤4m(SxÕîkú*õŸõrŸ>+¢3pÖÚT?ë‰ÿþñ[gìÃÖüüÍì>Š´´œr>8|ö&\ûx]?íKÂÄKïÿìWûn9@í§£4 µ‡½U¨Êè¡ÌóL¼Çú¡ßüð«aXól$QúÎ7œ²¨ˆ0¾£NKvÃŒ;­ë‰B:6š¢Óú4×¥n´•ƒWVãߥªëUÁÅ1ä«Ý{ ¡”j©èÖlPš âª)ãþ‘þÙãÇ(MRkS[»€¨Ñ=¯P $IBȉ§»¦=cÆT÷Å*5ëéÙï]齚Äí·÷¦²ÄŌǥ=§›U~\͇¯}}T=6”9SÝñS˜J¯˜`D .XÇR"Õç-ŽýÖ[oÍ=sxç¶ÜS"å;®W´€ÊÞM •A§å¨Ó®»îÕogÅ:N :é!'M+ ¥L¿œ—56 ê™_ÿõ™ö«¢9¹Fñt†ûž¸Ê Óz׈Nw„0ÖŠ¢Z˜<Tz%ÍWËL02Pã¯ei›>¤¯¡§>‡ŽGÉ"«ZÍ‚oE*ùäpGÕ¡’ÄÅLˆ!Î ÛÝæ ú †*’ƒë9Gõ=ù¯½kˆ^«e»×ÌN«1ðà2Sï .êNë”ÊÞƒ|”Oæ:—x—CGîZÕÊC‚ R¥ïYè)!„˜íN•¬Pêév,½UýiiËÚéãw”Ä$øöÚž¬ŒÅî „†Bì6¥2Âdd¤¹‡çÛ#{JÁn·+T!v›ƒÒ÷ú=„•š*› š<&`ËÁV^Ã2E–d†g::'(q/Çשaà•‘]²;AB5SþþQ±|Ù²¹cc¦,yyÊäá?~í£Ç†J©BÉ»QP\.füúW²âùÖ¡M‡.;‡.{~^tïsœ½M{•<¥÷æ<¥Ä=Ñx:÷˺¶§;µ\me¸öô}O¼½«§½mÕ™6ËÄû zéƒ>ê±£…Þ–lîT >:ìUòíþPªx–tª^rÔ£ÿ¤ÓœROav.OJ»Ô£î¬wú‘oæT``5ÝUõÍõD‹/á!È §Ï“º»¡ Ã^5;((Š,*²XÖlÿïËÍß™óÄÒpþYÓÿPÚ\b²¹OQª\·8–†aL¦fYVÜcó6›Ín·Ûív‡Ãáþb³YeY’%I–•ææ†e{sÃýaÎtñXÀˆ¯½¼>-ÆÀ€Kâua£§M ¢#Þ¾t Ôc?à4Ë3çÄ€TY\/‘^ÆX•A*?ðþï^þÖë›k |Ú¤Æe³Ê4"Rå²X[­R—.ªÈÄ/*T€[G·ìÎ;–wþ¶ó® ¿#k„aïÉyF®½R  ×ÖÚöÉæ?Ú¼qçá܃—-ÃÜKâ]‹Ý§Zä{ú„a}ñF®¿Zêq §†P¢ RõÔé¥mÒgÉa•º²P[2mˆ³ÅbjqI'÷˜#YêÑÒ•^tg]$à-á>›ël©kV¨Ô]UQ¡}þéá?øyŒ?}νWç­c†•$Ðj´/ÏMíªŒ# ß_0MU“J6V8Ò( ¼–†ã¸ÆÆN#Ë2! Ã(î…î݃ô”R÷Ú¸Š¢pW__ϳœ‹òAÕtðÏÛ¦ÿlå(.jÎúÍYßvÆQø‹—þr¹åô'‡—l˜0ù_~»ñÛžfLÍÞíœ,°=ÈÅÐÕ¿#S×ÚØhãBÃ@®¯¶€Ëqæ:MŽç'}÷ïÛ$Áž÷ŸßÛtƒ¹#SL×nZ Bûõ7ßL½mÓ Wû 1ÞItd™UµÞ½ó„aù9'²¾?C?¾ÿœ(*<ϸÎüòÕß_£^Ë­øž8½#¦÷M9Á÷ô Ëùâ3Ãr _l=·èådí”þåGÿܳW÷â°W>–1~±åÌÂïNÑNzñ—Ÿ¬e†gÊÞ]ÿz9rqœ£ÿ¯v½½´›N‰÷ìÞúo:V­ >šëô—‰U]*ð'•ÃžíZÕ)ÜÇALßËœ=I¦¼Àieo9†ÊR¤ 'ت¥¦:*KrsãHó­¡‚Be‰0L¡…³°ÇŽššL«MQQ]¢Ëå].Ñåjû"ºDQ¤”6™š›M&Žç}šCÈð\ý¾×¾ÿÖÆãe5w—®bo,/ñÉgJx•íôÛ?}së‰ W+®—«@‘DzLjÀ=ÔúNy$‚Ê^ð¿o¼ùÙù[­20 (N§(JŠL=q˱jçú]¢Ãå^7…áU¼Æ-6T±Zœ%‚V­á@v:-NXÁOÇ‚,Ú첬¸a8ר€RÙa]’{MN«å9蔈û‘*»\vQ–Ϩ4Ãr* /x_Éöhºë¦@wëS…rƒØÐh–(¢–ü×ëÏdŒ{_ecµJ=ˆïwΧè.YvWžÓi8¦§õâ×ÌRÅiq:(°ZµÞÓßNE«Ã&#¨ j¦{ëjÆîŽoæîÈ`× Ì*ÝWuA1.—HfÍœêÖæ^àA*—c½_ë$•Ùó&¸{:†-tªÞmõs •JÕ> (ÔétÊ’®Õj½û2Ý’o³ÙjjjY–áxA–d\0é4Ðl×¥½ñæ ±ŠÃjN pkã÷±ÛDë ‚ ˆOˆ¢ÈFGeújÖ˲¬TÃåYøJ¨ÕªbÀ¬0%¢cÖmµù³j­ ´MÂ!,Ë¡¦&“ÓéB–¡Š"Ûl6£±Éh4 ¼ŠãyI’G ‘Îý62è‡ÇŒ öóÓ©9p¯|ú¿ÿ·»Rp¢8‚ ˆïSE!³fNå8Ö—«9žcν¸¢ÈŠB†0 Ëó<Ïó¤»ýéÚ–Z“DY”dEöìjÏ°,ÏñO©"Ë2¥ îOtÖ;$QVÜëå–/°X,‚ w$É}¼Oß©À%"³,§Ñh†ô¡Íî!y†!”0*FE=‹qS @J)ExÞU&Ò#]k¯ºs` ë ‚ ÈÝÒ÷®µw„éŠâÅŽÝð> ‚ ÈýSz”ZAÜ1=–‚ ‚ f¥G©GAÁ«ô{ï,A|üÓ7¿Ž d!‚ È`•AAPéAA¥GAä+¥ôöó?Íœý³"Gç/¼–™ùÆEÇÀZuÞÜôòü‰£'>õ~…µýû_>ßЋ­ûã ‚ ‚<ÚJïºòû9 ‹ß¹Ù¶*žtëÙ é¿,~dEÑ~ñ¯*šõ·‚Ëç·<ZÒþ}ýô¥/¿’Ùî}BÔÂ^Îöhê܆4l ‚ yL¯ÚôÙΛ.+vZé÷Hïêjª´' ÕÎßÙ)O."¤‡ Rz?‹ ‚ ƒWéÕñ+³èžíWàºþÙN缧FkÜJOm¥›~ôä´¤„„¤©Ë¸ñ²•€b¾ø·™7!aôÄÌo¼}ªIîPàê¿ùÆœ‰ æ|ëíF¹G“ÎÊ}¿z~nrBBBÒK¾÷i•Ô“­.iJÆÜYúýkÑOM1Åüžï3×¼séËöþùnÒ÷ê½ïÆO{Á¿¥f~ï­Ÿ~sùÂ93§Îéƒ+v `Éÿ·—6×UôõÔ'žH_¿½V–kþb]ÚøĄĤ 3W¿Yd+>ëw[®;ÝͤŽï‚ òÈÄôªË–ªl)±ƒ£t[.»xÅ(µû„µð×/üêjæó‹‹óÿgÎõ_¿ðë/-ÔRðëùSÓÓ\*Øü-iÇ:É}±£äOÿôÃãIÿµ÷bÉùÏ¿ºã»ÿžÛнÖÛ‹Þú§ýáŽó—/ŸÝûæšDéÁV×4÷)óþgço&éƽ¶÷ô‰ýŸîËõ|ÿbã #…ÞÒo§'?Õye_ß¼÷ÐáVU¾ýÓÝu2ègýêOO‡E>ûÁñ/¿<öî“ᮋÿóÆþá?=téJiQþÆ/ˆàÄÚ#~°¿JèüAA\ß—1‹Vø?¿¹¨aÅÎúå‹Q]r+æåm\s~÷OOq“¿ùÃyŸ¼ºµä•§·Qæý~Ý8 †™ß~iòæßx.ÎÙÞ”þËI‹PDÏé…w³··ÌíFè/ç|Öœñëïή€ÄÉ!=Ùz•ßÝMš‹žè=7ݦï}¶;?™÷ÍCˆž5+裳·]«Â5S&¬Š±^)½\‘0"0zü8nê\ðœ×z}GGˆœœ,d¬^½ún”¸¡óŸ ýÇ;3^ ^ùíaü ÷Q©µÚæ÷Dˆ'TBbýí'[œ­Õ6ÿé!î™mD9Üß=ô-™*ŒÍßËJÍ$¾Õ®t£ôRs•5`R˜ºóÁîl9ºO³Ìt›~ÇÙnÓ”ÑjÝý„Ù!uÝ*@=ö;¿å·ÿýógÞ.‡¸Ù_ûákÿšŠÿÈãÁ† °d0QVVvWJlØìÕC_ÿ·ü ?=‚Òs~‘ÚÖr£ Fñàj¼Ù¢‰ðWùEhmµ­bs½E‰à¢‚C£þ°ÿ瓼a{7!.0Tßr½ÎáNz±¥î6MÚÜG³¥»ô½ÎvëgAɯÿ‰Ïþìƒg&/~ò½çð›iGÞœ¢Åê†<†ÏA_rrrîh»ûö>=;$ëWÿããß,ëR5£ŸœÃüÝûg›e¹ùÜû¿;ÀÎ^•˜¸"ƒü0¿Qp•ï|ï„Uñ\¼&[·ïç:ZeW@¶7–å>ßÔí8½fôêeþGùöáJ»¢8¯^n–{°åsš}¦ïuönÒdõCÔæŠ*‹;Îòý{¾¬4K”3‡ø KÎÈCA6>®œC4C“§'ÓxO^Ó=ñoýÁð}ßž‘”4ýÅ}Ñßÿë¿?¡gô“ôÖs¶7Ÿœ¿tź×.áéÈÍØWÞûͬ²_.O0fZöÿ{¿°¹ÕŒ{åÿ~1£ô%GNYô½¥Ú½-â{š}¦ßqönÒbž|!­ê?gNH^›S-š/oùñŠ©c'.ù½å©_|?Eƒ3òA‡ ™5s*îZ‹ _eÜ}}Ø{ ƒò/w­EAA*=‚ ‚ Ò#‚ ‚J ‚ È£F÷ïÓ'$$`Ñ ÈcÁ믿îËeÞ f!‚J°dÉ,yÄÙµk$''{cz>()}nr'‹N—½¡èT]§³BpRê쉡% G•‡ò¯4wtH3†¸ôEé£ô’Í!+“M°ŸÙz¸Z`ƒSÍ7tx8]½nŠ ›.æî+êØ!–ÑÇÏYôD¨J¬.ýä½üF¯­cÙ ‰KVÎ<¿ïÆ°q<0k»v$÷tu{ŒÎú'efŽ8Žb¾ž¿ÿd…Mè§]F™œ™–@("/=VTï¤>yÜ)O®›@Fn8·ï`‘·°Eî[@a=‚ ¾(=£O\¸8ºlgÎÁ €5Ä¥-NâO·õ·4;¦tûæC& ø qYÙYtËî2›»Z;?SuúÃwÊêáóWÅ{ö·—g?ßX:fõB׎mW»íEW,e6—5wÅð.÷ù}Çƾr& eeFtÍžrWíða3V¤Ñ¼›*m «MÍ^>iïÖ‚&Ù¯@¬=±kWke­U`ýÇ.Y:³1çh j=rŸúö°G땾—“CR’íG·•¶¸ÃtÙ|-?/6qžG®¹°))ŽÃ›ŠMž ^l*:»*%äz~ƒ ™*ZBÇŒs•×7™Zû?ª0¿‡\1Ënq./µN ä¡MS‰:4iÚŒq‘ZF‘D—‹b¸È¸úk’èâS£Š÷l©t÷€l¹y<7luj̥Ϯ;úö [M¹­­µÒr9ÿúSO„ó5•(õÈ@ðÆoÜ[SA¯¶Ò³†ùö—Þ)6Ý2ºÏY]PÀÐÔk¦{‰!«æ5,€»g¼©`Ë&¿ðèØø©S¢Ã ¤¹ôè/kœä9í˜ØF¥ã\ÈÔ'Îíùô˜ÑEXÃØU‹Ä «fëÊlÞƒ÷²¹Ú¤Ó0×J^uAv˜©V…/9"Á®]»°‹A{RzÙÜÈŽJ*ÛÅžŠjÝߥ¦ŠšDõä· IDATŠÚÏr+zÐn"tÔÒZsíB͵ '€Ñ%f?ZõáÁšûúf˜k<²µ´m\v9%* DʲŨ„Eè˜Ú¥Cü#Fû½ÌÜS Æ5âKv‚ È}¦× Rj(<§Éœ3:ÀÝ`ýâ3$éÚ•¯éü Ç´yBU¤­Ù`6>e„Þ“¦fTöÚ´¡êö“ºuyÍp“m}d ßÖ,ˆŽ‹Òõ?ÆU-RP¸Þ=¹žÑǦ­Z¯%ûíRkÙñê1‹fDkOÇÈôE1Wßtøâ—vÔ’µ‹Æ‡¸ÛŒ.vVªöÂù;'ü!‚ ÈéA±–îÝŤ/X3™W$I/ìÙ´&sE¶jïî‚FÉY•·“&gf?ãÏPDQ[Ë óÛú·©h1ÑQYëfûY¦ ë¬:³ãX}Gë¬8|*.ûùx»Åî’©ØX’_ëù©g<Œ&,J½&ÖIÁyûø®£·@´#2—ÍËç¸v¾ac',Ȉ‹e²&×î,0JÔ\r hÞ¢µÏJ8F¬½¿¿üÉÔ•ËT{ö¥~رö‹my)s²×¦1”"5íßqÑ(@_^Ø®îËe¦g¬þ†d…(-%w›ñ5>Aä~CÒÓf¼óîßï8š°dÉÜËAe222233_ýõ>é{ßµ6''gÆ 8EAwüE¿°þy†eY,A”°,‹J ‚ ƒZé9TzA¤pÓ#‚ È éYŽÃ‚@AÁ©ô‡1=‚ ‚ ê˜ÇéAd°‚ãô‚ 2ØczTzAA¥w_î?&ë™õ¯>7N÷ ä#³^Z?-ðïøví²†˜éËÖ¬}fͺu«R£TX Aû«ôÜ]ŒÓË-ŹÛÚåê±®IÈž#æîºÑË6¶Š£þf…e€wpXv@ˆ˜±*+¤dϧëœë‚ >âãã±î\’ù~À±,÷ðÞ²#œZÕÇ/rÓÙÝû‹]à#gÍ(Ú¶½¨w¶A”+¤3999«W¯Ærø˜žã¸¾zïù ¤ô¹Éaœ,:]ö†¢SuƒÔà¤ÔÙC9JŽ8* å_iîØ­Ž1Ä¥/J¥—lYq˜l‚ýÌÖÃÕ"œ²hþ¸¡ÃÃéêuSdÙt1w_Qsû.®ž]åTš`ué'ïå{oïÊM\²rvàù}7†‹àáXÛµ#¹§«ÛctÖ?)3st€Àqœ óõüý'+ÜìõË.£‰LÎLK ”‘—Ž+ªwGç}zBdÊЊü<”yAäÁ)=Ëö®ôŒ>qáâè²9[$Ö—¶8=Š?ÝvÖoÜÒì˜Òí›™$àƒÆeegÑ-»ËlîžiUìüLÕéß)w¨‡Ï_ïÙ(^6žý|cé˜Õ ];¶]í¶]±”Ø\|ÔÜûÜç÷ûò¼á'?Ø’g–Ñ'.]•ùñjOw»ÜRrhG‰§wœ HY™]³§ÜÕ_»|ØŒi4oç¦J›ÀêcS³—OÚ»µ IöÁ+Fª©½á:={fœ^qK9_‹}ø‚ È}Vú^ߧ熤$Û.mq‡é²ùZ~Þ-‰ñÈ56%Åqx±ÉÄ‹MEò™))!m J¦Š–Ð1cã"CüÔ,8*ötàö@};+¹b–Ýâ\^j äÛÏuhRÆògÖ®]ûÌÓO-ÏH¬ˆ ¢‹O*Þs²ÒÝ?²åæñÜŠøÔµo^ñ*Ö0iáx×é­}œ³»4`Î’±~ VBAä>ÂõÓ³†ùö—.¯#bÓ-£KðœÕ M]±fº—²j¾Qø{Æ› ¶lò ŽŸ:%:Ì@šKø²Æ9@ÎÓŽp˜*JÇ?¸©O.0œÛóé1£‹°†±«L»HÌÖ•Ù¼;ßesµI¦a®;”>¼ ›fˆõàß k`+?qjâÊx¿â‚fìÍGAîcLß«ÒËæFv|ˆPRÙ.ö|Pt°Pë Ù›*j*j?Ë­èA»‰`ÐQKk͵ 5×.œF—˜ýtjÕ‡k¤ûš§ÀÄXã‘­¥Æ¶®|—S¢Â@¤,[ŒJX„Ž©mQ:Ä?"Ði´û¤ÕŠíöµF­³}àŸÊe1¦GAî³Ò3,׳ÒK …ç4™sF¸;¿Y¿øŒIí/ÓËMçO8¦Í›ª"má´aØø”z|iFe¯Mªn?© ÐQ—× 7ÙæÐG¶uo! :.J×åS-RP¸Þ)F›¶ji¼–tRì{´K­eÇ«Ç,š­e<=#ÓÅ\=~Óá›gRÃŲ™ÉCx.xìdýõk-Ð#‚ ÷Sé¹>Þ§W¬¥{w1é ÖLæI’ÅÆ {ö­É\‘­Ú»» QrVåí¤É™ÙÏø³Q”ÄÖòÂü¶þm*ZLtTÖºÙ~D–)Ã:«Îì8VßÐ;+ŸŠË~þÞn±»d*6–ä×z~ê™Ï£ ‹ÒF¯‰uRpÞ>¾ëè-íˆÌe³GÆÀò9®‡oؘÀ 2âb™¬Éµ; Œ5—(š·hí³„Žk/äï/2uå2Õž=…F©v@¬ýb[^Êœìµi ¥„H Eûw\4ÊЗWîÒl¹¸÷Ô”O®Ör<'Õ|¹ç¨QÆJˆ ȃD±V•V˜Ä(-ãÓqä±Ç=Nßûl5±©ä`N‰×Ë(õªöÛ…{6vûK{ÙŽ-½¾Vª˜¯øèJ·gÜ3à»­m7oúËáŽÝtaç»vzÅìÍŹŸ{ýâÂ. €Ý¶üìÚXp^yZ75gvn>ƒ5Aû†«áJÉm§ì˜X\ë`YܵAdAÍF§"£Y·xðá}zAäqzG“Ñ¥‹ˆÕÕÝll•üƒxT¶5ܪ¨mq  f(x&)÷tlJûÓ#‚ »±IÔùéy[c‹û$ÙZ}£V Ž7~\|¨ÒÔ*zbýžŽ#ƒ ÜŸAd ½­±YÖùqŒ*0€·76;)€bojÿˆ-Kgo{ר§ãÈ Œé{{ËAyl­-²6È%@T‚ÓØä¤@—̪yÏËÆŒ ò<ó{:Ž 6¥çpœAd½¥±U¦Ê­’¢*·’Sj2:Â"Q\•{Ì^”A@z8Ž ¾˜¾Ï·ìAG*µ6˜Áoxâ0GWÓͲz£=""ÈÊÍZœ¦z³âVtFÓýqJ)¥”BÁbJÏáŒ<AA ôbK£• óWóœAfPš­ ‹Š Uª¯\.-»ZeÓxl÷ÇeYþýïo³Ùñ pø–‚ È`€Á£&w>ÄŒºÐØÑ¡í'¢Ú.`»;î’¤wÞy§µµõÕW_õ÷÷gœ©÷¸Çôw¥ô¬ÿ˜y g$ªN½û~‘õ8ÈGf½¸¸ùã¿2=ئå}°«Ž_ûíL¦®µmý[Î/L:üçWX y”p‡òÇŽcYöå—_öóóC±ÿ*)½ÜRœ»Ý©]®ëš„ì9bh⨿Ya6×Ç®Ô|ùøg— k=‰r³ŸJ4ã²÷‚<š?~|Ú´ijµúÛßþ¶^¯G±¼•ž{hoÙN­ê#d–›ÎîÞ?8ìJõ缶à‚âÂŒWKXy$1bD^^ÞÌ™3ÕjõóÏ?¯ÓéPìS8޲ヒÒç&‡q²ètÙŠNÕu:+'¥ÎžÊQÂpÄQYp(ÿJs‡|1†¸ôEé£ô’Í!+“M°ŸÙz¸Z`ƒSÍ7tx8]½nŠ ›.æî+jnr=»Ê©4ÁêÒOÞËoôŠ~Ù ‰KVÎ<¿ïÆ°q<0k»v$÷tu{ŒÎú'efŽ8Žb¾ž¿ÿd…{ƒ½~Ù`4‘É™i „R ²ñÒÑcEõNê“WšXþ#‡¶\9-bDäÑD–åÑ£G9r$##C¥R­[·N«Õ¢Ø?®1}¯oÙ1úÄ…‹£Ëvæl‘XC\Úâô(þtÛY¿qK³cJ·o>d’€—•E·ì.³¹×TTÅÎÏTþðr€zøüUñž)Ÿ²ñìçKǬ^èÚ±­û‘jÏ®r|ÔÜûÜç÷ûò¼á'?Ø’g–Ñ'.]•ùñjrÊ-%‡v”xÖudRVfD×ì)wõ×.6cEÍÛ¹©Ò¦°úØÔìå“ön-h’}ðÊ»X ±Ã-¥g]XyD‘$‰eÙ &ìÛ·oÞ¼y*•jÕªU:_½{ •žëõeznHJ²ýè¶Òw˜.›¯åçÅ&ÎóÜg.lJŠãð¦b“'ˆ›ŠäÇ®J ¹žß H¦Š–Ð1cã\åõM¦VGÅþ*ÌugÅá#WÜãÜŠ¥¼Ô:1‡6M%êФi3ÆEjE].~ˆá"Ðo]%ºøÔ¨â=[*Ýý [nÏ [sé³ëŽ¾½òn?ÅŒp^)ƹx‚<Ê1½,ËÇM™2e÷îÝ‹/V©TË–-Ójµ(ö½*=k‘oé-‘bÓ-£Ë³Ý« šºbÍt/1dÕ|£†p÷Œ7lÙä?uJt˜4—=ðeÍ@ íØŒ*JÇ?¸©O.0œÛóé1£‹°†±«L»HÌÖ•Ù¼‡øesµI¦a®;”>¼òn1h£ãä+{íXùy”cz÷kù0kÖ¬mÛ¶­X±B¥Reeei4ûA¤ô²¹‘"”T¶‹=,Ôz*BSEMEíg¹=h7 :ji­¹v¡æÚ…À賟N­úð`Í}‡Æ&Æl-5¶u廜­e‹Q ‹Ð1µ-J‡øG:ö»{OO®°aÝCÁINNÎcí¿ËårÇôî|·ØÏ;wãÆÏ<óLaaabb"Ïó(öƒE饆Âsš•sFï¿~ÌŠÛC"È`¤¬¬lõêÕul6ÛÏ~ö3wLïîÀw_¼xñ{ï½·~ýúŒŒŒÔÔT•J…b?(”kéÞ]Lú‚5“yE’d±ñž}Ak2Wd«öî.h”œUy;irfö3þ,E%±µ¼0¿­›Š•µn¶‘eʰΪ3;ŽÕwôΊçⲟ·[ì.™Š%ùµžŸzfÀóÀh¢´Ñkbœ·ï:zËD;"sÙì‘1°|Žkçá6&p‚Œ¸X&kríΣDÍ%Šæ-Zû,¡„cÄÚ ùûËŸL]¹LµgO¡Qê‡]k¿Ø–—2'{mC)!RCÑþ2ôå•WF(eEf\dAG™ö˜¾=¬€§žzÊáp¼òÊ+¿úÕ¯¦OŸ.ŠýcÙ–óñ؉“ï8š°dÉ’I“&a!È#KFFFffæ믿žœœÜû•K–,)++ëélNNΆ z¹ùJa³Ù’““+++Umð2ëÅÅÍÿí”éÁ®{첆Q³L¦¦”€õzþþS6\AñññX÷ .Ôø(½ÜRœ»Ý©]®Óš„ì9blc«8êoVXÄÍòC²Ë?±d¢å@ÎG"€*rÖÊ…cMÛ/¶¢Ö#Ê999ûæ@,ñ-;©Ulm‹¦³»÷—Y”Á`W2._nt7œµÅׄáÁ8v‚ ‚<ô˜žJJŸ›ÆÉ¢Óeo(:U×Y¾‚“RgO å(a8â¨,8”¥¹cã6Æ—¾(}”^²9dÅa² ö3[W‹lpÊ¢ùㆧ«×M‘dÓÅÜ}EÍí›Úzv•Si‚Õ¥Ÿ¼—ßèµÛ-4qÉÊÙç÷Ý6.‚†cm׎䞮nÑYÿ¤ÌÌÑÇq‚@Ì×ó÷Ÿôô“÷Ë.£‰LÎLK ”‘—Ž+òìØÛ§WÎêKæ¹ÉÑ7NÞ²)¬ßˆäè¦Ë—D¬€‚ ÈCUzFŸ¸pqtÙÎœƒ-kˆK[œÅŸn;ë7nivLéö͇LðAã²²³è–Ýe6÷š ªØù™ªÓ¾SîPŸ¿*Þ3ËC6žý|cé˜Õ ];¶]í¶]±”Ø\|ÔÜûÜç÷ûò¼á'?Ø’g–Ñ'.]•ùñjpÊ-%‡v”xÖu`RVfD×ì)wõ×.6cEÍÛ¹©Ò¦°úØÔìå“ön-h’}ðJi-Þ÷Ŭç^z•i´û[Žþí“k6\{Ay¨JÏ II¶ÝVÚâÓeóµü¼ØÄy¹æ¦¤8o*6y‚x±©è@~쪔ëù 2€dªh 36ÎU^ßdjuTìÿ¨bÀüvV>rÅ,»Å¹¼Ô:1‡6M%êФi3ÆEjE].~ˆá"Ðïå‰.>5ªxÏ–JÏ<:ÙróxnØêÔ˜KŸ]wôí£OœŸ¡:úç·®™e.0iî‚9#>;pÃŽb ‚<<¥g !òí/½%RlºetyÞ´`uACSW¬™î%†¬šoÔ°îžñ¦‚-›ü£cã§N‰3æÒ£¾¬qç^‹1REéø2õɆs{>=ftQÖ0vÕâ1ÈêƒÙº²NÓåesµI¦a®;”>¼ÆôåÔökf$Sɱ³ñKât7‹,(õ‚ ÈCSzÙÜÈŽJ*ÛÅžŠjÝߥ¦ŠšŠÚÏr+zÐn"tÔÒZsíB͵ '€Ñ%f?ZõáÁé~fˆ LŒ5ÙZjlëÊw9%: oÊ£¡cj[”ñtí¾ÌÜ#¬@d¹CÖIfõ,@¥GAî'½Î½— Ïi2çŒp·X¿øŒIºvåk:Â1mÞ„PUÛ œaØø”zOššQÙkÓ†ªÛOêtÔå5ÃM¶9ô‘|[³ :.J×ÿWG‹®wO®gô±i«–Æk;-q¯v©µìxõ˜E3¢µŒ§Çcdú¢˜«Ço:|j'˜JoÍHˆzèÔiÂåëf|ÇAy˜1=(ÖÒ½»˜ôk&óŠ$Ébã…=û‚Öd®ÈVíÝ]Ð(9«òvÒäÌìgüYŠ(Jbkya~[ÿ6-&:*kÝl?"Ë”aUgv«ïè‡OÅe?ÿo·Ø]2Kòk=?õÌ€çÑ„Ei£×Ä:)8oßuô–ˆvDæ²Ù#c`ù×ÎÃ7lLà„q±LÖäÚF‰šKÍ[´öYB Ljµò÷—?™ºr™jÏžB£Ô» Ö~±-/eNöÚ4†RB¤†¢ý;.eèË+_î*˜9ÿégµ˜¯æí-jA¡GÄ'¨½æò•:O÷*#B¢†Eø ¸ø<2JbSÉÁœ¯—ÿPêÕ°ß.ܳ¹°Û_ÚËvléu Å|åÀGWº=ãžß}…·Ý8¼é/‡½‚å ;ß½°Ó+zn.Îý¤Øëþpiì¶å·`×Æ‚{ð ËãŸÞ8Ž•A{膉 d—¹þæÍŠ*mÒˆ\“¥GAµ †à@U]£C¢À bKÍ­Ûf—†!Ƈ8j¹uù p54XDFJëÊ«šl£ ‹®g (öƪÊZ“MV=,Xíª½rÕž8"€' Xn].'Ó†éIWØ›ðXÁ` ‚<>PE´6™œ¬NËÅV{­¢U34qü„ñ ‘¼©üV³D@q5·r‰ãÇ'E²¦ò«7Z´Ñ£'ŒSªœKõj»!fÌ„ cbý·oܶ(ªÀ ÞÖhqOVlÆVÒ2=™@PéAÕxkÅ¥óç/\,¾Va#†ú±Š½©Iò‹ ó¨üÃÃk“UÂù‡‡èXÂþÁjBüBÕ,0Zÿ A´ºÅÖÔLý#Ãô!œ>4ÒŸ¶íTðìÆV‰ÈV£™ Ò0=š@°÷Aä±€è†ÀˆÖ†ò Ö­Ge§([*.3€RÔ’BY ,ïÙÞƒ†°<ËPÂ0”R ŠKaô¼§žp*V±È”èýƒ…Zc«`o43#Ô ô`XìÀG¥GAîƒÜ3¼>4:¼¹ôvC¨>BÅò<3:ºÓ«ÂŠÅ‡dFqJ€•œ2ó€÷ Q×4µÚØ ®a”nM xëA3µ‚Âý¤Æ:³D4!LËíÚV—B*’£µÕêÓ :£ ò'-Õ V@¶6Ô´€_°† ¼_°Ú^[]káÕ˜{5`L ‚Ü«Ôs†ð`¶¬¶Iô 9¼¶êö•"§L^­ŒŒ6øåé£b#++o\ª‘ÕDŽˆr/8F8C°¶ò–EèYÑÞ£ •AñYÛ5I¼…ZÙöoÁ/b„_D§«õÑcF·}gýGŽóoï ŽÜÖ‡ Oé²i'áƒâ&ÝчÐÕò8½÷Òõ©Ê²ø— *}|dÖKë§Þ¿G#>ãÉÌ(~@“ì‡Ï÷=¿½ÂúÉzfý«ÏÓ=(‹+¿÷Å.‘ùtæжôÖ{ϧ,]5?TW™eªÈRKYÞ¡sµÎ;&c(Žú›ñ¾I[àÄy ¹§Ô@|¾Ïùí¹¥8w»S»\=0Éi²çˆ¹»n8p~–]±¾èö‚Å)Aj’»6ÂýlWŽì(´4Ÿ ‚|”^6žÝ—7Š—?ÛRæ`4C§/_9íàæ“ 6ž•›ÎîÞÿ"»ˆ©“áìGòÀêe?|¾¿ù}ÀN­RB~–] å¯O:»û–ëÎsÎÚ¢ÂZÇ#ùª(}—Ë^u*·lÕì„¢O‹- }ï7•&X]úÉ{ùw¨1럔™9:@à8Nˆùzþþ“¶Ž§»*lâì̱AŒ$+Jë•“7G­ÈP•ŸÙ³Û³åœ!2e”íB¾¥s?£‰LÎLK ”‘—Ž+ªoëj`ƒSÍ74Ìyð¯;çÍaPë…úã[÷–´*}úÜ»W½ü– š¸dåìÀóûn Áñ¶kGrOW;¦¹””>79Œ“E§ËÞPtª®ÓY!8)uöÄPŽ†#ŽÊ‚CùWš;Ê1Ä¥/J¥—lYq˜l‚ýÌÖÃÕb{A §«×M‘dÓÅÜ}EÍíÙêO~{¾ûý²ÛûÝ÷å.PËóæÌI‘­r>A¥ïµ”•qéCøb‹  }ï7>jîŠáÝ]\rhG‰G™€”•Ñ5{65{ù¤½[ Üý±²ñìçKǬ{nÚª§Z/íÙöqƒ‹Q´Š§Ñ›Ï}yÕËoå¦óûŽ}yÞð“lÉ3ËÀè—®J‹üø@u¿»Ÿ}âÂÅÑe;s¶H¬!.mqzºí¬ß¸¥Ù1¥Û72IÀËÊ΢[v—ÙÜe¦ŠŸ©:ýá;åõðù«âI{ŸÍçKǬ^èÚ±íj·-’þä·ç»ß/»½ß}ßî‚£¶Â•À•×Kø@•¾Ë3Xt€š÷m’Q‡&M›1.RË(’èrñC 9:i\Ë‘­7Ü«'+ö['÷_LÊêNåt!ç5»wPGtñ©QÅ{¶TzúdËÍã¹a«Sc.}vÝáu™T“·ùP…€â4[|) _½ê gÅá#W̲[¬ÊK­yè·ÒsCR’íG·•¶¸eI6_ËÏ‹Mœç‘k.lJŠãð¦b“G´Ä¦¢ù±«RB®ç7È’©¢%tÌØ8Wy}“©ÕQ±ÿ£Š«=½å·—»ß|¸û}ÞÙnrjõ ÔãATú®ÊË«ÁÑìËî\ÈÔ'Îíùô˜ÑE\HØE IDATXÃØU‹ÛO²ú ¹ºÉëé+µT5¹TÝ$CeJBhÇoƒÙº2›·²¹Ú¤Ó0×GŦŠZûÝ-ää»W=B;,REu¤XCˆ|ûKo‰›n]žéã¬.(`hêŠ5Ó½ÄUóÀÝÅÑT°e“_xtlüÔ)ÑaÒ\zôÀ—5Ϊ?=å·×»ß¯Òðåî÷qa•q‹ä>‘““ƒ…€<ÎJÏúÇÇK7vû¦²‰±Æ#[Km¹.§D…Ž‡³‘,\®j0ÎhÐØ5¹µÎ¦Ñ1ÐÚþ`–-F%,BÇÔ¶(ÿˆ@§ÑÞßg·Ï^Ýk<ª:yZ¢Ÿ\{þÄ%£ä»Wìø¡¤²]ìù è`¡ÖÓiª¨©¨ý,·¢í&‚AG-­5×.Ô\»p]böÓ©U¬¹¯×½ßý~Ý£¸û¬>Dm«3Ëøç PjÁùx‚|u|Ñ6£Ÿ½pÒ€0rÍ™½y¦DP6ß÷Å#çÕ]¢˜Î|ðw¬p©îÌ‘:,APé Š¥ìà–G®ÃðÑô AAî1‚Å"@ATzAAPéAA¥GA•AA€Ã"@¤[Ž=Š…€ 8DéùȬ7ü·S¦û´ >cÉèŠ]Gn‹—d?|¾ïùíÖ̼…3U§Þ}¿Èú@,>¬üÞ»|D擉Ww©ê~ÁE.hìܹã9†e\Õ_È»ÒüßÂþ…^À')‚<²ÔÔÔ˜Íæ~)=œ²tÕüR]e–¨"K-ey‡ÎÕ:ïØóEqÔ߬°ˆ÷+'làÄy ¹§Ô@|¾Ïùí¹¥8w»S»\=0Éi²çˆ¹»n8p~–]±¾èö‚Å)Ayí£Ð£½xaØÅí97m pã/žeÞ|´Zħ ‚ 5½)½l<»/o/¶¥ÌÀh†N_¾rÚÁÍ'::rÓÙÝûï_d1u2œý¨q`‘íÏ÷7¿©UÊCÈïò ²ñB¡üõéQgwߺ3¬gÔÜ̓Çoº7’LÅ'ÊŸšÊWßF©GdÐ*}—Ë^u*·lÕì„¢O‹Ý×zvSi‚Õ¥Ÿ¼—‡³þI™™£Žã˜¯çï?Yaëxº«Â&ÎÎÄH²¢´^9ysÔŠ Uù™=» ;mÝ.D¦Œ²]È¿c£\F™œ™–@(ýÿìy@“WºÿÏû¾y³C@‘Å@”ME°.X– Šˆ".¨­Ó¹S;K§·3sogé8ít:3¿;ÛíÌÜöN;ö¶uW\PÙDQdĈ@!, †„¬ïòûƒ"BÄÖóùÞ÷ä<ç<'ç{ž'ç}@HÅí’Ò†>Kªsؼ1ÌÛÝPøq–!›Çì»r:·iˆšÒfÛVÙ(‹ ÷ìˆw©½Ôº(Ì(Ó¶\λ1C/ÍÇ…Á±V¸3H“Á¨ëo¸þèû\™®ÁÑñá 4‚2}gUQ¹uÒåûÅnŽõçZ=IéµL]åéânÓhGùXýwìp¬½“{ß¡zm{ß/КÖZµ$Ò“ÙÑ>Nꉆzën¸"ê;ðùäyRz©’J±óñF(´à¤à^Ò|&º¹©è\“YQAÄŽ8QOŽyvEù¡É ®µg¶SåˆÖ¥§ùªŠ?<_?î§g†@ìEv\×?"ô¸{TZ ]–u¼SK€ñÄÑ©Û"sOWäcIEÍ…cÍ!û¬Iß9t;'óH¿eñ(ó”mË橬²Q–TÖ^* }#Á§âËSej ¼ ”ôÏ#Ž'Q^Цd‘4ëD¡ŠãûÅ$Çzá7,WÃRR}›Ïž,$¸0,)5‰>•-ÕŽôK¼QºñÕ'íz`ûlL@Fs6Ž5‡ìÞd<—yo‰#íÜûÕkÛûöyA/—£ÅF{ßä?Â3\#6E>¼rV $r,…ÒŸÿÎ?¬W4l½på§ÌoVótuz-3öï½à4sYõæŠï‹s¯ý»˜ñTË:Ž©ãä[oüé¬úG7 3æ#O¥û稽³R¯®æ—Îoþ¿wWó&ì;Swîïò·k” 6þìOoƸaÓQ‘iZC™ô€ÛW a/ŽÛ¶gïÞ½{víÜ4ß•géÆ‚È0Õå²Öaj$YÐQ‘_¯¢'²ëÆ1 ꬧[„íÕ˜SÑiΚ¶+y²€hßG¹Fˆž²“Y·úÊ ¶ç÷^»­š ƒ¬øò]59"VíÍî.3pc~Ä ]Iq³jD–HuKyYšGÃ}U„¾8¿qÐ,Z&eCA9º*btƒ2Õ‚P?O7g6ô²üÃ3–Ž¶Õ^Þw;¼?¥HÝ ãÆ›|c‚ÄnõÊ» à9§é¯ÛsùK¾[:¶çgèò«‹ùÜÅÛþvÇŽÞÁæo|ÿøù¬Ó¤yOäþÁ ßþ^Ñ„Û‰(ÅåŸE ]–ozí³#0¶|þÚ¦å.nÑ?+V<;YSëço]ÎXî4“Š ÂÖÇ.qBžvÙÈ<Švý÷—¿‰ÌPõ“Ùmï\ÕË Ý»ºôÿº?áìl”~ô½ÿ[øî™ÜÜœœ¼?$þøƱ®i!Ó|Qœ ôíù®1ÜVoOäßÊ9Sª0Ò`üÐô䱞â Én¥U“U—ÒÈšàch’FP€+ëŠõJµÖ6êîA¶;½¯û¯I)“ëèé}Oì¶jRè±iŠ¢gä›Ïw#Ü´N4›” £98¸BwtZÆZ+1ÄØø`$Å¡¬:uÜy¡H°z•ÈkسÚÞ¹ª¦ÿK™ ©üÎGë¸ã¯ 7•‚´O8á/ß•¤ùý=Ý~oÞl)=6/ €hͶ#&Ä\‚Ċ˧›–d®Ñ@Ð̱ÉY…¹2ïŒ>ëĘç-dLüêղݸ(uI‚r÷à¢r56ý{¸:G'f»­zÒx”ã½rM3)¯½v[AØoÕ¶ÌÙÔ9*ö¸PäÊ”›#JYL~>O6‰v#L>—Ö õ´Ôõ´Ô](7(uWt×W…=³úô˜mï;ä£ð>Æsck{Õ-‰™î«RÖ;מ»tÊüXyn=ˆþþËæôŸ…°Œw¿<æúÝ_x}š;iüóoüO>ó÷ Y›©íð›?úâZ5ýóÚ‚]®S<Æ{Ÿ¾þ“cׯÔb;6ÿæ’—þú— {ó`ä`õg‡Þ?/#QÐÌ€=３ÊG5Pð‹ŒïuŠ Ö´ jÉ€W¾/.ùWQ·š}ëèÛ=¶ÊšçzuÃo:ÑB²X †×ÖŸÄgï{wHòúÿ|t0Àzm3\ýÏ×—~æŽÚc•¹Éÿzý'GoÔó_ó©ó?ýCჇòáÐ_ÿÛ/€”gÿüõo«•Ò‡Û³û¥Ã¶U6Êïñ½Œ·[¿õß*ŽÖhIgpÛô›^tž‰”ÖµœyïçŸÕëq'>ß%8ãÍe\–fþîÐÿ5èQÊd \Ö~ïw?OñëC¢'ï7¯ÿ&§‡í&À1—Åó5Â%Çž±áH{M§õΙ6A¯Ó Ó ?ýóOb\1ûƤmÙð¾=^@Ý7ðxç“›šuqã%œÿÂÏŸ}~mýÖ¹1Œ ÿuÅÇîi¥’¦£ô(W“Øyé„ÆŽX•Ò«áBÖ5H€òÄ/¦¦0ós-.ç§Æ,9wù¾†¨“oT¼?›îŸHéÛÚèäE<´g4‘@K¯tïÞ¥8­CK€ñ—Änö½—[¥wtÔÚmÕ“¦g|bcýµ•Ÿ•Ó²ªúgÇú¥Yw`Îq‰Á\ʬô¤²öš~[Âr]A½yWƒ¿(Øo­mÕPŽêÞ…¥_wŽlu`p\Úh¥q¤VϹà÷ä&Â,q­•9ÛÞw¤Þð>ÊõòAÚóTã•a{¯Ûº½~¶¨c$„/Œ ¥ênõÏ»Ô#ÎQßý讪þ¶òÖGåQßÿÁ¼‹ŸŽ,Œ^ýçQ᎘W âýžÛš»÷}Ìw>9¿ýTúyŸÝäúö§ç£\1L½Å¿ýö·>ùë±ïú1ÝÞýCzÁ›ü_^ü4’*:ù¶áhAÁ*ôê÷mIþ ¦­²²ûÄ÷ߺ÷­/.¬_À¤òêíÙ_êó~ã§û<ªŒ†ö¢JÖ‹oº öY5Òäûäü¶SÉ’¿fìòÞó᱋K¹äP7ÿÒ†-LþÃéd »ù³ýeM„SXe£,sÉ¿üêØ’·J\xòmO½ç_Ýý~uîVr^w÷f½ñƒò-ÿ<ó{ c÷¥÷¿ÿ^¥î‡–™«ëØ«ÿVºýËÓï‹Ù u-ÇÿýÛÿŽžø(Ùm¤ÏÔÅ?yGõF~eì<ZUöÖî‹”½cÑö⢿ý,Ýüû'Ñþé¾wË#þËs´^ÛÞ·Ë ˆ <–÷»âvC\è¸|Ãk×_ÞúñAbÓRa—tᯯ]Ht›VrkŠçécÄ‹‘]û^ ¡TwŠ2kF±3ïÆÊq÷reˆ 40<¸r±¤C­n*hHؼw?B# Ô$¯+Ïoß½c++'§ZAJݘoHÙ…Fƒ^y§´´#ŸPçjÑ” AµÕóÏ&ùÕ̲ˆõ©{cPšF¢¿!ÿ\½‚´X™¼q©‹ÀÛ²µû‘Ý6mžÂ*[e§Å’­ñK|Á¶õƬâV-ê²<1ÎOŒ&­”gU™ãw–G¨ˆahÎ+ëœ^îœnνˆÆ&f¬Ä)‚ Mu9—„’´TVnvÕaè*Ë¢WHR÷ÌÃ@™L„i¨½ºÜ’ߦMšAÚ?i_¼3B’4Šº*Ï•Zi—AV|Ý/õ•ƒ¸N£3’´i ©\N9ÜÞ)¼ÿäõÚö¾=^¨sà2FcöãúÍò‰[»ï‰NÛkùÄy¾©ð6TzfÀþýª7ÏTuåìùsëáŧSïpÕoölýÌ7ºêÙï[¾}W¹óï'£Fb1€»Ç¿ýAýöß—dükã<kIÊŽ0> #¶¦,wF€q¾7S¥¥¦,k¸ýÏ£‹Þ;¾~0áºÿiÿéM`›©·Yéœäj=iMe•%=Ì^ñËÌßE» †³‡»a·U“áóÛߤx‹Çmwÿ¿6-pXé Mÿü—ðÝ£©¢Ybz&þüPñù·Ìßc}ý‡Ÿ ~s~§˜mÎhúeüñgE»ÿu7á§K™À^+jûÃz!:#JO*jÎ}\c£e#;¤'-þ°1ïh£Õ?ê>¸ýHö®¯6çXíh¼ŒLø»6Ñ_uU±Ê³&o,} (݃ª‹Çª&¶º:ëXõÚlÓ*[eimkññ[eqê²þY—õhb8Ä—ñ ¤à®fú?ß›”M…'š¬þqçƒf`ÝÕ9''n³Nzî”ÔæøTß-8|wº}5e{§òþÖkÛûvxà/„i®žx\¾õwÿñ.ÌÕO’Y¸é{žqþ²êË<1ðð)ÕÊ]yè“Ìm‚‘?”™–ä¦^©),ù‘ÍǸçÊ%ë•ÄÆyS$*m—5ô´°V.±JŒ²}Vûq‡&Xý`LŒ"™³ì³ áúÅ.w™Þn»­š|±†Œ)ÂÀPh7û¬Uß·N4sü^ àšS:¦þ{²ë5û¶þÅê jz¨ Tš`¾äµçºê®_üÛ‡WëP¾ÛßýãwWðgfŸÛ¤í54ÿí¥Ÿô|û¾:äÏCmê>±çû3R£=ÞŸÊ 4IÐþøžzBvæ+Æ›_$y2lqÚ¯^½ðZ^ŸdïB»‡Ñ3òÞ{†KPÝR¶5§#MèEÜ(hÖÀ_᧠oõÛŸþcŸ×:gƽ‡€¦Úzú$…qwFý­~2|Ñè¬h꩹ïàÂp°,Ë+ÀxäÞpÚj˦(½ìFËpп>x-s{ØÜk^ØLXe»­zÒxtðÆÇÿ}¶“þ­Ÿì `ÛoU á«fÍŽ¨Q±×¶”K‡—›#KbWÄ.ÿü¿cø“å*{úwïI/¯Hzù?Ùwî•¿[}é÷lÇƆmŒ÷Ï]|ïØ6ÿ‘e‚óœY¨f®Ç¤U¢hÐm™×c[])ã0Éd¡c‹6ÓÔcœŽ•s¹ålØ™‘‘‘‘±ww‚ >ïÖà$"HÖ^ºú´dÞn«¦ý]ªüòóËÝF(s Ñ[yùöC¸Úz0×ðØÈùÍZì…^ÊëRíÈ¢¼ñ«Ÿþ¢hÐîaºÎë­jÓY¦þö«—*ûíqº å/œ|ýOW͹cwÁ{?,Ùü‹x£eYÁßÙ×óÎoòå& Jÿ|(G…NgâÞñôê5913VM¡©öZõ„‹­ÁÒ÷~ü›ÿúRîíÁš–Uÿmð·ÏÊF6É»²ßýñé>KŒŠ/ùÖ >øÏ/Õ–_ Ý׎ü«ÈÜc´2û•”÷oXvÖÒú>Y?Âc!©Â7«¥v¤Ñ{ù·¯e?òà朌I‹ÒõÝ(£ãÖ/zìwlæâÔõ-ü¬AC¨‡×ÿöf{¢çtž§ŸË˜žÒH mg–¡UÈœRøî[_ß/ÚV¿ù±Æ¡Õ7~wðÝÜ×» wü·ë]Q§Uo<þíMPB±Ä[þ02ç¿Þü$üÓ×X€”_üùëÝÖRÙP9pukEœWÿü¿zÑüÐ5?æý7óþ-æ­p¡ÇDœ‚Ò¶JQüÎw”P¯)ÿþ÷o‹:>ÿá?(®Äâüë×! 8Ëß:òö?ßþö–÷ ¡iöÒ½übŸ?j àÝ·¾(l¯ú¯ÿr0µ~õŸ¿9þëÕÀÔqò§#6OZ0þY÷óƒÉ6±yοí¿:´îwº º†ò­]¦ƒçd¯½¹dtf¶ùÉFé?ðgZÛ¯[¶v³–¼ôÁèÆnÛ}eÛ*[e)Eѯ¾s¨ „þÖÛ¼O³ÁhýòÇï^*6½ùÉòO_3?L ®9^®ç§~ð˨éåα[ÿö¿Ä¯œú-ƒÍf:½ü?¹¿õÐþï¨þöÑwYüÕ¿üýׯ^ÙÞi¤îäÄâxÇü©y±ˆ`N‹ÑœO~»‹b1“Ñyí¿82¶Ò˜tl8Ô^Ô#ý?ýaò&% 7üÀÏþ÷Òïö½:ô÷ÍVáYvã¾Ñ·k«PØ¿2 †Ös§¤Èy *êžú_ßGË¥ºgʪiCôÕ¶þ죃Pæçýýì7þ°}Z2o? ØÁb )Ï~ûß?iÒÑM2W¼þ×C‹'AæâþÏgΪ銀ø¥Ðés ;ìõ_‡ÍÚ§C¥‡@ GÀ&ÿ¿ÉÐ*È7˜½‡@ *=@ çWéqϤ×_]ã2{‹|aÔv‰×ŒnqÄæYo¯M°y!I{^}ó@ØSÛî;Wí•zqÉ.‰÷d‡í N‹^HÞóÒKû÷ïùåôØ%|¸†@ _¦xï}JúÆ@¤»KM@S$¡’–Ý’ƽƒÒ÷µÉ4³öbÌ%°áLãÈ)-æSÇXWvóÑÏÊÇ©16/X"Y*`2 &Qß/ϯiÇfw–{x¼$Tˆ$E Ý­hóO‹cµWædW?rt;Ó3Â_[W>îT”ã¹B(@h ¤âvIiCŸ%Õ€¹FlÞæín(ü8Ëž°n1ŸÍcö]9Û4DMi³m«l”Å„á[vÄ»Ô^j]æ”i[.çÝèž™—æãÂàØ +ܤÉ`Ôõ7\ï}ä*Ó58:>|ƒFP¢ï¬**¿ûp¬Q¾_ìæX¡Õ“”~PËÔUž.î6v”åпqÇþÇÚ;¹÷ª×¶÷íñ­i­UK"=™í¶$ã/Y¨¹Ûd‚“yŽ”@ª¤RFì|¼Qc`ôÔ1ÜkCšÏD77k2ë/*ˆØ'êÉ1Ï®(?49ÁµöìÑÖa  Ѻô4_Uñ‡çë‡ÇÙ'{‘×õ=î•C—eïÔR`¿ža>‹± 2Luùtë05’,è¨È¯NšHå¸nC‹Î:¨C¸Ñ^9§:ÍùRÓv%Ï}w´ïíó÷õV·=e'‹d:@Ô{:Ã^«&à +¾|WMŽˆU{óp¸ VzÆüˆº’ÌfÕˆ,‘ê–ò2qP‚Y®î«"ôÅÇÍ¢eR6”‹Ó#Üî—Ë@ ÊT BBýŒí}ÊÁ!½,ÿ°lÆF­öÚð¾#Øáý)½@ê œ  ú&^Î*oý¤.Ïcy¼dcÈÅœÆ!ø;=y®”ÅÙ@o×%†Ûêí‰ü[9gJFŒš>öÒŒ'$»•V³/¡êR':F‰&iE«s1ž+Ö+ÕZÛ@ª»Ùîô¾~ì¿&¥L®›^@f¿U“bu~'MQ3b|7òÁMk‰4);FËy\¡À;:-c­•bl|€ƒ0’âPV:î¼P$X½JäÎG6—Üì1ÌÐø™¬½6½ïPoØãý)¼€ (B“¶1¥ë¹u©ˆ¿sÕBia7ÌàC çHé±yDk¶3æ$V\>ݬ°$sðÜø IDAT‚fŽMÎ ,Ì•y§Ë"`ŒyÞBæÀñÕP¯–íÆEÁX`Ej”»•«¨±éßÃÅ Ð9{ÙmÕ“Æ£ï•k‚œIyíµÛ Â~«°en̦ÎQ±Ç…"W¦Ü¼QÊzdòóy²I´aò¹´f¨§¥®§¥î@¹A©»¢»¾*ì!fsXÙö¾C>šïc<7¶¶W=åOÊ g òõf:Ï £\qLr`gÙ]±*¥WÂ…¼‘·õ£/nÛ'Äi’@ÛVq® =ùÚóÌœOÏ Ú•Š]<Þ¨}ÜD^ÈÎŒE_äu>ýIwr«¦S´õîW?ú¬BAÂ7gà^ /GuŸÈlÔ|ý7ÚÁóé!眯×ùô — ºeâ‡ÏiMsÁMZäŠ?SVM”»P (Ï®†2?·#MèEÜ(hÖÀýôäù™øæ°n”¿)r>%{*sË'Ar°öÒÕgΪiB V~ù9ps Ñ[y¹vJÿt 4ÒÂSÒg­GžM« yÂv@ Pé!@¥‡@ •@ Tz@ Pé!̺ÒãžI¯¿ºÆeö ø¨í¯}sŽ#6Ïz{m‚Í IÚó꛸O«Æ¹jï¬Ô‹{HvI¼§8lå-Ývð‡?yi)ÎäëÏï½OI߈tw©IhŠ$TÒ²¢[rø3_(}_›L3k'{b.á ýy7f´GlžåöN©jÌ;kpÚ6C"Ä L]oÊ»ØjxÊí«zM} “#„‡­ÎQÛ'*´/·ˆE]¾Æ[¶lsy¥'5—Êüqòü)©PŽ÷Úm;Öž¬èätRY“?{‘Çê• æðÀ̾DÖ›g·½O„ÁfQsÐÞ¹ªŠºjòåµ^5ÙŸ£Àô\½Òtí|?×ëkõÛÖ”¯µ‡“•ÞžXV×u=OšØp¦qäàZó©c,Ž+»ùègåãÔ›,‘,0 “‰¨ï—çWÈ´c³;Ë=<^*D ’¢†îV´ù§Å±Ú+s²«9ºéᯭ+wP.Êñ\!‰  4 Rq»¤´¡Ï’jÀ\#6o óv7~œeOX·˜Ïæ1û®œÎm¢¦´Ù¶U6ÊbÂð-;â]j/µ. óÀÊÀ´-—ónÌÐKóqapì†î Òd0êú®?ú>W¦kpt|ø  DßYUT~÷áX¢|¿Øͱþ•-ÕŽôK¼QºñÕ'íz`ûlL@Fs6Ž5‡ìÞd<—yo‰#íÜûÕkÛûöyA/—£ÅFûc'×¢‚Ðun…•:àOôä9Uz@™ô€Û—ÕDØ ‚×D…y:¡a2ñùüzF`,ˆ S]>Ý:L$ :*ò냓&R9®ÇТ³ên@´WcΩNs~€Ô´]Ésßí{ûü}½ÕmDOÙÉ"™ŽPµÆžÎ°×ªÉ0ÈŠ/ßU“#bÕÞ<•ž1?b…®$³Y5"K¤º¥¼L”`–k†ûª}ññÆA³h™” åâô·ûåý$ƒ2Õ‚P?c{ŸrpH/Ë?,›±Ñc«½6¼ïvxJ/ºA'ˆ‡‚¾qŸÍ XçÛ^žºƒ@ ϵң8èÚ32ÜVoOäßÊ9Sª0Ò`üÐôäÑ‹OHv+­f_BÕ¥4²&øš¤A Çʺb½R­µ ¤º{íÎAïëÇþkRÊäºém©²ßªI¡Çj¤)jFvta|7òÁMk‰4);FóöqŒ+xG§e¬µ,Œp0FRʪSÇŠÄ«W‰ÜùÈÃæ’‚›=†?“µ×¦÷ê {¼?…EhòñAŒ».q6í^:Òí"/:qEßÅ[JxÎ0yž”›@´fÛ¦b.AbÅåÓÍ K2×h hæØä¬ÀÂ\™wº,Ƙç-d<þ1äP¯–íÆEÁÐèÄLj”»•«¨±éßÃÅ Ð9ŠÙmÕ“Æ£ï•k‚œIyíµÛ Â~«°en̦ÎQ±Ç…"W¦Ü¼QÊzdòóy²I´aò¹´f¨§¥®§¥î@¹A©»¢»¾*ì!fsXÙö¾C>šïc<7¶¶Wý˜€eÙŸýŸåvÀÎTô”yòõg:»‹Q®8&9°³ì®ÆŽX•Ò«áB6R’'ŽIO pBFƒåþšz~\Ì’‘먓oT¼?›¦'Rú¶6Z´ˆge'=,½Ò²9Jä„šcÞ%±›}ï]iÓ;Úv[õ„p|bc_\ÎÒÓ²ªúG²~©`dU†9Ä%>LO*k¯é×$,_À²t.ƒ¿hYÄbKqüS÷Æx³G/r\Úh¥^¤VÏótÁ-ËÈÏ‹ëø–sÛÞw¤Þð>ÊõòAÚÛTPÃ!Œé1׈Äñbd×¾H@(բ̚ÑGìÌ;¤q€rܽœDb ®\,éÐ@«› 6ïÝÐ5ÉëÊóÛ·GïØÊÊÉ©V€R7fçãRöFa„Ñ WÞ)-íˆÆ'Ô¹ÚF4%HPmõü³I~5³,b}êÞ”¦„èoÈ?W¯ -VG&o\ê"ð¶lí~dc·M›§°ÊVYÄi±dkü_°m½1«¸U‹º,OŒó£I+åYUæøå*bšóÊ:§—;§†›s/¢±‰+qŠ HÓ@]Î%a†$-••›]5@ºÊ²è’Ô=ó0P&aj¯.·ä·i“föOÚïŒ$b†®Ês¥V;Ñ ²âë~©¯Äu‘¤MMårÊáöNáý'¯×¶÷íñ@—1³û&_maÂðM‰Ë¼}<©¤ˆþ 5 ¸$€@ _kÌGBÃWŽûo``à–-["##Ÿž!œ ]©ØÅãÚÇMä…ìÌXTñE^§ñ©wÏäVM¦hëܯ~ôY”Ž9÷Jx9ªûDfã7`ß]\\œD"yï½÷:4åóôOpCIIÉÁƒá˜@žYzzzÔjµíoôíÚªgäÝ — ºeâ‡ÏiMsÁMZäŠ?SVM”»P (Ï®†2?·#MèEÜ(h†Ûë!Ès4ñÍaÝ(/ ~Sä|JöTæ– N"‚ä`í¥«ÏœUÓ„¬üòs8à梷òr/ì•þé@i¤…§¤ÏZ<›VA ò„,ì@ ÒC J@ *=@ ¨ô@ ÒC ™u¥Ç=“^uËì-ð…QÛ%^3úæGlžõöÚ›’´çÕ7„qŸVsÕÞY©÷ì’xOvØkqúß}yoÆ{¶®tcÀ9|Ý™â½÷)é‘î.5 M‘„JZVtKnwæ ¥ïk“iL³e"æžØŸwcF+pÄæYnïªÆ¼³§mì™ù8N`êzSÞÅVÃSnï\Õkêkx˜!<|}âSêL½×Îœ¬×Ðpj€@ Ï…Ò“ŠšKeþ8yþ”T(Ç{í¶k OVô?r:©¬ÉΟ½ÈÎcõJPsx`f_"ëˆÍ³ÛÞ§ Â`³¨9hï\Õ HE]5ùòZ¯šì#üþC ç]鋱t]×ó¤éñ gG‚ó©c,Ž+»ùègåãÔ›,‘,0 “‰¨ï—çWÈ´c³;Ë=<^*D ’¢†îV´ù§Å±Ú+s²«9ºéᯭ+d¡Ï’˜@BÓ!·KJú,©Ì5bóÆ0owCáÇY†ð„u‹ùl³ïÊéܦ!jJ›m[e£,& ß²#Þ¥öRë¢0  LÛr9ïÆ ½4ÇnXáÎ M£®¿áú£ïseºGLJ/`ÐÊ@ôUEåwŽõ!Ê÷‹ÝëÏ#´z’Òj™ºÊÓÅݦюò±ú7îØ?àX{'÷¾CõÚö¾=^ 5­µjI¤'³£}©gÌ_µ}ßr%ûo]*lP˜à,@ž¥*©”;oÔ€ÒH NJîµ!Íg¢››ŠÎ5™õDìˆõä˜gW”šœàZ{öhë0PŽh]zš¯ªøÃóõÃã숽ȎëúG„wJ‹¡Ë²Žwj)0ž8:u[dî骑|,©¨¹p¬9dß5é;‡nçdé7¢,¾e^cزy*«l”%•µ—JCßHð©øòT™š(/(%=ÆóHA·ÃJò‚6%‹¤Y' Uß/&9Ö ¿a¹ê–’êÛ|ödÑ À…aI©Iô©l©v¤ÏXâÖ¯>i×Ûgcz2š³¹p¬9d÷&ã¹Ì{®HiïäÞw¨^ÛÞ·Ï z¹Ì-0Ú;¹Ö$¿vñâP§|˜›º%eÝÀ‰’¨õäyRz@™ô€Û·I a/^æé„R„ÉhÄçó뀱 2Luùtë05’,è¨È¯NšHå¸nC‹Î:¨C¸Ñ^9§:ÍùRÓv%Ï}w´ïíó÷õV·=e'‹d:@Ô{:Ã^«&à +¾|WMŽˆU{óp¸ VzÆüˆº’ÌfÕˆ,‘ê–ò2qP‚Y®î«"ôÅÇÍ¢eR6”‹Ó#Üî—÷“@ ÊT BBýŒí}ÊÁ!½,ÿ°lÆF­öÚð¾#Øáý)½@ê œ  úÞÚžvíèBõNùý/,Ä{:¡ÔC çIéQœ ôí9ò“á¶z{"ÿVΙR…‘㇦'^ÄxB²[i5ª.¥‘5ÁÇÐ$ =VÖë•j­m Õ݃lwz_?ö_“R&×Moc•ýVM =V#MQ3²¯ 㻑nZK¤IÙ¡0š·c\¡À;:-c­•bl|€ƒ0’âPV:î¼P$X½JäÎG6—Üì1ÌÐø™¬½6½ïPoØãý)¼€ (B“SbR¯¦XðATò|)=6/ €hͶ#ÆÁ\‚Ċ˧›-?s’FA3Ç&gæʼÓe0Æuo JÓBô7䟫W«#“7.ux[¶v?²±Û¦ÍSXe«,â´X²5~‰/ضޘUܪE]–'Æù‰Ñ¤•ò¬*süÎò1 ÍyeÓËSÃ͹ÑØÄŒ•8E¤i .ç’0C’–ÊÊÍ® ]eYô Iêžy(“‰0 µW—[òÛ´I3Hû'í‹wFH’F1CWå¹R«hYñu¿ÔWâ:ÎHÒ¦¦r9åp{§ðþ“×kÛûöx ÎËÙ}¯¶´÷.å¡kãv‹H ¡TM…Ùj Näk’yâHhøÊqÿ ܲeKddäÓ3„´+»x¼Qû¸‰¼‹*¾Èë|úÏ?OnÕôaŠ¶¾qÀýêGŸU(`˜8wà^ /GuŸÈlÔ|ý5<..N"‘¼÷Þ{‡’J¥“Ývâĉ'»¡¤¤äàÁƒpÌ@ Ï,===jµÚö7úvmÕ3²ÝˆáB·Lüð9­i.¸I‹\ñgʪiƒr åÙÕPæçv¤ ½ˆÍªC çgâ›ÃºQ^@ü¦Èù ”ì©Ì-œDÉÁÚKWŸ9«¦ 5XùåçpÀÍ5Doåå^Ø *ýÓÒH OIŸµy6­‚@ ä #XØ@¥‡@ •@ Tz@ Pé!@¥‡@ 2ëJ{&½þê—Ù[4à £¶K¼fôÍ9ŽØ<ëíµ 6/$iÏ«oã>­窽³R/î!Ù%ñ¶}ØÆ÷]»5c}ûÒ£½Xp–€@ _o¦xï}JúÆ@¤»KM@S$¡’–Ý’ÆùBéûÚdšY;Ãs OìÏ»1£8bó,·w HUcÞYƒÓ6öÌ|'0u½)ïb«á)·w®ê5õ5N±yÁÉR“Á`0™ˆú~y~…L;6²ÜÃã%¡B” )jènE›Z«½2'»ú‘£Û™žþÚºòqå¢Ï’˜@BÓ!·KJú,æ±yc˜·»¡ðã,CxºÅ|6ÙwåtnÓ5¥Í¶­²Q†oÙïR{©uQ˜P¦m¹œwc†^š ƒc7¬pg&ƒQ×ßpýÑ÷¹2]ƒ£ãÃ0he úΪ¢ò»ÇúåûÅnŽõçZ=IéµL]åéânÓhGùXýwìp¬½“{ß¡zm{ß/КÖZµ$Ò“ÙÑþ˜Ô3=#¼eåePæ!ÈóªôR%•2bçã#PiÁI)À½6¤ùLtsSѹ&³þ¢‚ˆq¢žóìŠòC“\kÏm¦Ê­KOóUx¾~xœ}±Ùq]ÿˆÐãîQi1tYÖñN-ÆG§n‹Ì=]5’%5Ž5‡ì;°&}çÐíœÌ#ýF”Åw¢Ì³·-›§²ÊFYRY{©4ôŸŠ/O•©I€ò‚RÒc<t;œ~FyA›’EÒ¬…*Œï“ë…ß°\uKIõm>{²hà°¤Ô$úT¶T;Òg,ñF ëÆWŸ´ë€í³1=ÍÙ\8Ö²{“ñ\æ½ W$Ž´wrï;T¯mïÛç½\fŒ íãO®E¹ 8òVƒ÷ÚÔu~}XnâøÆnßzfò!Ès¦ô(Îú‡öÌ| ·ÕÛù·rΔ*Œ4?4=yô"Æ’ÝJ«Ù—Pu)ír¦IAz¬¬+Ö+ÕZÛ@ª»Ùîô¾~ì¿&¥L®›^8f¿U“BÕHSÔŒDƒß|pÓZ"MÊ…Ѽ}ã ÞÑik­ô cã €‘‡²êÔqç…"qÀêU"w>ò°¹¤àfa†ÆÏdíµé}‡zÃïOáAšœ`Óz-gþpáçÕr@Û~ízøŽçƪ‡Pê!Èó£ôؼ€¢5ÛŽ0s +.ŸnVX’¹FA3Ç&gæʼÓe0Æõî™ÜªéÃm}ã€ûÕ>«€â1‡à^ /GuŸÈlÔ|ýcõ¸¸8‰DòÞ{ï:tH*•NvÛ‰'žì†’’’ƒÂ1<³ôôô¨ÕjÛßèÛµUÏÈ — ºeâ‡ÏiMsÁMZäŠ?SVM”»P (Ï®†2?·#MèEÜ(hÖÀ”<y~&¾9¬åÄoŠœÏAÉžÊܲÁID¬½tõ™³jšPƒ•_~Ü\CôV^î…Ý@ Ò?(´ð”ôYë‘gÓ*@ž0‚…]@ Tz@ Pé!@¥‡@ •@ Tz@ ³®ô¸gÒ믮q™½E¾0j»ÄkFßœãˆÍ³Þ^›`óB’ö¼úæ0îÓªq®Ú;+õâ’]ï‰Ûaì}óµý£¼ôÚ›þ,8I@ ¯9S¼÷>%}c ÒÝ¥& )’PIËŠnÉ ãÎ|¡ô}m2i¶LÄ\ÂûónÌhŽØ<ËíRÕ˜wÖà)zTë IDAT´=3Ç L]oÊ»ØjxÊí«zM} “#„‡­ÎQ°@<¼såüíj¹¹J†GüÎ 5|¥!ù&+=©¨¹Tæ“çOIõ€r¼×nÛ±¦ðdEÿ#§ƒÊšìüÙ‹ì¿ž€ "ÃT—O·S#É‚ŽŠüúउ&\®ÇТ³ên@´WcΩNs~€Ô´]Ésßí{ûü}½ÕmDOÙÉ"™ŽPµÆžÎ°×ªÉ0ÈŠ/ßùe—Ò´7‡»àÀa¥gÌX¡+ÉlVÈ©n)/%˜åšá¾*B_|¼qÐ,Z&eCA¹8=Âí~y? Ä Lµ $ÔÏØÞ§ÒËòËflôØj¯ ï;‚ÞŸÒ ¤nÐÀ â¡ ÏÆêÊw±án£ÎäùSzgýC{Žüd¸­ÞžÈ¿•s¦Ta¤Àø¡éÉ£1žìVZ;„ªKiœh›3MÒŠ Ðce]±^©ÖÚRÝ=Èvç ÷õcÿ5)er=­ÆÙoÕ¤Ðc5ÒEÏ„‹0¾ù঵Dš” £yû8Æ ¼£Ó2ÖZ‰!ÆÆ8#)eÕ©ãÎ Eâ€Õ«Dî|äasIÁÍž™’°ÉÚkÓûõ†=ÞŸÂ ‚"4ik#N"?òn®NäùSzl^@ÑšmG˜Š¹‰—O7+,É\£ ™c“³ seÞé²cž·9ðøÇC½Z¶C£3©QPî\T®¢Æ¦ƒBçè™ãv[õ¤ñ(Ç{åš gR^{í¶‚°ßªl™³©sTìq¡È•)7/F”²™ü|žlíF˜|.­êi©ëi©»PnPê®è®¯ {fu§™mï;ä£ð>Æsck{míªg{ ÷ ´p~€@ ߦó°2ÊÇ$v–ÝÕØ«Rz!\ÈÃFJòÄ1é)NÈh°Ü_SÏ‹Y2ruòŠ÷gÓôDJßÖF‹ñ¬ì¤‡¥WºC6G‰œPsÌ»$v³ï½+mzG;Ãn«žŽOlbì‹ËyCbZVUßâHÖ/Œ¬Ê0瀸ÄàчéIeí5ýš„å X–Îeð-‹Xlé1ŽêÞoöèE®€K­4ŽÔêyž.¸eY ùyq‚ݶ÷©w¼r½|ö6iCèñû²aÎä›Óc®‰1âÅÈ®}/€Pª;E™5£Ø™wHãå¸{9‰2Ä\¹XÒ¡€V74$lÞ»¡j’וç·oÞ±••“S­ ¥nÌÎÇ7¤ì£A¯¼SZÚO¨sµhJ Úêùg“üjfYÄúÔ½1(M#Ñß®^AZ¬ŽLÞ¸ÔEàmÙÚýÈÆn›6Oa•­²ˆÓbÉÖø%¾`ÛzcVq«uYžç'F“VʳªÌñ;Ë#TÄ04ç•uN/wN 7ç^Dc3VâA¦ºœK IZ*+7»j€0t•eÑ+$©{æa L&Â4Ô^]nÉoÓ&Í íŸ´/Þ!IÅ ]•çJ­v¢dÅ×ýR_9ˆë4:#I›šÊå”ÃíÂûO^¯mïÛã€:.c4fÛØÇpñ ¤ j Nä’yâHhøÊqÿ ܲeKddäÓ3„´+»x¼ññŒ) ٙ±¨â‹¼Î§¿zr«¦S´õîW?ú¬BßÆ2wà^ /GuŸÈlÔ|ý•<..N"‘¼÷Þ{‡’J¥“Ývâĉ'»¡¤¤äàÁƒpÌ@ Ï,===jµÚö7úvmÕ3òÞ{†KPÝ2ñÃç´¦¹à&-rÅŸ)«¦ Ê](P”gWC™ŸÛ‘&ô"n4k`À@žŸ‰oëFyñ›"ç3@P²§2·lp$k/]}欚&Ô`å—ŸÃ7×½•—{a7@ ¨ôOJ#-<%}ÖzäÙ´ @ 'Œ`a@ •@ Tz@ Pé!@¥‡@ •@ Ȭ+=î™ôú«k\foÑ€/ŒÚ.ñšÑ7ç8bó¬·×&ؼ¤=¯¾y Œû´jœ«öÎJ½¸‡d—Ä{ÒÃv0¾lúþýû÷í{i߶('¸†@ _¦xï}JúÆ@¤»KM@S$¡’–Ý’ÆüAéûÚdÓl™ˆ¹„'öçÝ˜Ñ ±y–Û;¤ª1ï¬Ái{f>Ž˜ºÞ”w±Õð”Û;Wõšú$&G[£`5â_Ø®)8qxÀËóÅ›BÏÖÁ÷éA o¬Ò“ŠšKeþ8yþ”T(Ç{í¶k OVô?r:©¬ÉΟ½ÈÎcõJPsx`f_"ëˆÍ³ÛÞ§ Â`³¨9hï\Õ HE]5ùòZ¯šìŽÇÎQ`Î÷w. Œ,. òÆæ‹®Œú!#œ' È7U鋱t]×ó¤éñ gG®5Ÿ:Æ⸲›~V>N±yÁÉR“Á`0™ˆú~y~…L;6»³ÜÃã%¡B” )jènE›Z«½2'»ú‘£Û™žþÚºòqå¢Ï’˜@BÓ!·KJú,©Ì5bóÆ0owCáÇY†ð„u‹ùl³ïÊéܦ!jJ›m[e£,& ß²#Þ¥öRë¢0  LÛr9ïÆ ½4ÇnXáÎ M£®¿áú£ïseºGLJ/`ÐÊ@ôUEåwŽõ!Ê÷‹ÝëÏ#´z’Òj™ºÊÓÅݦюò±ú7îØ?àX{'÷¾CõÚö¾=^ 5­µjI¤'³£}¼„ºo«7¬µVth)Ìyñ ‘òÎmœ$ Ès¤ôR%•2bçã#PiÁI)À½6¤ùLtsSѹ&³þ¢‚ˆq¢žóìŠòC“\kÏm¦Ê­KOóUx¾~xœ}±Ùq]ÿˆÐãîQi1tYÖñN-ÆG§n‹Ì=]5’%5Ž5‡ì;°&}çÐíœÌ#ýF”Åw¢Ìk [6Oe•²¤²öRiè >_ž*S“奤Çx)èvX*P^Цd‘4ëD¡ŠãûÅ$Çzá7,WÃRR}›Ïž,$¸0,)5‰>•-ÕŽôK¼QºñÕ'íz`ûlL@Fs6Ž5‡ìÞd<—yo‰#íÜûÕkÛûöyA/—£ÅFûø“k©¡ÆKW_<ðú›è€ÎÙUSò¯£-ZxH=yΔP&=`ãöíSBØ ‚×D…y:¡a2ñùüzF`,ˆ S]>Ý:L$ :*ò냓&R9®ÇТ³ên@´WcΩNs~€Ô´]Ésßí{ûü}½ÕmDOÙÉ"™ŽPµÆžÎ°×ªÉ0ÈŠ/ßU“#bÕÞ<•ž1?b…®$³Y5"K¤º¥¼L”`–k†ûª}ññÆA³h™” åâô·ûåý$ƒ2Õ‚P?c{ŸrpH/Ë?,›±Ñc«½6¼ïvxJ/ºA'ˆ‡‚¾Ç×TãX%ÿû×5Ép Þ¸~ñù‚V{ò\)=Š³þ¡=[”n«·'òoåœ)Ui0~hzòèEŒ'$»•V³/¡êRY| MÒŠ Ðce]±^©ÖÚRÝ=Èvç ÷õcÿ5)eòiÎÑö[5)ôX4E͈D`|7òÁMk‰4);FóöqŒ+xG§e¬µCŒp0FRʪSÇŠÄ«W‰ÜùÈÃæ’‚›=†?“µ×¦÷ê {¼?…EhòñAŒÎ[ºœº~¶EMˆÁ¦Òš€-~ܶ ”zò)=6/ €hͶ#LÅ\‚Ċ˧›–d®Ñ@Ð̱ÉY…¹2ïtYŒ1Ï[ÈxücÈ¡^-Û‹‚± ФFA¹{pQ¹Š›þ=\ £{¤í¶êIãQŽ÷Ê5AΤ¼öÚma¿UØ27fSç¨ØãB‘+Sn^Œ(e=2ùù<Ù$Ú0ù\Z3ÔÓR×ÓRw  Ü Ô]Ñ]_ö³9¬l{ß!Í€÷1ž[Û«~l'‚1’“uŠ 1f½Ê„@ ¯!Óy^åŠc’;ËîÚãPz!\ÈÃFJòÄ1é)NÈh°Ü_SÏ‹Y2ruòŠ÷gÓôDJßÖF‹ñ¬ì¤‡¥WºC6G‰Ì;cü%±›}ï]iÓ;Úv[õ„p|bc_\ÎÒÓ²ªúG²~©`dU†9Ä%>LO*k¯é×$,_À²t.ƒ¿hYÄbKqüS÷Æx³G/r\Úh¥q¤VÏótÁ-ËÈÏ‹ëøC䶽ïH½3à}”ë僴·©Szr°¹MµÂ Û{õæûjøŒùÇô˜kDbŒx1²kß $ ”êNQfÍè#væÒ8@9î^N¢ ±†W.–tè ÕM ›÷îGh„šäuåùíÛ£wleåäT+@©³óñ ){£0ÂhÐ+ï”–vDãê\m#š$¨¶zþÙ$¿šY±>uo JÓBô7䟫W«#“7.ux[¶v?²±Û¦ÍSXe«,â´X²5~‰/ضޘUܪE]–'Æù‰Ñ¤•ò¬*süÎò1 ÍyeÓËSÃ͹ÑØÄŒ•8E¤i .ç’0C’–ÊÊÍ® ]eYô Iêžy(“‰0 µW—[òÛ´I3Hû'í‹wFH’F1CWå¹R«hYñu¿ÔWâ:ÎHÒ¦¦r9åp{§ðþ“×kÛûöx ÎËÙ}¬¶HÅÍ‹Uë6îÚïD¨ï•å6¨ ÐC ¯;ÈÿgïM£šÊÒÿß}ÎÉH Š ¬€ŒåT ˆ ˆ(¢èÏêá–·{­îª^ë®õûõê^õ¦^ßußuÝî¾ëVݶÊ…VQ™Aˆ „y  !Ãî D…LþÕÏç$ûìgÙß½Ÿ³÷~ ó/…GŽ÷ßcÇŽÅÆÆ®Ÿ!üÐ3™Ä«úM†ÎÝÑð¯Òõ?×¼²Uk‡#;þõ—¾ÿ¡ACAÇÛ0ØÞ)ç çvèþ××p…B¡T*¿ûî»o¿ýV¥R­ôµüüüûBuuõ…  ÏÀ¦eddD«ÕÚþE?omÚ$·}²ÜBØWË>gt]åŒLÊÞTV­\°]¬©/n™ßØž&ñ&—wé`©ÀÎÀ·yãÂà¤#±[Ya85òäníô "HM·Þ{°é¬Z#ôô“‹?B‡Ûhȱ'UcP €Ò¯´NUq]µÙjdsZ¹‚…*Pz@饔PzøÏ#ø½°¹”$eçæžýí¿:èNü⪎½ýÀI¥·Sïóa{¥ýá«}nøz§ubKXZÎWú2B°nÕ¿Aåý$ù²=•g”>+„Ây[ƒâ’OäýoüÍçÒ_Þï€ Uú­ûr¾ùë_¾9æ÷6n+W~ü›¿üõ›œ½[í8ˆOϽ®,¸råZé‹Ùån$ã‡dó_6$,æ"?ô›ÿùóïÏeDKDH¢3ÎýþÏÿý›Còwƒ¤l(„[TJÈÄóQ³3JÆ{ÕSæuOë8ÔLGéª~ƒ“­Ü7>my7*_óxûorŒdYg‹·»ê»«nÞ}> 7ûðØ쉦š6¿YB"êoÓÒá¢XAgkÕÒ<áxÐSŒÅã.?r1úÞúú)~ûv—!4Õr»Ðè’n®ëÕošð¡lϽqèéϓνږšzZ\¶i7+÷O[ÞÊQšgÍÔùýÞO‹û?ˆî`moEˆå ÃNWz„b´ÝÍL|¤{gý8IH#Ãç›êfc‚ZˆwdOUñý•—FD¸íNKÚá…êþ~qõx0„4æèá_k¸¹÷ÎÙáŠà{E+BÄà ŒÒ<¯®i72s HÎNßiVOpÄ<6®ii|ã-qññ…¥/,ñGWHkõZxD%)Ã%8IÑôlwCoP–‚Û÷¤¤¸ù€ò¯˜ ý³ú÷Â÷Úz²¥ÈÆŠ£R>÷ñ„œñº‚»³4ZŒßÆåKy]—¨aÓ*i IÔ±SIn­÷zvDx²Î"ô¯ªJ;é*¶dWâ¡he6šæ'Ú½{Ë,Gº+>)j‹Ápfhº_ßýæmâ¢ÀÄ£‰ABRo hôž3ÿ¤ rØlOßp¤¼Ä–]JåN1‡Åbq8˜öu}Yƒz!ìŸCùÚn}{ZÑõ´j•±^œþ> K¬«Ò#dhìûlÏÞÝáíŸÉÁ1–UHsQáüé–5Ét[ñÕî S'8ö­`žÞ¾ÒöÅÓÍ—kW¶Ç¬¦¶èꀞFˆÊã3OÄÞ-hš¢ô¯«Ë;ÒLµ—o cþ'/$…/1²Ô¬‰êá$e+-B…§§H[o\Η}žå7Sù·[msïÕšXîMõ?z×WmóÉÖ"Ÿýr_öéÙç%…—&L8WäB[V´NU~M…ØÞ‡²|?˜Ö¬b•´ÔT뽚ð¯S|.^¯ÕR†fd'x]*vØýŒ C¤ËTEù3$B„(0!=Ñ›ýØú©kDF¦_×k÷§I„[‘–™Æ\/VYÜ2\ùa%÷ñOÿì3 „Ï÷pv0foßp¤¼ÔLçý›–vÃÅ1§²‘’>“£ùÚn}ûZÁ0ª6ÅËŬ¾qÆ%œ‰‹(M[/v§WhœËóÖIr½è¯Äì\+Ùñž‹[ß0Ap¼wGIÀ%;¥ë­+UÇûñ¬£}wç˜!R79Òýb̈¥Ÿ¥xllÕ´¬m±3Uµ=s4BÑóý em3Ë•¸óÓóKu«Yeý9R{­¨e„¢Z‚k·U+ÎÕÔ•UÝZjA¬úºæ¤nNØFÈÚ=_]Ù5³ K”öU}m?‰[äšå±'ÆPYÖ1m-óT{y=¾'fqc&9­žÙèåîÊ#A]ösù³Þ}Û*/ÆÛ¶Kq"'///çÌéŠÐ­R¡3B?ØÑú«¶5?mä» aw=뿦GˆÑ½lÔþêäþ¡;?jiÄ['ËLÃ5·ó» ðwYB)1¦Ò/}Ji‡§y|üµa•—¬¶ÓB 5¼t³938eZnƒC1Ža1k´Ê<¥_ÛdÉ~«VnÁ·924픹!r§†—:šÍSý“Å¥C$bŸø¬ÜýKÄà±'ùB .Ž©¦ëW]·ËäÁ{÷ȼÅbc¢ï«ôbà )Ù¾Pʲ3‚ß=¸¹}ÒúKxûb}½3 ô¬çšÞ_‘)wóÏñxY}»vЈ×'!Cèíãê¯è+¸ß£gÌCšÃ3ÏgH„QÓªæÇÃAÓâF‹š4¤u§2á|oY®ÜÈ ãPÝê~ËègTW> Ìüõö¼nÞD1æÉÎúQ!„0¹òxR€–Á/)i™Ù}ôè>¹7}2‰¼UÕ;Ï óèƒÂÚ˜ä̼œa0Œœh/»Ù¦¡Â\)‘þb¯ƒý5£ˆpÛ¯.¨DÄ–°”›WL‹B´¶£¸Œ}(#ïAšŒ†©55ýñìeu®µÏ7?šZ™m>™Æ¦Þé&ö±ní~gc·íº²m•­´˜‹¿òxR€:‘l*ªìÑãn‘©Š@9nm#„âz†ËXÆ®ÒÚµùÎé¹®»wðÄÔÜ86M’”yòYÉ=I®2+“{·¸i’4Ö1ÑÊÌœ-†h³™4Ïö5×[ýÛŒY7Í¥MrÅ(ŠÁ ãà“›5Kv¢­Ø7*/£í,oO9šwc0n}V_Öw2þÔqnI‰åhÅGæk»õíi„»†ìfu/³%ݳó·y üò|´i¨þvµÚ£öæ_ Š{ï¿!!!ÇŽ‹ýO¯~è™LâÎÕeÎ b°ӹ;þU:`ÚDV­Žìø×_z<øþ‡ ¬&7¶wÊùÃù…ºµ¼ìQ(J¥ò»ï¾ûöÛoU*ÕJ_ËÏÏÿ¸/TWW_¸pZ6-###Z­Öö/úykìôµåðp c^-øœÑu•722){SYµfpÁv±¦¾¸d~c{šÄ›|\Þ¥ƒýx|Š!ªà}í'‰ÝÊBé‘'wk§WAjºõÞƒMgÕ¡§Ÿ\ü}£!ÇžTA5J¿>Ð:UÅuXüBV°PJ(= ô€ÒJÀ†)=.HÊÎÍ=ûÛ?~uÐøÅU{û“Jo§ÞœÃöJûÃWûÜðõNë8Ä–°´œ¯þôe„`ݪƒÊûIòe{*Ï(}V ¶Ã’„§ÉÍÍ;{îÜé¤1ÀiJ¿u_Î7ýË7ÇüÞFHåÊó—¿~“³w«£ =÷º²àÊ•k¥/f—»û‹’yÌÙ૘‹üÐoþçÏ¿?—-!!‰Î8÷û?ÿ÷oÉ]°MSq„[TJÈÄóQ³3JÆ{ÕSæuOë8ÔLGéª~ƒ“ï¬Ü7>my7*_óxûorŒd¹ù0.Ü™~Ä£ûNþ•Ë—~¾Z1š~Ћ #öcK°'šjÚüf a\ˆ¨¿MK#„‹Bb­=TKó„ãáE1»üퟌ¾·¾¾GŠß¾Ýe@MµÜ.4º¤›ëzõÌf©7¶çÞ8ôôçIç^"KM=-.Û€´›Ž•ûƧ-ïFå‹(ͳfêü~ï§ÅýïÇQÀy¬ÞŠºÞ…à@ätÇþÓQÛØÃCfà°Ò#„£ínfâ#Ý;ëÇIB>ßT7„ÐBt¶#{¨Šï¯¼4"ÂmwZjÔ/T÷÷‹«G^!¤1GGøøZ»½ÚÍWß+Z™"Æa”æyuMû¸‘A˜K@rvúN³z‚#æ±qMKã¿h¹ˆ‹?(,}¡¥m¥µz-<¢’”ᜤhz¶»¡7(KÁí{RRÜüNèvŽWLþY½Ž±Çª¥Eö0Vü£È•ò¹¿ˆ'äŒ×Ü휥Ñb¤4._ÊëºüCýû3›VÙHKH¢ŽJrk½×³#“p¡UUúØI—æ³%»E{°(³Ñ4?ÑþèÝû\9Ò]ñIQÛX †³0Ã@Óýúî7oë&M ’zE¦õœù'•Ãf{ú†#å%¶ìR*wŠ9,‹ÃÁ´¯ëËÔ êP¾¶[ßžV`t=­Ze¬§¿ï=©''ÛÛ–V«XŠi_Àù8Oé24ö}¶gïîðöÏdÆàË*¤¹¨pþ‹tËY8Ñ IDATšdº­øjwЩûV0Oo_é ûâˆéfá˵«Ûã@VS[tu@O#Dåñ™'bï4MQú×Õåi¦ÚË·†1ÿ“’‰Â‹‡YjV„Dõp’²•!„‹ÂÓS¤­7.÷ÌÑçË>ÏÎò›©üÛ­¶¹÷jM,÷¦ú½ë«¶ùdk‘Ï~¹/ûôìó’ÂK&œ+r¡-C6­S•_S!¶÷¡,ߦ5«Xe#-5Õz¯&üë߆‹×kµÂ…¡Ù ^—ʇ^âÂÐ#é2UQ~Å ‰! LHOôf?¶~ê‘‘é×uãÚýi!Ä–D¤e¦1׋U· W~XÉ}üÓ?û !Äó=œŒÙÛ7)/5Óyÿf§¥ÝpqÌ)…l¤¤Ïäh¾¶[ß¾V0ŒªMñr1«o|eKs$öMÝ)HÀÆk{d¹­»Ó+4Îåyë$¹^tŽWbv®•ìxÏÅW“˜ 8Þ»£¤aÀìœÒõÖ•ªƒãýxÖѾ»s̈©›é~1fDˆÒÏR<6¶jZֶ؈™ªÚž9!„èùþ†²¶™åÊ‹ ÜùÆéù¥ÃíjVY¿FŽÔ^+j™0!„h£Vg‡àÚmÕŠs5ueU·–Z«¾®9©›^ó²¶ÆDÏWWvÍ,È¥}U_ÛOâ¹fyì‰1T–uL[DË<Õ^^ï‰YܘIN«g¶……z¹»òdP—ý\î4w´­òb¼m»'ròòòrÎœ>¡Ý*:c›­¿j+PóÓF¾»på_$!K=ìÞv»~ÐN]Ó#Äè^6jurÿе4â­“e¦ášÛù]†…?ø»¬Î„¡”Sé—ú/)íð4σ¿6¬âÕ´–J¨á¥›­È™Á)Ór´ŠÁp CˆY£Uæ)õèüÚ&Kö[µr ¾Í‘¡i§ÌÕ‘;5Ô¸ÔÑlžêט,.B ûÄgåî_"†=É'ZpqL5]¿êº]&Þ»Gæ!ÂÞtU—7Ž8KÁV*/Ë}ïÉTQKÉ¿k4&!Bžî” íjýUZÃpŒ¡èWóQéižªÛe¯æÀs€ó•!ÓP]Ñvm¿¡v¯ûvxJ§¡=<øè ýv õt3jæiÓRZ !å¼´ k‹„3¹ÌcfÇôÖ­Ýïlì¶]W¶­²•sñWO ðC'’ME•=zÜ-2U(Ç­m„B\ÏpËØUZ;°6ß9=×u÷ž˜šǦI’2O>+¹'ÉUferï7M’ÆÁÚ"&Z™™³…Àm6“æÙ¾æz«›1릙 ´³I®E18a|r³fÉN´û†Cåe´åí)GóÎa ÆÂÍ£ÏêËúNÆŸ:Î-)±­øÈ|m·¾=­€p×ݬŽâwãq}û·³Gâ³ò,ÊO¸n5W°ÂüKáQqïý7$$äرc±±±ÿéÕÃ=“IܹºÌ¹ALv:wGÿJL›ÈªµÃ‘ÿúKßÿРýÜÛ;åüáüÂÝZÜó …B©T~÷Ýwß~û­J¥Zékùùù÷…êêê .@ëÀ¦eddD«ÕÚþE?om‚{ïm9<ÜBØWË>gt]åŒLÊÞTV­\°]¬©/n™ßØž&ñ&—wéà-<Ÿbˆ*x_û„ÁIGb·²ÂpjäÉÝÚéDšn½÷`ÓYµFèé'„Fßhȱ'UcP €Ò¯´NUq]V¿,T€ÒJ(= ô€Ò°aJ ’²ssÏþö_t'~qUÇÞ~à¤ÒÛ©7ç°½ÒþðÕ>7|½Ó:±%,-ç«?}!X·êß ò~’|ÙžÊ3JŸ•‚íà.;>KÏù¯ÿ:wîÜùóÙ‰"˜Ÿ°l§ßº/'/Ù=»ü÷;}–Ù¸òã¿Ë@=— O¬vñ6=÷º²à5b{Êò]æc~Hf²¹ôNχ—½a.òäœ3qÜ¡ŽG‹÷Þ‡y›®_½ß«g6GÅnQ)!¥ÍÎ|(mïUëÌëžÖq¨™ŽÒF—NŠi¼rßø´åݨ|ÍãíC©é1’Ÿ—ÄQxûu ö§ŸÝº48O#ÄÙ~àÔ‘(MÁÓ7pŸNPú‰¦š6¿YB"êoÓÒá¢XAgkÕÒ<áx| ŒÅã.?Z1úÞúú)~ûv—!4Õr»Ðè’n®Û42Ûsozúó¤s/‘¥¦ž—m@ÚMÇÊ}ãÓ–w£òE”æY3u~¿÷Óâþâ(šÖõÓxçK<^ÌzúÆ„Vz„b´ÝÍL|¤{gý8IH#Ãç›êfc‚ZˆÎvdOUñý•—FD¸íNKÚá…êþ~qõÈ+„4æèá_k`·÷B»­Î÷ŠV&„ˆ1†A¥y^]Ó>nd朾Ӭžàˆyl\ÓÒøÆ/Z.ââã K_hi[i-p=¢’”ᜤhz¶»¡7(KÁí{RRÜüNèvŽWLþY½Ž±Çª¥Eö0Vü£È•ò¹¿ˆ'äŒ×Ü휥Ñb¤4._ÊëºüCýû3›VÙHKH¢ŽJrk½×³#“p¡UUúØI—æ³%»E{°(³Ñ4?ÑþèÝû\9Ò]ñIQÛX †³0Ã@Óýúî7oë&M ’zE¦õœù'•Ãf{ú†#å%¶ìR*wŠ9,‹ÃÁ´¯ëËÔ öÊ×vëÛÓ Œ®§U«Œõâô÷Ù’pB°'D×Ýi†¡ §)=BÆƾÏöìàÝÞþ™l ±ÑcY…4ΑnY“L·_í:u‚cß æéí+]a_1Ý,|¹vÕa{ÈJ`j‹®èi„¡<>óDìÝ‚¦)Jÿºº¼3 ÍT{ùÖ0æòB2Qxñâ#KÍŠ¨NR¶Ò"„pQxzŠ´õÆåž9á|ÙçÙY~3•»Õ6÷^­‰åÞTÿ#cŸUKŠ|öË}Ù§gŸ—^š0á\‘ mYAÒ:Uù5Õòo:V³ÊFZjªõ^Mø×)¾ ¯×j)„ C3²¼.•;,¸0ôHºLU”_1C"DˆÒ½Ù­ŸºFddúuݸvšD±%i™iÌõb•Å-ÕVrÿôÏ>Bñ|gcöö GÊKÍtÞ¿Ùii7\sJ!)é39š¯íÖ·¯ £jS¼\Ìê[>-!‰>žî>]ýãOªÍãÜàìØÛCiÚÚx±;½Bã\ž·N’ë5Æp¼³s­dÇ{.n}ÃÁñÞ% –`ç”®·®TïdzŽöÝcF„HÝäH÷‹1#B”~–â±±UÓ²¶ÅFÌTÕöÌÑ!DÏ÷7”µÍ,W^\àÎ7NÏ/]Ô­f•õkäHíµ¢– Bˆ6jíyßk·U+ÎÕÔ•UÝZjA¬úºæ¤nNØFÈÚ=_]Ù5³ K”öU}m?‰[äšå±'ÆPYÖ1m-óT{y=¾'fqc&9­žÙèåîÊ#A]ösù³Ö©¶Ê‹ñ¶íRœÈÉËËË9sú„"t«TèŒÐv´þª­@ÍOùî•~‘ÔTËþ_ÿçO]•‡Ã\aOÎ\Ó#Äè^6jurÿе4â­“e¦ášÛù]†…?ø»¬Î„¡”Sé—¾N¥´ÃÓ<>þÚ°ÊKVÛi ¡„žZ"8äÌà”‰»\•P †cBÌ­2O©Gç×6Y²ßª•[ðmŽ M;e®FˆÜ©¡Æ¥ŽfóT¿Ædqé‰Ø'>+wÿ1$xìI>Ђ‹cªéúU×í2yðÞ=2ö¦«º¼qÄ褾³RyYî{O¦ŠZJþ]£11¢ðìt§dhWë¯Ò †c e»Óó#-÷î‹NïÙ®ª>ÎSz„LCuEwÚµýF„>8-‡­·É”NC{x ðÑúí@ëéfÔÌÓ¦¥´"BÊy1h0Ö gr™ÇÌŽéyîÍÒΰj›íµêc×£|Ÿ¸}¡®ÔhëÃçÒ~«&‰ÝîœÎE±gKdRΨe22¥QÞ*U¯ ÝG$`t³#¯ž¼zöá‚ÐÌ3ñƒ?UŒŸ²çn¡rMUA—ÆêÊ7I†³Ñ}òmwž~L»êVÚ¨G|.cöb§Öö Ï}0j‘:-ßGºà…äl‹LIö_ÊŸÒ„^V&Æ˽ö˜ÃÌ©ê†ÃŽ¹à–Õe@âQ¿—u½GÓ’OÛDŠ„!B¸‹ß¤ Ã,§ô½½ŒlÇRO«#VÙÆn«>¾objâÁHᬎ\“UÍ-|eòNñÂ\‘p V¤îZ4ìK‰Ü¶¨H,ÑŽÝ1þÖãeæ%øð?ˆŒ‰r¼o¬Ò‹ 3¤dûB="\(OÈÎvÁ6ºOZ‰o_¬¯wæ¥çzïù<܃·ð`ŒëshçèÓ!Øy€SÖôþŠ”H¹›ŽÇËêÛµƒF„¸> Š@oWE_Áý=czОyî2_Û­oO+ Ü5d7«£x™ÝxÆ1Õ¸ïÁg%l†¢Ò÷6ܬ¡` `…ù—£âÞûoHHȱcÇbccÿÓ«‡z&“¸su™sƒ˜0ìt•˜6‘Uk‡#;þõ—¾ÿ¡ACÁÏaÃ`{§œ?0œ_Ø¡[ËË…B¡T*¿ûî»o¿ýV¥R­ôµüüüûBuuõ…  u`Ó222¢Õjmÿ¢Ÿ·6Á^[·Ð0æÕò‡Ï]Wy##“²7•Uklkê‹›Aæ7¶§I¼ÉÇå]:¸÷€O1Ä@¼¯}Âà¤#±[Ya85òäníô "HM·Þ{°é¬Z#ôô“‹?B£o4äØ“ª1¨@é×Z§ª¸®«€_È ª@饔Pz@éØ0¥ÇIÙ¹¹gûǯº¿¸ªco?pRéíÔ›sØ^iøjŸ¾Þi‡Ø––óÕŸ¾Œ¬[õoPy?I¾lOå¥Ï*ÁvpáÎþøü×Np’ÒoÝ—óÍ_ÿòÍ1¿·R¹òãßüå¯ßäìÝjÇA|zîueÁ•+×J_Ì.w÷?$ó˜ÿ²ÁW1ù¡ßüÏŸ.#ZB Bqî÷þïß’»lž^„[TJÈÄóQ§†¥ ã½ê)óº§uj¦£ôFU¿ÁIwV¼•¯y¼}È79Fbk>Ìó=>~÷þ€‰AkÁ–`O4Õ´ù͸Q›–F…Ä :[{¨–æ ÇËb,wùÛ?}o}}¿}»Ë€šj¹]htI7×õê7Í Çö܇žþ<éÜKd©©§ÅevÓ±rßø´åݨ|¥yÖLßïý´¸ù8 ¯½q懷&ÞðÆ '*=B1Úîf&>Ò½³~œ$¤‘áóMu³1A-Dg;²'€ªøþÊK#"Üv§¥FíðBu¿¸zäBsôp„¯5°Û{¡ÝìpE𽢕 !bŒaFižW×´„¹$g§ï4«'8b×´4¾ñ‹–‹¸øøƒÂÒZÚVZ«×Â#*I.ÁIŠ¦g»zƒ²ܾ'%ÅÍï„nçxÅéŸÕë{¬ZZdcÅ?ŠŒQ)Ÿû‹xBÎx]ÁÝÎY-FJãò¥¼®Ë?Ô¿?ƒ°i•´„$êØ©$·Ö{=;"<ÙgúWU¥ti>[²+ñP´‹2Móí޽ϕ#ÝŸµÅ`8 3 4ݯï~ó¶qQ`âÑÄ !©7P´aZÏ™RP9l¶§o8R^bË.¥r§˜Ãb±8Lûº¾¬A½`Ï¡|m·¾=­ÀèzZµÊX/Nß2RÏÚs€×\ûå¾ìÓ³ÏK /M˜p®È…¶¬ iªüš ±½eù~0­YÅ*i©©Ö{5á_§ø6\¼^«¥. ÍÈNðºT>ì°û†I—©Šò+fH„Q`Bz¢7û±õS׈ŒL¿®×îO“!¶$"-3¹^¬²¸e¸òÃJîãŸþÙg@!žïáì`ÌÞ¾áHy©™Îû7;-톋cN)d#%}&GóµÝúöµ‚aTmŠ—‹Y}D®ÅÅ៻wT<™G^Ñðãµ=²ÜÖÆ‹Ýéçò¼u’\/:Ç+1;×Jv¼çâÖ7LïÝQÒ0` vNézëJÕÁñ~<ëhßÝ9fDˆÔMŽt¿3"Dég)[5-k[lÄLUmÏBô|CYÛÌråÅî|ãôüÒEÝjVY¿FŽÔ^+j™0!„h£Vg‡àÚmÕŠs5ueU·–Z«¾®9©›¶²¶ÆDÏWWvÍ,È¥}U_ÛOâ¹fyì‰1T–uL[DË<Õ^^ï‰YܘIN«g¶……z¹»òdP—ý\>ä¬w߶ʋñ¶íRœÈÉËËË9sú„"t«TèŒÐv´þª­@ÍOùîÂ~‘˜0øs¿¾úntÀ'[Ó#Äè^6jurÿеôº­*LÃ5·ó» ðwYB)1¦Ò/÷(íð4σ¿6¬2ÚNK%ÔðÒÍVäÌà”i¹ Z Å`8†!ĬÑ*ó”zt~m“%û­Z¹ßæÈдSæj„Èj\êh6OõkL—!ˆ}â³r÷/,‚Çžä-¸8¦š®_uÝ.“ïÝ#óaoºªËGŒNê;+•—å¾÷dª¨¥äß5ƒ! ÏNwJ†vµþ*­€a8ÆPv`¶4ÀC,>òÅÎ…j—y3©ÑãwZ¦ Î0NTz„LCuEwÚµýF„>ؼîÛá)†öðà£3ôÛÖÓͨ™§LKi5D„”óbÐ*`¬->Îä2™ÓóÜ8z{¨À«V±Ù^«>v=Ê÷‰ÛêJ¶>|®!í·j’ØíÎéX{¶D&åŒZ&#Sêõè­Rõ ÚqDF7;òêÙÈ«g.Í<?øSÅù){á*×Tti¬®|“‘d8Ý'ß&pçéÇ´¸I]üÃÿgýƒ|:¿2ÀZ°s/­ížû`Ô"uZ¾tÁ ÉÙ™’쿆?¥7½¬.LŒ#–z ì1‡™SÕ ‡= sÁ-«Ë€Ä£~/ëz Ž¦%'ž¶‰ B!„p¿IA<†YNé{{ÙŽ¥žVG¬²ÝV}$|ßÄÔă‘ÂY¹&«š[øÊäâ…¹"á¬Hݵx˜žšj}hØ—¹k²D;vÇø[kŒ”™—àÃ[üP 0&Êñ¾±J/6Ì’í õˆp¡®þŠ¾‚û=zÆ<ô 9<óÜy†D5­j~<t0-n´¨ICZw*³Î÷ðv‘åÊ 2ÕÝ©î·Œ~Fuå£ÀÌ__`ÏëæMcžì¬¥Bs‘+'øaÙü’’–™-ÑGî“{Ó'“È[U½ó 2>(¬IÎÌKÀÃȉö²›m !Ì%@‘é/ö:Ø_P3Š·Ý)ñꂪADl KY°yÅ´!Dk;ŠË؇2ò¤Éh˜zQSÓÏ^VçZ;ðŒPqó£·«+›O&¤±é‡wº‰}¬[»ßÙØm»®l[e+-æâ¯<žà‡N$›Š*{ô¸[dª"PŽ[Û!„¸žá2–±«´v`m¾sz®ëî<157ŽM“$ež|VrO’«ÌÊäÞ-nš$ƒµEL´23g !Úl&ͳ}ÍõVÿ6cÖM3Aig“\1ŠbpÂ8øäfÍ’h+ö ‡ÊËh;ËÛSŽæÃŒ…›GŸÕ—õŒ?uœ[Rb9Zñ‘ùÚn}{Zá®!»YÅã+϶IÔ‘ÔÝ>¾^tZÌÄí§˜`'Xaþ¥ð¨¸÷þrìرØØØÿôêᇞÉ$î\]æÜ & ;»£á_¥¦MdÕÚáÈŽý¥ÇƒïhéØ@ØÞ)ç çv¬ißB¡P*•ß}÷Ý·ß~«R©VúZ~~þÇ}¡ººúÂ… Ð:°iÑjµ¶ÑÏ[›à[·Ð0æÕò‡Ï]Wy##“²7•Uklkê‹›Aæ7¶§I¼ÉÇå]°½€O2Ä@¼¯}Âà¤#±[Ya85òäníô "HM·Þ{°é¬Z#ôô“‹?B£o4äØ“ª1¨@é×Z§ª¸®«€_È ª@饔Pz@éØ0¥ÇIÙ¹¹gûǯº¿¸ªco?pRéíÔ›sØ^iøjŸ¾Þi‡Ø––óÕŸ¾Œ¬[õoPy?I¾lOå¥ÏJÁv¸þÙ_ÿî|ž%‚sÎñ8w8 À°5dlÝ—“—ìž]þû>Ël\ùñßåF žŠË'V ˆBϽ®,xØÞ‡²|—ù˜’™l.½Óóáeo˜‹<9çLw¨ãÑâ½÷a^ƦëWï÷ê™ÍQq„[TJÈDéc³3JÆ{Õ:óº§uj¦£ô†Ñå„“b¯Ü7>my7*_óxûPjzŒäçGËG©3=ü÷µ6NVú‰¦š6¿YB"êoÓÒá¢XAgkÕÒ<áxxQŒÅã.û'£ï­¯ï‘â·owBS-· .éæºM#ó±=÷Æ¡§?O:÷YjêiqÙ¤Ýt¬Ü7>my7*_Diž5Sç÷{?-î7!€uSz„b´ÝÍL|¤{gý8IH#Ãç›êfc‚ZˆÎvdOUñý•—FD¸íNKÚá…êþ~qõÈ+„4æèá_k`·÷B»­Î÷ŠV&„ˆ1†A¥y^]Ó>nd朾Ӭžàˆyl\ÓÒøÆ/Z.ââã K_hi[i-p=¢’”ᜤhz¶»¡7(KÁí{RRÜüNèvŽWLþYý{‹,[O¶ÙÃXñ"cTÊçþ"ž3^Wp·s–F‹‘Ò¸|)¯ëòõïÏ lZe#-!‰:v*É­õ^ÏŽO6ÂY„þUUéc']šÏ–ìJ<íÁ¢ÌFÓüDû£wïsåHwÅ'Emc1ΠM÷ë»ß¼­C\˜x41HHê m˜ÖsæŸT›í鎔—زK©Ü)æ°X,Ó¾®/kP/Øs(_Û­oO+0ºžV­2Ö‹Óß·ŒÔ³¶î9y6 >¹ IDAT’AÇ©‰–{í3Œ]8Ké24ö}¶gïîðöÏdÆàË*¤¹¨pþ‹tËšdº­øjwЩûV0Oo_é ûâˆéfá˵«Ûã@VS[tu@O#Dåñ™'bï4MQú×Õåi¦ÚË·†1ÿ“’‰Â‹‡YjV„Dõp’²•!„‹ÂÓS¤­7.÷ÌÑçË>ÏÎò›©üÛ­¶¹÷jM,÷¦úû¬ZRä³_îË>=û¼¤ðÒ„ çŠ\hË ’֩ʯ©–Ó±šU6ÒRS­÷j¿Nñm¸x½VK!\š‘àu©|Øa¥À…¡GÒeª¢üŠ!B˜žèÍ~lýÔ5"#Ó¯ëƵûÓ$Bˆ-‰HËLc®«,n®ü°’ûø§öBˆç{8;³·o8R^j¦óþÍNK»áâ˜S ÙHIŸÉÑ|m·¾}­`U›âåbVß‘kÍ£ïÜ™£"¶„Ëø|2¿z´{Çk{d¹­»Ó+4Îåyë$¹^tŽWbv®•ìxÏÅ­o˜ 8Þ»£¤aÀìœÒõÖ•ªƒãýxÖѾ»s̈©›é~1fDˆÒÏR<6¶jZֶ؈™ªÚž9!„èùþ†²¶™åÊ‹ ÜùÆéù¥‹ºÕ¬²~©½VÔ2aBÑF­=ï{í¶jŹšº²ª[K-ˆU_לÔÍ ÛY[c¢ç«+»fd‰Ò¾ª¯í'q‹\³<öÄ*Ë:¦-¢ežj/¯Ç÷Ä,nÌ$§Õ3Û½Ü]y2¨Ë~.r–tÙ*/ÆÛ¶Kq"'///çÌéŠÐ­R¡3v·ÙÑú«¶5?mä» —ùEÒú‘¾Ñ9Ê2UyQÿz{äv6Œ]8oM£{Ù¨ýÕÉýCw~ÔÒˆ·N–™†knçwþàï²:"„RbL¥_ú:•ÒOó<øøkÃ*/Ym§%„jxj‰à3ƒS&îrUB1Ža1k´Ê<¥_ÛdÉ~«VnÁ·924픹!r§†—:šÍSý“Å¥C$bŸø¬ÜýKÄà±'ùB .Ž©¦ëW]·ËäÁ{÷ÈÎä2™ÓóÜ8š¥aÕ*6ÛkÕÇ®Gù>qûB]©ÑÖ‡Ï5¤ýVM»Ý9‹bϖȤœQËddJ=¢½Uª^A»1ŽHÀèfG^=yõì!¡™gâª!?eÏ!ÜB嚪‚.ëKnÊd$ÎF÷É· Üyú1íª[U¸[w°&šI»°;—´~´oxîƒQ‹Ôiù>ÒG"g[dJ²ÿVü”Þ ô²º01ŽXè-°ÇfNU7vô€Ì·¬.ú½¬ë58š–œxÚ&R$ „Â]ü$ñf9¥ïíed;–zZ±Ê6v[õ‘ð}SF gu䚬jná+“wŠ抄k°"u×âazjªõ¡a_Jä6®u"ÈíØão­1~Pf^‚oñCXÀ˜(ÇûÆ*½Ø0CJ¶/Ô#Â…ò„ìŒ`l£û¤õ—(ðöÅúzg>Pz— cyGw»s,ß’ŒwyÖêäCüÇ®éý)‘r7ÿ—Õ·kq}2Þ>®þŠ¾‚û=zÆ<ô 9<óÜy†D5­j~<t0-n´¨ICZw*³Î÷ðv‘åÊ 2ÕÝ©î·Œ~Fuå£ÀÌ__`ÏëæMcžì¬¥Bs‘+'øaÙüÅóôroúdy«ªwžAæÑ…µ1É™y 8Ã`9Ñ^v³MC!„¹(R"ýÅ^û jFá¶;%^]P5ˆˆ-a) 6¯˜!„hmGqûPFÞ‚4 S/jjúãÙËê\kž*n^rþÙæ“ ilúánbëÖîw6vÛ®+ÛVÙJ‹¹ø+'ø¡ɦ¢Ê=”ãÖ6B!®g¸Œeì*­X›ïœžëº{OLÍcÓ$I™'Ÿ•Ü“ä*³2¹w‹›&Iã`m­ÌÌÙB`ˆ6›Iól_s½Õ¿Í˜uÓLPÚÙ$WŒ¢œ0>¹Y³d'ÚŠ}áò2ÚÎòö”£yç0cáæÑgõe}'ãOç–”XŽV|d¾¶[ßžV@¸kÈnVGñø‡³-ýË{¥ø~Å¿!ŠÆè™ÎŠâ- Cvƒæ_ Š{ï¿!!!ÇŽ‹ýO¯~è™LâÎÕeÎ b°ӹ;þU:`ÚDV­Žìø×_z<øþ‡ ,7¶wÊùÃù…ºµh¸B¡P*•ß}÷Ý·ß~«R©VúZ~~þÇ}¡ººúÂ… Ð:°iÑjµ¶ÑÏ[›`c-‡‡[hójùÃ献«¼‘‘Iٛʪ5ƒ ¶‹5õÅÍ óÛÓ$Þäãò.,ÕøC TÁûÚ' N:»•…†S#OîÖN¯ ‚Ôt뽛Ϊ5BO?¹ø#4úFCŽ=©ƒj”~} uªŠë*° ø…¬`¡ ”Pz@饔€ Sz\”›{ö·üê ;ñ‹«:öö'•ÞN½9‡í•ö‡¯ö¹áëÖqˆ-ai9_ýéËÁºUÿ•÷“äËöTžQúضCˆüöÏÍËÉ={6;Þ›‹ìÄÖyú­ûrò’ýѳË¿Óg¹‘+?þ»ÜÔSq¹àñÄjQè¹×•¯ÛûP–ï2óC2“Í¥wz>¼ì s‘'眉ãuÌËØtýêý^=³9*Žp‹J ™(}lvæCiÃx¯Zg^÷´ŽCÍt”Þ0ºœpRLã•ûƧ-ïFåkoJM‘ü¼$ŽÂ;p<d§¹w–üûʘ‘a §)ýDSM›ß,!Œ õ·ii„pQH¬ ³µ‡jižp ×åà\¥G!FÛÝÌÄGºwÖ“„42|¾©n6&¡…èlGöPß_yiD„Ûî´Ô¨^¨îïW¼BHcŽŽðñµv{/´Ûêà|¯heBˆc„QšçÕ5íãFa.ÉÙé;Íê Ž˜ÇÆ5-oü¢å".>þ °ô…%Ø i-p=¢’”ᜤhz¶»¡7(KÁí{RRÜüNèvŽWLþY½Ž±Çª¥Eö0Vü£È•ò¹¿ˆ'äŒ×Ü휥Ñb¤4._ÊëºüCýû3›VÙHKH¢ŽJrk½×³#“p¡UUúØI—æ³%»E{°(³Ñ4?ÑþèÝû\9Ò]ñIQÛX †³0Ã@Óýúî7oë&M ’zE¦õœù'•Ãf{ú†#å%¶ìR*wŠ9,‹ÃÁ´¯ëËÔ öÊ×vëÛÓ Œ®§U«Œõâô÷} õ¯u}-È<ŸFé24ö}¶gïîðöÏdÆàË*¤¹¨pþ‹tËšdº­øjwЩûV0Oo_é ûâˆéfá˵«Ûã@VS[tu@O#Dåñ™'bï4MQú×Õåi¦ÚË·†1ÿ“’‰Â‹‡YjV„Dõp’²•!„‹ÂÓS¤­7.÷ÌÑçË>ÏÎò›©üÛ­¶¹÷jM,÷¦úû¬ZRä³_îË>=û¼¤ðÒ„ çŠ\hËèMëTå×TË¿éXÍ*i©©Ö{5á_§ø6\¼^«¥. ÍÈNðºT>ì°û†I—©Šò+fH„Q`Bz¢7û±õS׈ŒL¿®×îO“!¶$"-3¹^¬²¸e¸òÃJîãŸþÙg@!žïáì`ÌÞ¾áHy©™Îû7;-톋cN)d#%}&GóµÝúöµ‚aTmŠ—‹Y}ïG®ÅÛø£=FŸý™Ÿ iƒ¦«®ªu|øØ?^Û#Ëmm¼Ø^¡q.Ï['Éõa8^‰Ù¹V²ã=·¾a‚àx†K°sJ×[WªŽ÷ãYGûîÎ1#B¤nr¤ûŘ!J?KñØتiYÛb#fªj{æh„¢çûÊÚf–+/.pç§ç—.êV³Êú5r¤öZQË„ !Dµö¼ïµÛªçjêʪn-µ V}]sR7'l#dm‰ž¯®ìšY%Jûª¾¶ŸÄ-rÍòØc¨,똶ˆ–yª½¼ß³¸1“œVÏl ôrwåÈ .û¹|ÈYï¾m•ãmÛ¥8‘“———sæô EèV©Ð¡ìhýU[šŸ6òÝ…þ"16—ÅÙmz\ðó¥üâ.qò±pW84€×ô1º—Ú_Ü?tçG-xëd™i¸æv~—aáþ.«ó!B(%ÆTú¥®LJ;<Íóà㯠«88m§%„jxj‰à3ƒS¦åv93ƒá†³F«ÌSêÑùµM–ì·jå|›#CÓN™«"wj¨q©£Ù<Õ¯1Y\:„@"ö‰ÏÊÝ¿D¯{’O ´àâ˜jº~Õu»L¼wÌC„½éª.o1:©ï¬T^–ûÞ“©¢–’×hL B„(<;Ý)ÚÕú«´†áC-Óƒž¿u®âÇæQ#BHß÷ðQÔ©`׎¦7àÍÀiJi¨®èN»¶ßˆÐ§å°õ6™ÒihO>:C¿h=ÝŒšyÚÁ´”VCDH9/­ÆÚâ#áL.ó˜Ù1=Ï]€£·¯N±j›íµêc×£|Ÿ¸}¡®ÔhëÃçÒ~«&‰ÝîœÎE±gKdRΨe22¥QÞ*U¯ ÝG$`t³#¯ž¼zöá‚ÐÌ3ñƒ?UŒŸ²çn¡rMUA—ÆêÊ7I†³Ñ}òmwž~LûáVZ?ôjÒÅH½f’ kzìÆ΃Öö Ï}0j‘:-ßGºà…äl‹LIö_ÊŸÒ„^V&Æ˽ö˜ÃÌ©ê†ÃŽ¹à–Õe@âQ¿—u½GÓ’OÛDŠ„!B¸‹ß¤ Ã,§ô½½ŒlÇRO«#VÙÆn«>¾objâÁHᬎ\“UÍ-|eòNñÂ\‘p V¤îZ4ìK‰ÜƵNY¢»cü­5ÆÊÌKðá-~( åxßX¥fHÉö…zD¸Pžì‚mtŸ´þÞ¾X_ï µ\]·©Ü?ÞÊF!–4 Ìüõö¼nÞD1æÉÎúQ!„0¹òxR€–Á_™Æ¦Þé&ö±ní~gc·íº²m•­´˜‹¿òxR€:‘l*ªìÑãn‘©Š@9nm#„âz†ËXÆ®ÒÚµùÎé¹®»wðÄÔÜ86M’”yòYÉ=I®2+“{·¸i’4Ö1ÑÊÌœ-†h³™4Ïö5×[ýÛŒY7Í¥MrÅ(ŠÁ ãà“›5Kv¢­Ø7*/£í,oO9šwc0n}V_Öw2þÔqnI‰åhÅGæk»õíi„»†ìfu/;Û¢gÚî>Ú“zò ›EŽ4–Tk(»°¬0ÿRxTÜ{ÿ 9vìXllìzõðCÏdw®.sn†ÎÝÑð¯ÒÓ&²jípdÇ¿þÒãÁ÷?4€xl lï”ó†ó ;tkY«+ ¥RùÝwß}ûí·*•j¥¯åççܪ««/\¸­›–‘‘­Vkûý¼µ ^÷Ùrx¸…†1¯–?|ÎèºÊ™”½©¬Z3¸`»XS_Ü 2¿±=MâM>.ïÒK€O1Ä@¼¯}Âà¤#±[Ya85òäníô "HM·Þ{°é¬Z#ôô“‹?B£o4äØ“ª1¨@é×Z§ª¸®«€_È ª@饔Pz@é°É§æ‡d&›Kïô|8ba.òäœ3qÜ¡ŽG‹÷Þ‡y›®_½ß«g6GÅnQ)!¥ÍÎ|(mïUëÌëžÖq¨™ŽÒF—NŠi¼rßø´åݨ|ÍãíC©é1’Ÿ—ÄQ°B¾yQwëyó¨%K–gÒéP-\i€s”~¢©¦Ío–Æ…ˆúÛ´4B¸($VÐÙÚCµ4O8^cñ¸Ë/‹}o}}¿}»Ë€šj¹]htI7×m™Gˆí¹7=ýÙÉ~ jêiqÙ¤Ýt¬Ü7>my7*_Diž5Sç÷{?-îYOŽ·4/ùÁJ=4/ëHà ¥G!FÛÝÌÄGºwÖ“„42|¾©n6&¡…èlGöPß_yiD„Ûî´Ô¨^¨îïW¼BHcŽŽðñµv{/´Ûêà|¯heBˆc„QšçÕ5íãFa.ÉÙé;Íê Ž˜ÇÆ5-oü¢å".>þ °ô…–¶•Ö×#*I.ÁIŠ¦g»zƒ²ܾ'%ÅÍï„nçxÅéŸÕë{¬ZZdcÅ?ŠŒQ)Ÿû‹xBÎx]ÁÝÎY-FJãò¥¼®Ë¾é°i•´„$êØ©$·Ö{=;"<ÙgúWU¥ti>[²+ñP´‹2Móí޽ϕ#ÝŸµÅ`8 3 4ݯï~ó¶qQ`âÑÄ !©7P´aZÏ™RP9l¶§o8R^bË.¥r§˜Ãb±8Lûº¾¬A½`Ï¡|m·¾=­ÀèzZµÊX/NŸ­IÄ–Ÿ™îÇf¹pžÒ#dhìûlÏÞÝáíŸÉÁ1–UHsQáüé–5Ét[ñÕî S'8ö­`žÞ¾ÒöÅÓÍÂxßÈö8•ÀÔ]ÐÓBy|æ‰Ø»MS”þuuyg@š©öò­aÌÿä…d¢ðâÅ!F–š!Q=œ¤l¥Eá¢ðôiëË=s4Âù²Ï³³üf*ÿv«í½WÏ,±Ü›êd`ì³jI‘Ï~¹/ûôìó’ÂK&œ+r¡-+HZ§*¿¦ZþMÇjVÙHKMµÞ« ÿ:Å·áâõZ-…pahFv‚×¥òa‡µ†I—©Šò+fH„Q`Bz¢7û±õS׈ŒL¿®×îO“!¶$"-3¹^¬²¸e¸òÃJîãŸþÙg@!žïáì`ÌÞ¾áHy©™Îû7;-톋cN)d#%}&GóµÝúöµ‚aTmŠ—‹Y}ã+/Øq‘ÜW×õÔk¯í‘å¶6^ìN¯Ð8—ç­“äz9Ð9^‰Ù¹V²ã=ƒÆa‚àx†K°sJ×[WªŽ÷ãYGûîÎ1#B¤nr¤ûŘ!J?KñØتiYÛb#fªj{æh„¢çûÊÚf–+/.pç§ç—.êV³Êú5r¤öZQË„ !Dµö¼ïµÛªçjêʪî…7»´®¯kNêæ„|¬­1ÑóÕ•]3 ²Di_Õ×ö“¸E®Y{b •eÓÑ2Oµ—×ã{bÜ KrrZ=³-,<ÐËÝ•G ƒºìçr§íl´U^Œ·m—âDN^^^ΙÓ'¡[¥BglJµ£õWmj~ÚÈwâ6gW~þÆîØ‹€“×ô1º—Ú_Ü?tçG-xëd™i¸æv~—aáþ.«ó!B(%ÆTú¥¯S)íð4σ¿6¬ò’ÕvZB(¡†§–938eâ.W%ƒá†³F«ÌSêÑùµM–ì·jå|›#CÓN™«"wj¨qéâÒ<Õ¯1Y\:„@"ö‰ÏÊÝ¿D {’O ´àâ˜jº~Õu»L¼wÌC„½éª.otš„­T^–ûÞ“©¢–’×hL B„(<;Ý)ÚÕú«´†áCÙêÀ˜‹,ê¾;ÃNWz„LCuEwÚµýF„ˆ†Ÿõ6™ÒihO>:C¿h=ÝŒšyÚÁ´”VCDH9/Ï:±¶øH8“Ë×ö[5Iìvçt,Š=["“rF-“‘)õˆzôV©zíÆ8"£›yõläÕ³‡„fž‰ü©bä“î4#ÜB嚪‚.Õ•o2’ g£ûäÛî<ý˜­]õ<ï`ìe¹F-àÿoï>⸺8€Ÿ­°T)ŠbgA‘n§Š€[DÅž¨‰‚¨Iˆ(ö.±Dc/1‰bï`‹‚QPŒ±€Jo˲íý°€”]Šð"Êÿ÷çavïÌÝ{gï™;;3 vjøäiAÊóäüJ£–8/—g¬'? ÉmaåîÖ¾3~IA¡†aÉ)L·™IG#õšTG–Ÿ™Üe@o5fñ첃ó€¶O"ŸÖµ¬øíÝûš.N4XDDLµ¶½ûvR•ÉEúgÏd&­Ëži­K­ªVãZ½'^gOg+œ£ŒUK_To¦.+’Ô}ߨf/.Ì붔·#15Ú9 hªÆøÐûdÉ7Qݨ ãù³lIžÏyú"_†Q êoNßÞÅݪNû‘O®œˆH©; téhd¬ÕÞåù‘‹ÿÈD¯®G[øŒ'C’™ýwr'‡þö)Ç£ÒÅ%W*sˆÉ30R3ÕN(#á«Èð+/‹G?á‹K·:úLœÂä Š$2QZܵ)C­ë ¾Ú2†ä•ÞOßÎH:¤¯ø¯ËÏ2¥\?aëæ3Ú‰)“1â·±çþ¼Ÿ.!b¨upq·jßÌÐáå‘«)ÄÒ±tw|qär±´»¸Ë무,‘4÷áÉsœ~G÷f‰‹„…ñW¯¾tä(Œs1™ÍšE—¹ÿ¹Ê5³ôì>ó0×if\riw¹ »«n«ªkUUY†Z{×A};´¥ÁnEÇ/ýWÀÔ±òtéØŽYÒGDD*­,LØÂGg"kwî\šÿèt8ÓÙs”=G*KDiÿœ:«;Êu¨Êé“QibaRÄq™«ÏHmƒ¤"‘X”ó<úZÉùm™(/SÖ©¿__-†D"c²„I·ÿ¼ZæJ4¥ûF>¯,7î|¬û€Ñc2›)JùçÚ¹çC‡ R9uªøÖŠ÷ÜnÕ½_“^ ¦ß’ýðdWã±uZIbs¥´ –GÃZXÛWXÊçó½½½íììšzóðÌFø°ÂSpß C£Ëç£ZßÜ{&±¨Õªö¸&ƒfŽ7¸¾e×Ít<åÃá¹ëvôa^m"¹‹‹‹««ëâÅ‹ƒ‚‚”½-,,ìýÞpåÊ•)S¦ w­×¯_çææVý~…çÞWuÂCǬ‹ì_Å7ŸËò¿#3Ñã4ªZÕS½e³ôk'£æ?잦k$þûü£ Ó¾^vÍÙD ¦äõíÓ™J‚ $3æìõFW«Z’fÞÞ·þ¡‰So_NE3"}Ãæ%\ø=µ€Od‹&@¤€ÓÑ°ƒ+!"ooo"ŠŒŒÄ_üÅßÆù÷òåËD´xñb™LöX¹÷~ömÛdЈ%''Wû>v°ª»ìÂÃÃåc 46ŽŽŽ—/_Æ]vMY=Üe‡0Ðháë 5ĬzÒ€h´sz4Ô5Ò×uÒ ÍºýKð‚ù_šrð¥è“k:ῇ–ìŒ/¬×U> õ›òûkIC—­;q꥟Lù|Ö…¬†z,û‡ú¼ÿ—í vþ°3Ni’CqÚ­=ÁþsçúÏ]öÛ?ò4:˜Ó@ýÌéG¸{ø®‰y÷ÔÕü»+Gx¸XpäE "7³Y÷ÉÁËW„ÌtnÁRðrîõåk£>ÐUš}÷çé¼&¬>™$"QÒ©Õ¼¼§ÿ|7»ñÓ8±nÃÑ;œ©{WôR'üNL IDATñës¡¡§<–Ù½¾¯vuQRøúMçþó/뇹GÙDDl#ÏY3úµ¬é#û¤¹OîØ™"e2IÆní6qb¿öjL’fßÙ¼6BÕºmÁë\¡¤õ€!­îŒN+´½p¦“>«ª²Åc}Ág¶ÿrñ•”Ãa±š;Œ´½±xWí°ÙþƒÊõÂÇ'nj{ŽÑaÖ¤Ve>rÜSµ©;ç«ŸÙúkÔÛ¼ôÂvã‚¿unÎ""iÆÍŸ×ýñL“˜ë´j³ŸIù#ˆªkUEYÑ«Óký’ìõýƒó B EÚ=¿œ9”¯^SR™0)b϶“O‹X*jjšmûèPîÕÂÄ«ûvœú¯ˆ)‹dš]Oۧ廪IÒÿÞ³~Ï­t®¶›©iØL 54x²™JMöº|^qê•;¯¾ˆ„BA!éuó›>ÒJ~üX§íVÝû5馎½—Þ®ãñ‚®6¼ í¬fæ¦ÿó©ö#ºj³Do£Âﻸ©”\{! ê<§_|KÍbˆõ±?bógñŽÝï:d¨úpù,¤õÀy µ~XW¼žVs–÷¹µh¹ †óÏÀ厗ƒ6ð,ì©^ëJ Ÿ ÙË7o……6“HœqwߊeÇg.lÌÑî6išËjãWÍãË¢—~±]´pýzsfìO‹Ï'õÕ†SUY"’¦]Z»%qÀ÷+ítX$͉=¼8¦åäýó¿»:qŒú8·WcH×ÄÔ¡OGu‰šµàä eÕ–ýwâ¼ÁÁv:,""¦VWßéž—CÔMœù2W­‡vÙ«ª«UñDWªÒiÜ’)VšL"b©ëéÖ !j\+eÔ­¦|ÙGŸ-V6Nº§^R#½èù‰p­‰  ä=ÄÑï>fBtä–â )Šžþ®ñår×V\yS1î7}Lôá‰ö~òß}¸FÖÏ.^¸£jÝÎØP_ÃjÚZ«zÛsªú¼ÒügWÿv>>KÂæªòÔÏÒ=‹dÄ«ë)Žô~µ½ÀÖn¥–{-SLTñdAFäæ÷„õÐg½ŽØº=zÁ×vZÎeæôááá666Îà=çôDLÝžCô¾Yfî¿IŸIy T3U³ K5äÿä\-9y@$ÉL·ïUî'{¶¾™QÞÓ ©³ª^iÕe‹Ò“8|£2¿“r[v6æåW^ ƒÉfÊ$Òš¯¹´ ÏȺ£fí®6¨q­”b0ÊD2“I$#ªcp+zû’ÓyHÙÍ*Æ–­yŧtÄY‰)Ï/3Í”‰ó„&9’â8Æ6òY´üíÓ¢oüþGìÓ4Y+§IÓ|:©ÕÏunJ?¯èÅ‘ÍéŸÍoÌc’Lœv)xm½l±&½_]/Ȥ“]yߤ\=Ëñ}}q[9Mt}õßY6îeçôò'Z¼çœžˆˆ×yܼæ]Õ‰*\s_Ç{ªÞ§8KǘõôI–´S‹ÒQQœ–ðJ­µ&«Že¹Í[‹Ï%:w. «E)qI­¦ßA;ïE¦˜šsë£VU«q­Þw>šwüpÄN'¯‘n­¹5¯•‰èÌK³Ei°/LºŸ(èXüª¡M'›Žß}c¥¦d›…éY¤Ó¼SþzôEÒÌÈeAû;¯›jÊ­·]KÑyˆW‘÷L¾Xèh,oI›§Æa >ô>YÑ3^æjwh^éO©¨PÂá”0X\¶X(*ÿ;}xx¸©©©ÂÕÊÏêa°h²˜UÏéKߦÝɚ߬ҨÅÕmžó0±ˆH&xvöçíѹ5'mõŒÇ¯…%CJìíø¬šÜEÇÔé3ÖìÒúC±Ùò{šEiwöl¸×kœ­F]ËrÚtOß¹çv†˜ˆH’óÛŽ›ùLóLv ;[Fll†¤~jUµ×ê=¶rcöü´çà™ŒæzœZÕjwîÎí)Eòûö殟®d–ÌQÙF^£5o9ó¬ ¸Ge¢´磋[L–{cÙ¼}qù%/e¦d1xe¢ÙûîÕÄcMNÒy;’$ãî¾E«ofË>ô>YÐ3ãbd6v-*{s ì“~=ùŸ@FDÒ¼‡G œºë3+ÞO¯,ÒNì`N_Õœ¾¥F«¨O{Mœ;¶‹:CV·Õ®[qßFí¶ žl§ÍT5=èâò€o™\†”Óª÷ aü›CwšïÓšCÒŒ?¯;öLHÒœ§qÙ±ó£Õ˜ õÎc&XjÈGt5«É#n­˜1¹PKW‹Ça¨˜¸ŒéÈ "’fßݵrÇ{ÒàÕ9ß~û™Aê© GÇ3—îP™?ÉV‹I*ý~wü—å{%L†LÆmã>í{wc‘4ûή-g¢R´ oJâä³›÷Y-šÔ™Ä©—¶Ê무,SÏuÎ á¶Uab.O]ÃÈi„®û…Šf´í¼úŠWG¤øŒ0*mÁ*×,J<±nóÕä×K.íæyÎ,½°»ê¶ªºVU••fGï\¹ãÎ=ÙÒí¼ù_ÚkK’ÏlÚu;Zz¼ã|Ÿâ› .Þ/Rs˜9΢vçΙ:}¾õ—ìÞ´à¯B—ËViÓÎ7Éó~ Y™7Ë° G­óøyŒð˾#–[U…£ÒÂf_ñÁ"ƒ©ªgȸµ1ð—72‡Ä"u‹áßo÷îHCé¾Q§ÏËÔs™ÖïçõR†´H¢ÚÉËoºÍ’ý‹Wå;§øÖŠ÷ÜnÕ½_“^ É›kçŠ\ç´WpR…ÓzpÀ_7d32Òë9þ[¬J×Þ+œÖã2= ¢ª2ÜÌ›7¯©7OndÐrqÀR×Ê÷ J3./œÿÀ7ô ÕFT«Ú+Œ]é;ë©ß¾-#ŒÙø:|0…ñ[gâ/ î«[›‹8Äb±<Ã|Öîíí]!KMXX˜²—*¼ n>Fõá¦É?„K”y™ÑCñÍçL]‡¯‡0b…ªVµ&É|’ÒÚoÎ „ùª(9Ž;|šS­Â|å¯gxx8&ôPY•×Þ7ÉkK3nnßxü¹PÆIÙ¦Cg7TÒFC¯£]­j‰ÕjèúMø|hÜöþh_Ë2 Ÿ‘gjjZvjŽ_è šHïèè؃=S·×W‹z¡VÐÈ)|F^é¯õ˜ÐÀ»‚9=ÀÇHÙ׳ôºUÌטh€&7§ ¢×¯_£éàÓäââRùbøèàì="= Ò"= ÒÃGAðÏ"W×û…h Dz¨9IúÍ-3¼{˜óù|s['Ÿ¯6ÞΖÖ4ôÞ rª[è­Õ¸F^3gõ7ä Ï-Üe×èHÓ/,˜þ‹`ÊÖ‹cmuEoÿŽŒ¯a$•ˆ¥ \W–~÷!ƒÑe˜ÓC-ˆRbžˆ:ûìi¨ÎUmfdá6òsm&‘¬àÑoó†ôìÌçwî18ðP|¾ŒH5ßÑuöú ±Ÿ¹9 ˜³~æŒÃ©ÉÆ9vëæ<ùXŠ¤Üq@Ê…e~N–f|³ÎV}|×Ä ˆ¨(ùüª nÖ|>ßÊíËÐé’¼kó˯AQ©2'JÎÞ ¢æ;ºúo ž4ØË­Oû dèJÌéA®qkÕ=«~ü©`¸³]W3.CþB~ôÊ)+ž ÿåÚï6³uò—SVv8½¨+‘0ùòƒ.ÇŽ.n§J$ˆÉòœÃÝt:ÈRµÂj÷7‡œk|qoßæŒÜ—Ÿ«q¨0nÓ‘®OoëÓ,õ\È„Ù :„o]±iĽwkÜ ªXJarD‚õ±ÃÁƲ'ÛGŒ >éq`xKºsz¨€¡í¼ì÷Ðá¼;;æöèÙµ÷ÈÅá/…D$ˆ?z¾ÈmîÝtÙl]ûIîâóGâ ‰ˆ­ÓoÊ vª "FU«e©0óÅ¿È15M,»ê³ñaÇ2œçMsj¥Ââ™x̘btïhL^u¥”n€£ï>ÉÓ˜K¤bâà ûêî«"ô%"=( ʼ¶îÓVî;u#ö~Äþ‰ÜcßþöBLâœä­vúÜ♿~;mÁël±4 4«Ÿ<«Z|³~V›[KGö²èîýuhĉ8óEzÖÿþŽŽŽŽŽŽ.#6üÇç¤Õ”R¾+©ë¨Éw'‡Ë”Šqú‘ª ù*ÝGM²c¿Œy]Dl-CµœçéÅ3墴gÙ¼VÚ""F¹É<ƒd c,SÛzÌ’}gïÄÞØù¹`ïÜUÑEÍŒôZøl9)wíVÌí}ÃZ2˯¡R©t "=Ô…èå_k7¹õ$%§H”›x}ÿŽ(i‡î­Uˆg>ÄsaíÞ»YIÖ½½kϳúï\ñ×x–FsÕÜIy•/Â>?wêNb®XÆÖÔÓ×â²Y 5óQ>êg—nº’$’D–píRL†¤ü*—b ‡é¡N]¢¦Ï{yô‡Q.ݺZØ{úŸÒœºa¸‹H½ÛümsÛœýªwçν¦ž5 ض ›FŸËm;dŠSÒæ|›ÑaÉeOµKsãÿ~h.ffÖÞëó>_`ËãYÌÚµÊ!aù`[s~—ž>söFgI*¬AT¹zà£Â8vÐÂÚ Ðd………%$$ )>½oôƒ˜(Ìé>e¸ŸŠçhDzød¡éàäëë‹é>aø¾á„•Q‹b‚¨ùŽ«ã…õS‰ú][=Õ)&صï’Xä¾Àœþ#ŽñTþìhHHˆ|.¥(ðÝ RòìzDúFªÂ ò›è)S‰XÊb3¾,@“„1³!&ô¥ažÁ`0Ê€Hß(•ÆxFU”ÕpX±i„á˜}‘wî\ýeHK%:­2õ÷è‡7~vŽ_·ðdª„¨8çlç…§ïÇÅœhñçìgÞVˆ¹<þÈuáwââ¢~­ùG`ð¥LQÑ«“ç4gýýàï^®™¶!V@%)q]ºöðáµÍnOWNYyGh2`HËØßnfH‰ˆòc_a¸}ÞU­Úí*XUžŒŠ³ë.>z!âìÒîÒ¨•Ó6eŒ8õ êð—â?ϧŠåe•­¼lY5ìJˆôŸ–N¿ÙºësØzöC=ôä a«Ï9ËÒëæãiÑ‚ÇbitôÕ@­øÈç…DDL §™»é°XÚ¶æ8åŸ?ño¡â”¸bcaÆq¿E¤Iˆrc_ãz 3ãU»Ýšd×<:vYê>ǯ«&‹Ý¼ÏW3ìUKÊ*YyÍ2ó@eø¾áÈd2ùl^&«eBW–––¼§\GúF,#’d¾HÏŠòïïÈeÉD¹ Ó„´KØJsî\¾bÿ•„tƒ)ÉËaõʈˆXÍLš«ÈßÂÕk­™÷<["O‰Û­\JÜ›Ù"¶ûp“͇¯¾õñL»Îë¿“Ï£¢ê¶«xU¤S6»®8'¹@»—>Gþ™Ô Û—+^¹N 3ó"ýàëë"ÿ©¾BŒWzEž²Ì³e{®™‘^ £ ç–Ú)É9#|üK@諱{Ïîï¬Å>\îõÅù ’¬—iB""*JOÌUo¥Íz—·‡Ê¤Äeµê;¢ý†ß/?ÖŽ¸¥õÙ¾N*DÒ궫dUT6».[«•ZAJŽ˜ˆCD¢¬7yRåJU13/ÔÎÞ7„Ú=}LyæÙ²x sΖyƒ´0G¨jØÁX“E’´Ã_•¼º÷n–D’soÿú5÷U•¦Äe¸ŒèôïöþÖ8°·&Û­Iv]U³¡.² û¯¥IˆŠžßu#_Z³ˆôQBBB…é{å%ï(Ï<[>Ô+Ê9[öõ®S½<}†ž{^§‡¡JÉúôËZ;È΢ÛÄ¿ŒælúÖ’GÊSâ²Z8ùvIýûiËÁŸµåÔh»5É®ËаŸ·q|Áš!‡ú-Š·è]|"¿ú•@-!k-À' Yk>qˆôˆô€Hˆôð±S–dÉgé¡QÆ­êÿÙ†„Œk_÷ñ;™ùîQ=’¤ýCù|ëé—²eõ¼f@¤‡NôòTØݶœ[odHшôPÏr/O²÷Ù—$bŒàî÷NýV>’²YvüÌùCóš¡4álq–؈ëŇO_¼t`xbhðÉÔŠq˜mØöL÷–êÇÌŸl©Vü»ÂÇGþ|ÓåsW;ïÏZļ"!"ÊS’dVÙr…kDz¨KÛå‹&"Ž±×—ŽÊÎGß}’§1—HÅÄÁAWžë¶üÚšYzzvÑRiÙg KëâgåÄþv2³ëç}ôymx·|tèt¢˜Hi’Y%Ë®é¡ÚH¯ÙR»8ý K«¥Z^rIÂÙvå²Ä¾ÎS]GMÞ¿ —))Wú^–s÷×s¹–Ãûè1‰ÛÆËÛèéáãÏŠäIfÛ+H2«d9 Òƒr ¶ SR$‘ÇeIA¶àÝeqâÌç©ò$ó$|ó,»|ÂY"*Ÿ%¶ÖdÙ·^Î)ºó›••U·a[_H_üqôIá»$³DÅIf‰¨lòÙòË‘”Smm×2õÊß©b¢¢§v_Ëy>eù‘¡»¢2%’¬èÝë¯iT™p¶ö¤é‘û¯3\VŸºrñÂ… .\¼ô»¿Ù›‡ãdJ’Ì*K> MÐøúÄdDðÄþCûîl¡ßº»Cwý¥/q¼ûçnð±»›J&žþeÎ.\ðUïÙ2m³ÏäYbßãù5’ÔKïjùìíס…Zñ"½¡S{m]|蟹+çm?oþÐæºúæv½õ“åg4ì.€ÆYk?‚¨ùßém?1×׸@ k-À'‘àS†ßé?<û‘çÐ €9=`N‚©©)á½%$$ ïÐwˆô€!ïÓæë닾CßÕÎÞ|Êéé‘é‘é›Á?‹\]Cî¢%0´"Ò䊯wã—°÷œ¼!â­­PÏdOŽ-™èaÇçóù–Î#÷DeJJ¾€]GÌ”IÞ^ ö´r 8‘$B‹5\×þ;¹ú«½-ø|>ßÚuäÜí×RÑþˆôŸ®Íšëãb£Î­rNÚömðÕl4 @½P?Û?åóE1fs~½{'|Y¿¬mF¯»—'+óñë³A#g_³[õÛŠÆ´YÃuÍä¡ó®¶zêÛáëü üy;óDúOƒÉb±¸šmœü†µ/|ÿ¶ô€V’±j¢Gw >Ÿoé:nÕ¥Tùþ/L<»bb?>Ÿß¹›·ÿIb"¢¢äó«&¸Yóù|+·/Co¤ã« Ÿ4f][³ñÍ¢-ýùz<®–IŸÉk7 ÉÙ»ìxrÉ·¤(éÄü‘w× ñl‰‡‰6p×X‡ìX8ÂÞ¤§ml3pvèoR4ô ¢æ;ºúo ž4ØË­Oû ŠÕ0$"Ò4¤ù .¼à´î¨[fBÁã\~'..úèך_Ê”‘ vãWÚþ÷ôšQfê ¢Â¸M_Fv^xú~\̉€Î^p?>½-´æÐœUzl­iéã¬ñøìƒ\‰“ÿ ô¸ÿæ_ƒú¾{4X× w2(ßì ÅC “#¬>}ñÒቡÁ'åGé?Ewý¬­ºv¶²Y0hñBWÝÒWXzÝ|<-ZðX,Nƒ¾¨ù¼PöW–˳=Úð˜LU}3ûÎ:,ćËpž7Í©• ‹gâ1cŠÑ½£1yhZ"Iî›==£Ì2–¶¦$'½@JDÒ”ÈÛ™í¼Ùê Ì7‚®Q:ôqôÝ'ys‰TLt_Ý}UD„!±œ”jœ8 ÿØꮣ¢¦£¥Æb‘ tšŸsÿàòû¯$¤‹LI^«¿P&.HÊofg ZnâÌéYQþý¹ ""™(—aš#6†.hòXš-Ô„/3 e¤VQ$Ù©¹L-=5&qm‚ÖØ…ŒŸ¡qhóØN•£4h×(úˆˆ˜ê:jò)+ƒÃeJ Å2"IV‡Di DzøP\-ýæú:•áã_B_Ý{vg-–ðár¯/Þ±›kd?M-¤NeNò³›éµ0Úpn©í PŽj7[ÿ#‘oûû´úÊòbÿºšgúuMF>1´»ÏÙó³Èoªß,nXèÈvªh³ëš~v*sŽD¼é?¸ì |™Â¡Oi`«ù(ø§)´*ÎÞd¤…9BUÃÆš,’¤Ý8þªˆˆx澃´¯,½”(J…i£ã³$Ä3å£~vé¦+I)Ii ×.ÅàòU""F3GÿéfwN[sæqFaQnâsfý¡î·ÀǨ4º0uúÌÛ·iÀóÑG_¡Í¬kfu¹÷㔣w“r‚œW1áf„§(ú”ÁˆHÿqãuèœèåé3|ôÜó:= UäKgíXÖûQˆ·µ¹¹íÿCòdD<‹Y»V9$,lkÎïÒÓgÎÞè,z9n‡ ;‚,ãÖŽêeÕÕÞ+ð¬Ö—»Úk–;ÆÒw :°Á5îÇÑóO$ã†î†êšvcw]Úçõîiݬ­{›–n?¢§‘•¢¡Où@‰!±ÜñÓÑ°ƒÖöع a˜šš&$$ ÞCXX˜¯¯/ú}W+b¢0§ø”!Ò Ò"= Ò7]ae4Ð&ë3!£ &صï’Øš.¯ÕJàÿOÎi ODAAA¥KBBBˆèÃ^F ˆôPoʆùÒq¿ Ò ú a¾”â›SQó=抆;g^þûÅÛ –ÍôŸVåóDT”|~Ãw+~½™$P5vü"dåôž’¿FvdÔé>úE–y;Ð>ôú.OÂ{A^¼M'¿ë _£è¿mÃ}o~uz—·>“ˆ î|ç1‡±þìÒnj¥•¤G¬ Xrä΋l‘Ša1A«ýû°Hš{÷<ÿЋI ã|µ$TüˆheË+×°·K雡))Ðè;@¤oÚ„É ÖÇËžl1*ø¤Çá-Yò”‹®OoëÓ,õ\È„Ù :„‡Z9·X}ñQ¾O¯Ì¨¨Bõ'—žuTó‡×™1oÚ¹ñËDr–¡û¨Ž a—S%$ËŽ:tMmÀp³òyo&Ã}tì²Ô}Ž_WM»yŸ¯fØ:(]®¨†o”¼0§oÊ`Šò0*L¹Èíábž¿þïÿ\^{8»ÞÝ»òîK«kÏ \,›•=Šc¸Ž¶X³ûBò0ï¸_ÿÖ<«c…‡F+N†›\ ÝK_ž Š¡fئ8á­8GÉrE5ÌËTüf@¤ÿDøúú†„„(ü©¾VWä)ÉB+¡¾&¯Ÿ<)ÐrœnÐQ›ÿfGxx"¯Û”–å»–©çèg»,ôÔ=›1­†}ߎS~劓ájµR+HÉqˆH”õ&OjBTÅrE5DÝQøfh É#š (»"¯V”¤\déÚ8èÅí>–oÛÝ€£næÚòþΓYæ.•òi3´»ûõLÛµpmŒÉˆþ­+á)L†«j6ÔEvaÿµ4 QÑóã»näK‰ªX®°†\%oÌé?ò¹»©©éûÍæKB½Å¬]«˜ —¶š+SÕïØ}pÀ2g"®q[Í-©ÎíT‰É²vl)‰Öv3רTœ¡e;Ê…1ö³¥ýZU:Îë:5ÐyF —§^+]Sû†*2"††ý¼ãçÍâÚ\WßÜ®·~²|MJ–+®¡²7@ƒ@ÖÚ&DübÇàa‘SÏì¨s9M²Ö6%²Ü»û~Mï3ÑIЄ`Ð×…!8áIDATo¤iálí']± œë Í@{4!ø¾iÐé{ï¹çvÀœéáÿª>SÍ6¾Í"}SÇ5òš9«¿!‡÷‚œ_ nœµ%ð;}ãÃÒï>d0‘m˜ÓzŠO§§]›?ãpjòqŽÝº9O>–"!*J>¿j‚›5ŸÏ·rû2ôFº„H5ß±ï‚+Ç;YðùöC_Hzq*øó^ùü£ÖßΖIR.,ós²4ã›u¶êã»&VÉ„$=bÕDî|>ßÒuܪK©eÔ 0§‡÷ á°bÓˆ{s¸›NYªÆm¬œ¶/QÑ«³Wtv1Ï>6}ØìÑw§Ž}Ç.šîÔBQæXE‰kU»W*«‰®@¤‡ú ™¬¤«f­DU¿¦¶õ˜%ûÆ,¥ßÿÕâÜU=/¯é®Vé] ×*(Â)[+häpö¾ci4WÍ}‘”WENØš\'|~îÔÄ\±Œ­©§¯Åe³ Ÿ‡«0q­‚²åkˆôð¾¸m‡LqJúÁÁœo3:,Y³˜µk•CÂòÁ¶æü.=}æìΪѕïÒÜøß¿Ú£‹™™µ÷ú¼Ï—Øò½×uj sB —§ÏðÑsÏëô0TQ\¶B­ÐM²Ö|²µà‡H€Hˆô€Hˆô€HUcу˜(4À'é ´C–¥ë±IEND®B`‚gnm/vignettes/fig-LCover45.pdf0000644000176000001440000012257412515147523015714 0ustar ripleyusers%PDF-1.1 %¿÷¢þ 1 0 obj << /Pages 3 0 R /Type /Catalog >> endobj 2 0 obj << /CreationDate (D:20061218162601) /Creator (R) /ModDate (D:20061218162601) /Producer (R 2.4.1) /Title (R Graphics Output) >> endobj 3 0 obj << /Count 1 /Kids [ 4 0 R ] /MediaBox [ 0 0 720 720 ] /Type /Pages >> endobj 4 0 obj << /Contents 5 0 R /Parent 3 0 R /Resources 6 0 R /Type /Page >> endobj 5 0 obj << /Length 40898 /Filter /FlateDecode >> stream xœìÝA%9v¦é½ÿŠ»¬ZEÒH¹¥h ¥ 1hÍB‚JBÒÕP Aÿûáá94£™Ç•{$Z€Ëç]TE~vŸkv##Âós Ú¿¼üõã_©9ï)z×Ê#ÖìšÄìO_ÿôøo~éÿÜ'Öÿÿ›ÿô]Ìÿ÷å¿ÿ?ÿø‡—ðøÏý~ òú㿼„]ÉšËåñú)»ÍÏüËã[G÷±þ±ÉŠÜÿϤ´¹¶?RnnÓ•šÛîò6VÚ‘’ú×s"—àj¾Ø§¸…UL®l3O±O,☰¼L````````````````````|1ct¹ÖÀ¾¾X—;Ù™gÿ:çïù˜ÿvïŽûúV9fÚ8ƒêŠŸYÎàyW¼»X©—òX9r–:ع#Ås9'úe»ØS,®Ä«ØÒÌ*Ž‰‹Ø',/_ÌÐ.WX銵˵FöÈIû×cþžçü·—¿üþòáñýåÿ~}dòÃ?¼þƒ,ˆý$û™…úøþúøÝ/¿|ÿsùû¯YXBråÍÂkAýÇïý¼š‹û#o^N¿_y®…™åÊŸvÔÙ»Üi¯®mcåÈ9D·i»=ìIÞõœH=_íÙQ7¡1ÈBËú³0&qLX^&00000000000000000000¾œ!]®6°¯/2!]®5²G¶þõ˜¿ç9ÿŽººÜO¡U¹÷ø§:ê±P:êúŽ:G—ûyÅ}^yv1Î,Wþ¼£Þd×T6•#K Ÿ’~fv¤×Ú2‘v9ÁÕ>E©åO1ÉÆ–§Ø'qLX^&00000000000000000000¾˜¡]®6°ÒQk—kì‘­=æïyοÙs£¯¯õ‘ctUïg®²9‡e9ƒç]ñîbx¤º¹mì¢9‡M.AÏ]ìÕ…¸L¤(·v¯ö‹œôE¬ûÌ*Ž‰‹Ø',/_ÌÐ.WX銵˵FöÈÖ¿ó÷<çß¿Ÿ¹/ýZu©üÔýÌsatÛö–Ý–>p#t“£s ²FÿÈd'ê4³|dOËíâ]ŽTú‡;¶ËÖœý~óz¤„N¶e"Ùeµg¹Ýdo”E ²Q‰eýé‹8&,/_Î.WØ×™.×Ù#[ÿzÌßóœÿ@¹]]’Sh®„÷«æ]öc>-0ËrÏ«æÍU}cÒK9ûèvkþõHéGò¶L¤4wŽ¾UÍýõþ¹.b’°<Å>±ˆcÂò2ñÅŒ8º\?X©šµËÕüË‘­=æïyοÙs£¯¯á‘cgÆžeÜ?mYÎàyW¼» »^gm™Gξ¹ìÜÇ‘â½ÜÊ}N¤"[N¯ö‹KûU¬~fÇÄEì–— Œ/fh—« ¬tÅÚåZ#{dë_ù{žóo»â&Û{ä¸I¥ÜÏ`nßg–3xÚïcËçÔOy=·æì«ìS1ÎÝŽÔ]n‘>'Ò6»ò{WܤÐ^Ä 'mY?1±ˆcÂò2ñå érµ}}‘ ér­‘=²õ¯Çü=Ïù÷·°òYêê(›H¿»…Å]ÚFµ¼ë%fçëÌrÏ«æñDCù0š^úÈÙ—fÕ¬Gj“>ýœHIÞbµOÑ·UL.…™§Ø'qLX^&00000000000000000000¾˜¡]®6°R5k—kì‘­=æïyο_5KwÅ'¹ùgžè7V·åtÔÙíò¿â|W^eÏ ËråÏ;êÝùòH5Ù¦šåÆëYÏÛ‘êeëés"yyîâjO±H»~÷8³Šcâ"ö ËËÆ3´ËÕV:jír­‘=²õ¯Çü=ÏùtÔÙù~ ­¸í';j]X³Üxý~GÝœ—Gÿ%yà`¿ò\i3Ë•?í¨û‡³É†$›ky¬9?WÎ#»w­,isåÇ÷Q7i×18¿Í¬? cbÇ„åeãËÒåjûú"ÒåZ#{dë_ù{žó訫l(R|vµ½_5×趢Õò¸ºæÑ(k– x^53L©ÊÃeåÈÙ·ÙíÐv¤mLΉÔ\ûaÕÜ_÷q“ÛÒÌSýû!Ž ËËÆ3´ËÕVªfír­‘=²õ¯Çü=Ïù7OìëKÑ­2ƶ"µÊmÐ–å žwÅŵöHû.´¬YJíh·rÛ‘½ØÖÑ–·êöëS OqÛVqw%Ï<Åm[Å1ay™ÀÀÀÀÀÀÀÀÀÀÀÀÀÀÀÀÀÀÀÀøb†v¹ÚÀJW¬]®5²G¶þõ˜¿ç9ÿþmÉ5ÉMÓ%ŒÝ;~fë ](Û~äòû™«kû#ÇÝÅÑ£· {{X–+ÚQ7ï6ÿH9Z»­9û(wtÏÌŽô¼_&ÒîÚõ^éÙQ7WÒUlefýY±OX^&00000000000000000000¾˜¡]®6°¯/³ËµFöÈÖ¿ó÷<çßÜÏÜ¢”×yó®Œ;ª[rm›YÎàyW¼ÉHÚ›mЬ9‡sï ;²ï.‡e"SXíó\úI/brqŸYÅ1±ˆcÂò2ñå érÃÜû¢i—æÞ–­=æï9<Ûû¢e—õ|qlݪ‹if9ƒç]qq->Òæ]Ô&\sMn·½/ôH9ïY9ç}Þy|¼é"ö¡Î<Å>±ˆc¢œwAc```````````````````|Mct¹ÖÀJW<ºÜÙÈÎ<û×9ÏÇüÛ®¸º&ÏÐk²WÆë#úàržYÎàYWý¸ŒTìžhËÙ'y ¬œGÚx*à9‘’ ?Ü'¹ærk›Y?1qû„åeã‹Úåjûú2»\kdlýë1ÏsþÞGßs÷ouœArµÌ,gð¼+Þ\ÞúÕìº{…åì«ÛµçžGê6zîc"móÎãkW,çR÷UL.ú™U‹8&,/_Î.WX銵˵FöÈÖ¿ó÷<çß}¦ž”Ï©=JÈnKï>/úì²< /˳åª ûÌrÏ«æâªlæÜd_ Y9röMî¬ÖK×#m<¤ïœH»‹íbŸb¨«¸»ìgžbŸXÄ1ay™ÀÀÀÀÀÀÀÀÀÀÀÀÀÀÀÀÀÀÀÀøb†v¹ÚÀJÕ¬]®5²G¶þõ˜¿ç9ÿíÍmÉÕÕí‘c±-,Bp©Î,gð´+^ªé$Ï»ÛÇÊ‘sÈ.Y‘nGJr{Z&¶*§´Úó\š´Ü±†™õÓ±OX^&00000000000000000000¾˜¡]®6°¯/³ËµFöÈÖ¿ó÷<ç?p[ru¡Ÿä¾ËÖÄ·Û’?°0·âJù- ÷$ üm ó›w|{uˆ.¤þIT—Æv!!ÉÛ–å³~ÞŠo.•þ‘Dyn¡¬9{ï²ý,é‘"ÕX&Ò.?¬ö£«~“ qfÇÄ"Ž ËËÆ—3¤ËÕVZqír­‘=²õ¯Çü=Ïùo÷;¨¥.—‡þE©çå ªómf9ƒç]qqûþHy“›ÀeåÈÙgWí!€v¤yÙ‘äœèç®;yœb¿ìEì±Í<ÅVqLX^&00000000000000000000¾˜¡]®6°Òk—kì‘­=æïyοµ”ÏÛØØ9‡Ü–\¥îÍqwelÖƒK~f¹€§UsôÎׇܩƇ£9K·m5¹©ýÃX'R|V57—âUÜÓÌúaŽ‰‹Ø',/_ÌÐ.WØ×—ÙåZ#{dë_ù{žó¨š«óòl¿ìüö~Õ£óeì½½«cr{˜Y.À?­š7·µñahË®Yvè­e·#u—ÒûœHÞÅî€ÑÏeßV19ŸgVqL,☰¼L````````````````````|9Cº\m`¥jÖ.×Ù#[ÿzÌßóœ¿jŽÁmylÂÑòªæì¶}TËql¶ûûÄ™åžWÍã)…©…qaË9l®mvéz¤&Ù%äœH›ì½Ú§è·UÜÝVfž¢ßVqLX^&00000000000000000000¾˜¡]®6°R5k—kì‘­=æïyΠjNòÁ”ƒ+íUsu¥>r›D÷ Ø‚Û¶™åžVÍ›w>»â÷¶°òYOa¯÷-,>°0•àR|ï ¹ñYÁW]wT§ÝÕ}f¹òçuq9=6tiÍÙ79 ýÌÆ‘¼m bysõº…ó³«mwÃÌ*Ž‰E–— Œ/gH—« ¬tÔûèr­‘=²õ¯Çü=Ïùooîg®.§Ñ çqŽ9ÈÌ–å žwÅMÞ"íYöy–sYvÙóÜõÈޜߖ‰þ–ò¸ÀÇÛ®¸–û³— 8,OQ ôCÔ ËËÆ3´ËÕöõev¹ÖÈÙú×cþžçü·{WÜ××þú]»oääR›YÎàiWœ;);{쮌;’5gŸå–ìqîz¤øèrX&r”[¬W{Šcð"ÖmfÇÄEì–— Œ/fh—« ¬tÅÚåZ#{ä¬ýë1Ïsþý½/¤|<«ïÝ-,òX$Û*oZ‡ï®ú™åžWÍÅ¥úHµ›cÛÍ9ç­H·#-É»žiŸ$¼WÍÙÕ¸Š» ifÇÄ"Ž ËËÆ—3¤ËÕVªfír­‘=²õ¯Çü=ÏùoonK®.•GÞ6Ù‰¸ŸA .„™å žwÅMÞ"UïŠ^ÝÈRjÇY“ë‘–].ËDNn¿nqŠ!.bñ.噧â"ê„åeã‹Úåjûú2»\kdlýë1ÏsþÛ½+îëwy^“ÊYÎ Éõ,Ë<íŠËØ#õK°«yl¬lçnGöâJY&dænw\¢KÛUÜËÌ*Ž‰‹¸—™— Œ/fh—« ¬tÅÚåZ#{dë_ù{žóïoaÑx9IïRz ‹’¯¼Er»tÔ–åžWÍÅ%­š«ÞÐXWn9»V/ö£ÛÊU,ufÇÄEì–— Œ/fh—« ¬tÅÚåZ#{dë_ù{žóßÞtÅÙy}^ÜÇì²{†e9ƒç]qqÛöH%º =÷ÈÙ—ñP¾qîãH yÚ:!Oókû<—²_ÅÖfVqL\Ä>ay™ÀÀÀÀÀÀÀÀÀÀÀÀÀÀÀÀÀÀÀÀør†t¹ÚÀJW¬]®5²G¶þõ˜¿ç9ÿíMW\Ýy+®Žs¬Þµ2³œÁ󮸹";lD¹ZÎ}äì“KÁÎ]´"§pN?﫾wÅU îSìç²ù™U§¨–— Œ/gH—« ìë‹LH—kì‘­=æïyλwÅ5¸²=r .è$©©-Ë<íŠktMvØØ•#çàe[qîv¤_MIËD®òˆ»Õ>ÅØV±ÿC˜yŠ}bÇ„åeã‹Úåj+]±v¹ÖÈÙú×cþžçü·7]q’‡Øå­ºPÆì.×™å žwÅÅÅ2šï2öÏÐ,ïgW<Ž”~d/ËD_‘ÜgWüUlqfÇÄEì–— Œ/fh—F+]±v¹š9rÖþõ˜¿ç9ÿíMW\åä¼õ±q_qó²y…e9ÿ´+n.ïýjªËzGòÈ9$—ì®i;Òš Û2‘“œòjŸçÒßôû¹ÄmfÇÄ)ê„åeãËÒåjûú"ÒåZ#{dë_ù{žóß^þòûË_üUx„Ç÷”ÿûõ‘]Èÿðú£|î'¹5çëãûëãw¿üþñýÏýåï¿Þzæ?~ïo\μEçÛ¸€ä¢ŸY.àiÕܢ순ö '*+GÎ!»d[5Û‘ÚÆ¥ÙÏ2úV5÷×cXÅÍå4óûÄ"Ž ËËÆ3Âèrµ•ªY»\kdlýë1ÏsþýªYºëô(a“ýŽoUóJ‘¾ÅtÔÉÕòȱ¹6 Øvéß-Ë•?廓 ­Šóa~š¥L϶í‡Ù«Ûã2‘žÝݲËñ*Ö<³Šcâ"ö ËËÆ3´ËÕV:jír­‘=rš·?Ûü=ÏùtÔI~H¹Hû~Õ\]Øy cgÇ潫qf¹€çUss©=R-²†\úÈ9z·Ù½âv¤>.i»×ØS¬®n§(çÊÌ*Ž‰C´ ËËÆ—3¤ËÕöõE&¤ËµFöÈÖ¿ó÷<ç?P5ï.å~’Áµös·Cë½ÿðævè-ŒÎo¿uáþn+¾ù wˆçX层òY'ÙÞIJ|ÖÏZñÍGW}ÿH¢ÛÃX9röÕmº?Ê<"dôËD r·ûjŸbH«¸ÉÝê–§Ò*Ž ËËÆ3´ËÕVZqír­‘=²õ¯Çü=Ïùo·ºXÖïrßó¸ó[Î`w)Í,gð¼+.²HÚ«ì "+GÎ!Î&x©eöКStíjO1k˽ˆû>³Šcâ"îûÌËÆ3´ËÕVºbír­‘=²õ¯Çü=Ïù·]quÁûˆ½l(²…ÑßZ–3xÞ7¹U:ÕÍmzî#çPæF#óHIºYÉÌ©“ÛÅ>Ïe/‹ØÏ¥Ÿ„eÇÄ)ê„åeãËÒåjûú"ÒåZ#{dë_ù{žóïî¶!åóÖO¡Vjï [˜ö,;…¼»÷E.éÃÃ6®<9Ÿg–+ÚQ‡(—6¶ÑÏld¹Â¢w+Ï#¥È¥Ÿ)þøáƒòºßW±ÿµ™§Ø'qLX^&00000000000000000000¾˜¡]®6°ÒQk—kì‘­=æïyοßQïJ?É=Hãü~Õœ¤û–j9ë%îÒ¿[– x^5çÓ#µ¤Ï]´œCue·K×#%Ë®ÙçÄÖœ/{Šy”䫸û™U±OX^&00000000000000000000¾˜1†­•ªY»\kdlýë1ÏsþUs’N8ÕìöTÍã1‰R-oc—æèå},Ë<¯šÇHª»«cj͹"aVÍzdß\¨ËDò²YójO±Ž’üû¹ø0³ŠcâuÂò2ñå érµ}}‘ ér­‘=²õ¯Çü=ÏùTÍ»ÛävÓûUs ò¤À£Kã¶ì¸ÉÛ–åžVÍ}¤”Gj›ËûX9²l'­j¶#{39'ÒÈ«}ŠR’Ÿâæ¶8óå$NqLX^&00000000000000000000¾˜¡]®6°R5k—kì‘­=æïyÎ{ùëÇ¿¼øþ®þ±þÿßü§[qÿ¹ÿïÏ/²í´ü——¤fNÑ»66¬ŽÚk–3½ X inR’Ëey]ãòúîêúºÆóõþ™„ý|Ýâòzr©.¯k\^ßåɈçëÏ×kMKŽ×-.¯'Ù‰ä|]ãòú.·Ÿ¯k<_oÁùí|Ýâòzr[Z^׸¼¾Ë(çëÏ׃÷ò Èc`æu"És&— ÍëÄ.ÿš,š—‰þïC]>æ™×‰äB['4¯]õë„æe¢¿T—{æu¢«qмNì.më„æebëÿÊ/ùÌëD»¼œš×‰}ÜœNh^&’_~¥ô ËëDÿ5¼~ê–׉ÝÅõS·¼Ldïòú©[^'ºº~ê–׉]žºLh^&JW×OÝò:±É_ÈX&4¯»ìã¾Lh^&v/9=',¯›üfµLh^'vùkË„æe¢z·­ŸºåubseýÔ-¯ŵõS·¼L´®®Ÿºåu¢«ë§nyèêú©[>'bÿµ—‰™×‰þ‡Î¶Nh^'Š<§v™Ð¼Lô_ëaùÔg^'6—Ê:¡y(㡸ç„æeB¾Ýº|ê3¯Û؈êœÐ¼Nèß:'4/›~Û÷˜°¼Nlò=×eBó:Q\Y?uËëDsmýÔ-/ý×z\?uËëDqyýÔ-¯íòGçÌËDÿµ¾þá9ó:Q.|μN´Ë 3/ý×úúGèÌëD‘oA/š×‰®®Ÿºåe¢ÿZ÷ë§ny(ò5Ï2¡yh®¬Ÿºåe¢ÆËŸ¦3¯åò§éÌëD»üi:ó2ÑâåOә׉rùÓtæu¢¹t™Ð|Nl¾«Ë§>ó:Qä΄eBó:¡>'4/A¾qLX^'ÊåOә׉vùÓtæeâòuçùuè9ñ—ßðåí¯ÿôòíþŠñíø¿?Œÿ×9Ë]›|zz‹EôrÅ»ëöìrº¬+Zײü¦µ¬Ûâ‡Ö…å+ëuaýØÂ-Ë/ÇeaJ[˜“|ØËÂü±&ô/Âû—mëÂ}6¡õ[.ïX>öáÄþïAÿ‚d]ø±'ö¯ìúŸþËÂýcNìNÜÖ…õcNìNÿsm]ø±'ö/ÚåÛÇ>Ùžq¿ü+×ôÃùã÷—þG†ü§jOèy}h,%Ͻ ï¯oú3#.¯÷?¶åuËëúµÍñºÆóõÜ¿FoþxÝâòz‰ò»ßñºÆåõ6þÔ:^×x¾^ä+üýxÝâòº^ðñúyýã-ûx—ôËG½%ýh\³|º¸ý|þÛëúÉÅÖ¥ŸY×ÿC³]b]î?wAßÏÿ̲þSšôí~êÝúÏtÿlü¦óËŠü Ðô×ñÏ,ëÿ^Dý+ÙÇ¿àýø)?øá×?=þÛãŸÿ?;6û•￲¢þÖ:¶1-ý?ꃞîßþîïþö÷¿ùÓ¿þø_÷Ë¿>þþ?þîŸþô±ßüúfôë6ýƒE°ÿë¾òÉÕË¿úg-OÓŸ¢ÿú§¿ûõ_ÿç?÷ËÓóX>ŒßÒÍßôÏÞ—­_õù_”¿Ì\ö c¿Ì_S×xL&ê‘üz4Ö,BìÿñÚ¿0°Ÿ”Žýðz®÷F¾¹žþkµÉî²ów­þÅD*3Ë/ûg7J¦8~¦Sÿ "ŽßO4gæóÛæ‘þ[P¬Ë„ÜZ/öýø²~kœYÅ1qû„åeã³ÖKk›üú2{ik—l]ò1ÏsþÝýdAÿA¾Ýb߃ú·öHqÜבcqI/1˽ˆ–åž×æiìó[æ·é4g_äfýpôˆ|óÃ/)ÏïÔÝkómܳwŠyÜÇSÎïüéÄ"fÝÎaäeãÒKk›,µ¹öÒÖ.ÙºäcþžçüûµyŒò—ôËå¯Á¼_›Û#÷é0þŽI¬òl-Ë<¯Íõ¯á—&+EVŽ,7úgûë/v¤eÙUáœÈ^þRÛjO±Œ¿ˆzŠÕE?³ŠcbÇ„åeãÒKk›,µ¹öÒÖ.Ùºäcþžçü·[õb“?÷ŠîÝåä,Ë<í½·ñJ’½~uW¦‘sèsv/½Ù£ì•sN¤,;Ü­ö)Ê6±§ÆÖ‘š§â*Ž ËËÆg5¬—Ö6ùõeöÒÖ.Ùºäcþžçün³?61¿Ý.þ…)gê½{Ÿy?ﲇ'½ò,ƒ²,Wþ¼oß\KýSl²¼¬9‡â²=¦ÁŽ4/O :'ä™DåbŸ¢<4é“Ëuæ)ö‰E–— ŒOkh/­m²ôíÚK[»|dë’ù{žóï÷í[Ï#õM¶^ÿ™¾]¦\Ú~˼ͧÝþì;fÙ þ·,ôòÝŸ_( ¼Yøƒï)ô+Ü’‹>Ÿ¹Êc€-ËÏîóï)ì.l”êüNÍÈYŸU«ÿ^Œ#¥Ñ}l"·ù½œû÷Š ^Ä}ŸYÅ1qû„åeãÓÚK‡Ñ&Ë÷´—ÖüË‘­K>æïyοٻfkÎçÑs·ñ¨Ó¤ð¶,gð´÷NÞmñ‘ú;¥±ëæì›|›dœ»iÞ¥¼LÈvôñbŸçRÚUlefý4ÆÄEì–— ŒOlH/­mòëË쥭]>²uÉÇü=ÏùÜK_]¬âãOßÙ>¦]z³éÍÛûÌûXL¬ÞÈ•'··™åÊŸ÷í›üµ€´'WÆd5gŸœ·ý~ôÈøîA\&r‘«=Åè꾊i\fÇÄ"Ž ËËÆ'6¤—Ö6Yúví¥­]>²uÉÇü=Ïù÷ûö\꧔›kØò=eÙ©fÜ–>n¥O»kuf¹€çµy‘B?•è¼®YæºÙMøz¤„ùY›oW\·“?ÅN-âî¶4óû?.☰¼L`````````````````````|ZC{im“¥6×^ÚÚå#[—|Ìßóœ³=Kª®ÄGÞÆ#YûäàÂ>³œÁóÞ»¹Ú²oº+5‡¹Îo è‘Ú¤¹?'²—cVûû¹bÌÛÌS짨–— ŒOkh/­mòëË쥭]>rÖ.ù˜¿ç9ÿÛÅwWó£Ä涟{b©.L±þàQ§oï3ïç]ûʘe3¹ò$że¹ò§}{—’ö*••#ç0±:>3;R79¡s"y¹Ý|µ§å›q3«8&.bŸ°¼L`````````````````````|ZC{im“¥o×^ÚÚå#[—|Ìßóœ¿oï |z”­ÉIþLß® s«®ý\Q¯ e—øýû¨÷ nml·ñ‘í.ç™å#{^Ôe·ïêØ-Gsöåøö‡)ÞË÷(ΉþCýásKsve»Š-̬☸ˆ}Âò2ñi í¥µM–¢^{ik—œµK>æïyοÙ&Wtãó2¾™P¼ÛÓÌrÏ{ï&÷° PõÜG–o Ì]âíHÛ\,ËD?%Ÿ/öy.:Å~.ÁϬâøÇSÔ ËËÆ'6Bmòë‹LH/míò‘­K>æïyοé½KpÛþÈ1º²3Ø\‹3Ë<í½K‘§¡n¶…ºæ¼ì$3ÎÝŽ´ì¶°L¤âêõ^òSôaûÇÐfžbŸXÄ1ay™ÀÀÀÀÀÀÀÀÀÀÀÀÀÀÀÀÀÀÀÀÀø´†öÒÚ&Kï­½´µËG¶.ù˜¿ç9ÿ¶÷Nò¬Ñ¼ÅÑ«÷3Ø]3Ë<ï½³«i½º‘s(n³VߎÔ(Ï(='Rž½÷ÞÙE¿ŠÅ¥zý¼ÆÄ"Ž ËËƧ5´—Ö6Yzoí¥­]>²uÉÇü=Ïù·½÷.ÝsŽÅ¶Dß½ÛüÌrÏ{ïæü¦w¯•šåÎò¹‹‹iqœû1!O9õ{ŠÕmí*–}fÇÄEì–— ŒOkh/­mòëË쥭]>²uÉÇü=Ïù7½w_ßÒ#Çæ6í½7—ÚÌrO{ï½Kqåyì¢9‡ñÃ8w;R³lA~N¤6ïf¿õÞ{p¹^ÅšgVqL\Ä>ay™ÀÀÀÀÀÀÀÀÀÀÀÀÀÀÀÀÀÀÀÀÀø´†öÒÚ&Kï­½´µËG¶.ù˜¿ç9ÿ¶÷N.l¼yaä ŠlpbYÎàyï]Ò]\ÂØ£Esɵyîz¤ÿ`ׯ9í²¹ËjŸç²ï«XœO3«8&qLX^&000000000000000000000>±!½´¶ÉÒ{k/míò‘­K>æïyοí½w¹µ±ˆcÂò2ñi í¥µM~}™½´µËG¶.ù˜¿ç9ÿíå/¿¿üÅ_…Gx|ÿGù¿_YNÁ?¼þƒé{z¤]ˆï¯ßýòûÇ÷?÷—¿ÿzëÌÿø}¼ýÞWn›k£Ú¯EF,Ë<¯Í“ŒŒó7škÎÁ»m^ºiý3Y'ò&íùjŸâ–V1»fžbŸXÄ1ay™ÀÀÀÀÀÀÀÀÀÀÀÀÀÀÀÀÀÀÀÀÀø´†öÒÚ&Km®½´µËG¶.ù˜¿ç9ÿövñâª<³È¼œAs)Í,gð¼÷®.ÈÍð›m°¢YÀi‹Ï#{v­-ÉËè«=ÅÝåí*V?³Šcâ"ö ËËƧ5´—Ö6Yzoí¥­]>²uÉÇü=Ïù÷oï B|¤²¹¶ÝoÿÀÂ,—ËÂdóðwoPoÞù:ö/›|d-º²Í,ÙÓ¢¾¿ÁÖß.÷CãÖvÍÙWçu¥)²—º_&úO‡}£âVÔ÷sÙãUì?o–U±OX^&000000000000000000000>±!½´¶É¯/³—¶vùÈÖ%ó÷<çßÜ Þ6õ¹iì=Ó²«qf9ÿ´÷N²UŒlE³ic>²<1´Ø^ìv¤&¹GþœÈ›KábŸçÒü*fêÌ*Ž‰E–— ŒOlH/­m²ôÞÚK[»|dë’ù{žóo{ï"{Àä­ã šóaf9ƒç½÷îv¹ý¾Øí÷šsØ]³ ÉíHI.çe"¥»}Š¾­buÛ>óûÄ"Ž ËËƧ5´—Ö6Yzoí¥­]>²uÉÇü=Ïù·½ws%?rÜ]–'‰fß_÷3Ë<ë½³÷®Å‡<Šttö–s(R´ËÊyd¯rËù9‘Ú¼ßýÚ{Ëë±®b›Ö-O1ÖU–— ŒOjÌ^ZÛä×—ÙK[»|dë’ù{žó÷Þ[ÖWyðf’Â]Î »­Í,gð¼÷N.ÈcEw9eY9rÇ..óHõÓÖœ’ËþbOqsi¿Š{žYÅ1qû„åeãÓÚKk›,½·öÒÖ.Ùºäcþžçü»â”}²x/?üÄó4manUªü÷ž§™}q^ö÷²¡¹\y•Ì-Ë•?ïÛw·ùGÚÇ?eåÈr…¶rÙ“ki™È^žzºÚS,®”«ØZVqL\Äþ–— ŒOlH/­m²ôíÚK[»|dë’ù{žóoï3o²‹­¡½´¶ÉÒ·k/míò‘­K>æïyο߷÷1<Šn«¨Í³|g@¶]/c[˜¸»=Ì,ð¼6/.Éf;A®XVŽ,ß(vk¿ÙõÑ©ÇD .\·xŸb–½âqw~ŸYÅ1±ˆcÂò2ñ‰Pf›,µ¹öÒÖ.Ùºäcþžçüjó$û¶Ÿ\ ?u›º.̵ºæ?зW·ÉcN£óã[ ›—e–åÊŸ÷íÍÙl§ ŸÙÈr‡¿Ýà>ìEv¶9'RvùjO±ÊÇ}Šý\b™YÅ1qŠ:ay™ÀÀÀÀÀÀÀÀÀÀÀÀÀÀÀÀÀÀÀÀÀøÄÆ–g›üú"ÒK[»|dë’ù{žóèÛwÙ·]ö®ñ¨Í·à²ì _]ûÚlI6…±,ð´6ߢ«ò Ø:ö»yXÓëMèóÈÞf%¯9õÜ.ö)†ºŠ›Kyæ)ö‰E–— ŒOkh/­m²ÔæÚK[»|dë’ù{žóßnÕ·¬ßõ©¡›öÞ»‹uf9ƒç½wq^ö®IòøTY9²Ü¿';w=Ò‚ a™Èã¢W{ŠY¾[pKšYÅ1qû„åeãÓÚKk›,½·öÒÖ.ÙºäcþžçüÛÞ{wMžÞ]Í{ò.í3Ë<ï½› í‘ú¹ê3R5示×s×#{qaèoÚÅžâ¸Éü"ÖmfÇÄE¬ÛÌËƧ5´—Ö6ùõeöÒÖ.ÙºäcþžçüûÛ¤ôAo9/í§¶%×…%xiÜ~aÚÇ“Mßݘ%ï Èòº×{Ú\)3ËGö´¨O㥩ìöèTÍò Óh{ÒعÝ~H»ü°ÚS ’±¿eœYÅ1±ˆcÂò2ñ‰ 饵M–¢^{ik—l]ò1ÏsþÍ ê)¹Í]ÌuÏõT\Í3Ë<ï½³Ëò4ÔÝå8VŽœC{ö©)yÜ\LlÙ±Ïsii‹‹afÇÄ"Ž ËËÆ'6¤—Ö6Yzoí¥­]>²uÉÇü=Ïù÷oPO›ležk•3ù™ûÌÏ…ûþûÌw—äŸÁ•ñÓìO3Ë•?ïÛûÈÓP7Y ŸÙÈ9æ÷ ôÈî]+Ë„|¸×MÖO1ÄUl.ù™§â*Ž ËËƧ5´—Ö6ùõeöÒÖ.Ùºäcþžçüúöâöø(1ÊníïÖæýíK5ù>v´É›‹ÛÌrOkó\“›ðƒ˜²räìG7>.ÝŽìÍmq™HA¶ªYíSÜÂ*F—ÛÌSì‹8&,/ŸÖÐ^ZÛd©Íµ—¶vùÈÖ%ó÷<ç?P›7×£ÄìÒÏm‡® sÛ¥°¿oß\•ç”6WÇ7 rq)Î,Wþ¼oÏRè§ZÜ>ž¹ªYvŠ·Üç‘Ú?š´L¤8ŸÊzïÛ“Ëþ*îufÇÄEì–— ŒOkh/­m²ôíÚK[»|dë’ù{žóï÷í}Aè']štÒï×æ»óå‘ãØ|F. ÉÃG-Ë<¯Í«Û¶ñ´R¯+GÎÁ{ÅÛ‘}w9,ÛèõW{Š»|á"¶}fÇÄEì–— ŒOlH/­m²ÔæÚK[»|dë’ù{žó¨Í‹ì»ž›—­T~æ6u]XBvÛ¶…)^¾3c•óìW^¢«~f¹ò§}{ ²}*› ÈÊ‘³ï„ÝüoGÚ&àœHIþÀjOÑ»ZW1ºPfVqL,☰¼L`````````````````````|bCzim“__dBzik—l]ò1Ïsþ}{“ý`r+®|`w—~-IvO÷.‡–ìZ›Y.àymžäfýTv·éÊ‘eÿš`;âØ‘–]ŒËDjr·újŸ¢ßW1»-Ï<Å>±ˆcÂò2ñi í¥µM–Ú\{ik—l]ò1ÏsþÛ­úÎ¥¸yÓ‡œö3h.ì3Ë<ï½w×ü#ÉãNÇÎ4š³Ï.ÙSDõHñEvš9'乨õbŸb,«X]N3O±O,☰¼L`````````````````````|ZC{im“¥÷Ö^ÚÚå#[—|ÌßóœÛ{7WƒöÜc ™=º­Ì,gð´÷ÞƒóòLÔ*w¸ËÊ‘åÆôh­¾Ùç®î–û¹mébOÑ»”¯â¾Í¬â˜¸ˆû6ó2ñY 륵M~}™½´µËG¶.ù˜¿ç9ÿþ6)}ÏcwñÞß&e×6²uÉÇü=Ïù7·‹ïÅÙÊN.çD?·˜/öy.u[Å*wÍ[VqL,☰¼L`````````````````````|bCzim“¥÷Ö^ÚÚå#[—|ÌßóœÛ{7·ÕGŽÅµÑÍ× ƒ–å žöÞÕËs7G‹?Î]ó؈ÜZ};Ò¼=UÔrŠ®´‹}Š>®bp±Í«a½´¶É¯/2Ѷ£]>²uÉÇü=Ïù÷·IÙ«ì'^üæjz›”]Þ9z×ô²<~Ó²\À¸’Öæý ò#íIÞGVjní¬ÍǑܪ|oà˜JÿÇOÕì¯Ç°ŠI-O±O,☰¼L`````````````````````|ZcôÒÖ&Km>zéÙ.Ï<»ä9ÏÇüûµy ²y I6p¹ÕæZX¥Ç¿oÏn/³Ý(_«ÛÂÌråÏûöÝù4>Åay™ÀÀÀÀÀÀÀÀÀÀÀÀÀÀÀÀÀÀÀÀÀø´†öÒÚ&Kß®½´µËG¶.ù˜¿ç9ÿf{–¾¾Éã7ç­ô-ÈBËrO{ïæ]Œýj¢ÛÆ÷ 4çd¿˜qîv¤ìòM…s"%Y±Úó\šËí*Ö2³~câ"ö ËËƧ5´—Ö6ùõeöÒÖ.Ùºäcþžçü›Þ»Eäñ›U*i9ƒ$ûºX–3xÞ{o.É]ôÕî¢×œƒ—ê~œ»)ý-Ë2±—ÒÅ>Ïe¯«˜œÏ3«8&qLX^&000000000000000000000>±!½´¶ÉÒ{k/míò‘­K>æïyο¿¿J n“[Î7¹ýgöW™ ³lþ[öSn¿aaÊE>´ß°°Œ›âßÝ ¦eùBŽI6x—Ÿ¤ÝÕ:³ü$=ÿÖ@qÅ÷÷‰ãfý‡åìw¹G_zõÈÞz¶ebÛ]ùáÖëý\Ú¾Š»‹ifÇÄ"Ž ËËÆ'6¤—Ö6Y¾5 ½´µËG¶.ù˜¿ç9ÿo $—ËCÚóŸíÛÇÂⳫ?ÙðÛ;nÎï(ê«l¤“ct»|dÅYfY>²çE}s»<0¶ºM¿£¹BaõãHÞwù‹ Ç„<þõfŸb(§X¼wi›yŠ}âmÂò2ñiÑK[›üúb½ôl—gž]òœ¿çcþ¾w¬ßåÙ¤^¿™P|’¿>`YÎàYï]|t­ö«Ùä!­²Rs‹ò— d¥)ÁË_#8&rö¬ö)ni7WâÌSì‹8&,/ŸÖ½´µÉÒ{^z¶Ë3Ï.yÎßó1ÿ¶÷N®ÉðEîs—3Ø]Ê3Ë<ｋ û#%ïê>Vj®mÞ\oG¤zþœÈéØ»çÞ{g—·«XÃÌ*Ž‰‹Ø',/ŸÖ½´µÉÒ{^z¶Ë3Ï.yÎßó1ÿ¶÷®Î7Ý€|4ïÁ»’f–3xÞ{7¹m=•êrç>²lã‹»)MÞòœØ²äÕ>Ïe‹ØÏÅû™U§¨–— ŒOlH/­mòë‹LH/míò‘­K>æïyο»!¹é[?¥½ºí§ö·…){¹Óü·,lrçú{’—\” È« cCò°¹ºÍ,ÙÓ¢>D—óC¶™im¬ÔܲKöÓ¡GrÝäÚ‰œ²äÕžbp-¬âæB›YÅ1±ˆcÂò2ñ‰¸Ï6YŠúÑKÏvyæÙ%Ïù{>æßÜ ’Keܾë ê»óqf9ƒç½w–‘”ûäØLÝrk.Ù¦2zD^ªÛÎXNÉù=ˆS^~‹ÛêÌSì‹8&,/ŸÖ½´µÉÒ{^z¶Ë3Ï.yÎßó1ÿ¶÷Þ]É£m̽‹af9ƒç½wuM+Úì1£–[q±Ú¹#r¯{ŒçDÞú¹´‹}Š±­bsyŸyŠ}bÇ„åeãÓ£—¶6ùõÅzéÙ.Ï<»ä9ÏÇüû³„âj[£·üþ6)ýíkÒçvŽKŒãNuËrOkóåq¤©Ù‚]VjnéØÏ\äÝ»ìÏ ©ôÛoÁ¥z÷2³Šcâ"ö ËËƧ5F/mm²Ô棗žíò̳Kžó÷|Ì¿¹]<ŽûÊs,ò@Q9ƒþuf9ƒç½wv[èWÓÜ>žRª9Ë–ìö”R;²'9rN$ï¶ýbŸçRö«ØòÌ*Ž‰‹Ø',/ŸØ^ZÛdé½µ—¶vùÈÖ%ó÷<çßö޻ܙžã6¶aég0NβœÁóÞ»J9Ÿú¹FmõGξ¿¡=ƒÔŽ”âê¾L¤ { ¯öy.µ¬b“Ýb,«8&qLX^&000000000000000000000>±!½´¶ÉÒ{k/míò‘­K>æïyοé½7/ÏË̱ÿQ¶èZ™YÎàiï½WÚ#•,ûËÊ‘³/òœÎqîv¤V·ùeBÎíº‡Ê)ú¼ŠÑmqæ)ö‰E–— ŒOkh/­mòë‹LH/míò‘­K>æïyοí½7Wü#÷ÂhÞ·"›„[–3xÞ{'Wë#¥Ý5]9²¼S›½·)»ÔÝçDösóó{ï\L«˜]3O1¦U–— ŒOkh/­m²ôÞÚK[»|dë’ù{žóo{ïâö¦ûšŒý½·æ¶4³œÁóÞ»ÊÖ)i¶7¹æì«\‚ž»)E6-?'žîs²íò¦q÷3«8&.âîg^&000000000000000000000>­QF/­m²ôÞÚK[»|äc_›¿ç9ÿ¶÷n®ÕÑsDZ¿wŠò@KËrO{ïþ±ay™ÀÀÀÀÀÀÀÀÀÀÀÀÀÀÀÀÀÀÀÀÀø´†öÒÚ&Kï­½´µËG¶.ù˜¿ç9ÿ¶÷n²ñ·ôÜúdÍdËrO{ïâÝÖôjÆÝìšÇåöLN;²{ÛÅÅrÚ]Ëû<—=¯bp~›Y?1±ˆcÂò2ñ‰g›üú"ÒK[»|dë’ù{žóozïÝyó.nã ’«ef9ƒç½÷ær}¤½ºÖÆÊ‘ÇN⶿·iE¶79'Ru)\ìó\ZZÅ$›Œ[VqL,☰¼L`````````````````````|bc ³M–Þ[{ik—l]ò1Ïsþmï]j£çnc§•RÏ3Ë<ｋÛ÷GÊÍ5mõGξ{´è‘âãØ£å˜Ð­WVûöŠ»\„å)ö‰E–— ŒOkh/­m²ôÞÚK[»|dë’ù{žóo{ïêŠìç½Ës0ûìÁÅ4³œÁóÞ»Éâ)o.Ôqî#Ëåû¼ß[´ìj[&rz²¿w}‹‹¸{y¸¦å)nquÂò2ñi í¥µM~}™½´µËG¶.ù˜¿ç9ÿ¦÷îë«ìç]ÍûÞÆ™å žöÞ{¿ŒüH{sAÏ}d¹³|³ÎÞŽô ½›ÝrŠOö÷ÞGÝ÷6³Šcâ"îmæeãÓÚKk›,½·öÒÖ.Ùºäcþžçü·—¿üþòáñýåÿ~}d)£ýÃë?È‚ 7¥'l~}üî—ß?¾ÿ¹¿üý×[gþÇïÏΗGŽmÞ.¾»f– x^›·¥‡Üü^Ç+šåœ›m“bGZ±mb,ç.Õ‹=Åìv[YÅ1q[y™ÀÀÀÀÀÀÀÀÀÀÀÀÀÀÀÀÀÀÀÀÀøĆôÒÚ&Km®½´µËG¶.ù˜¿ç9ÿövñ±ŠôÜaœcõ®ú™å ž÷ÞÍåí‘Jv±ŒsYÞ)&;w=Ò’<™óœHõnŸçRÛ"ös ûÌ*Ž‰SÔ ËËÆ'6¤—Ö6ùõE&¤—¶vùÈÖ%ó÷<ç?p»ø.û‰—X\l÷ÛÅ?²°¿Sð¿aaê§ìßÜ þ‘…ÙË–0ïÞÙ^ƒŒå°ËN4òYoòÍËòY?møkt{ì'¸Ë ÊÊ‘³ßܦ+íHéVØ–‰tüÅ€[Ãß_÷u;UfžbŸXÄ1ay™ÀÀÀÀÀÀÀÀÀÀÀÀÀÀÀÀÀÀÀÀÀø´†öÒÚ&Kï½´µËG¶.ù˜¿ç9ÿ~Ã_½ìœ^B“;æoµù¿¿üõã_^¼ó¾/XþÿoþÓ­ÃÿÜÿ÷ç— ¯?þËKŠAN(E/Ï•+ÕÖ\³\é}bæäfHn«ë„æubwy_'4/Û8ÏsÂò:Ñ?Ñ´Nh^'ÆOÏ2¡y™H~lLX^'¶±YÐ9¡y(ãïEœš×‰6¾ƒtNh^&ô/8œ–׉ââú‰Y^'šKeмL”èÊú™Z^'²l/´Lh^'šóëgjy™Ø£‹ëgjyÈã»ç„æu¢ŽGœš—‰þëe_?SËëDÿÇõ³¼NTÖOÌò2Ñ‚üE“sÂò:‘\^?1ËëÄîöõ³|NdïåIÃÇÄÌëD»?š×‰]~ãX&4/ý—xZ>±™×‰M~KY&4¯E¶­Z&4/ý%¿|¦3¯›|So™Ð¼Nù½p™Ð¼Nôßõ¶uBó2±EW×ÏÔò:Qd7¯eBó:Ñ.¿ÓͼL¤xùnæu"_~§›y¨n_?1ËËD²kØ9ayè&¬Ÿ˜åu¢ºmýÄ,/%¸¼~b–׉äöõ³¼Nì®­ÿZ^&öàüú™Z^'’‹ëgjyØ]Z?SËË„üá¸~¦–׉MþÚÚ2¡yü/š—‰æ/¿ÓͼNl—ßéf^'Êåwº™×‰vùnæsBžCÞ–Olæu¢Œ{Î ÍëD_󜚗‰]^>±™×‰,O3_&4¯ÕµmмLÄ(7Mœ–׉,{æ-š×‰ê²_'4/[peý*Çò:‘\]?SËëDu~ýL-/)È-"ç„åu"]~§›yØ/¿ÓͼLdùnæu"]~§›yØ]\?1ËËDé_­Ÿ˜åu¢e´~b–׉þ•Ñú‰Y^'ôaõç„æeb×Çú–× ÝÁðœÐ¼Nèßõ<'4/—¯zϯ‚ωþÕùÛ¯®ý§þ%{´/ÙÇÔãÿþ0þ_$ùalý±ÝíZ ò•ûûKû¿>ýl.K·.ÝÇïn—¥écK³Ïò…÷eiþàÒþõvI×¥åƒKûO|¾-Ý?¶´øþ_RÛuiýàÒþŸýËÿËÒöÁ¥r;Øeiô^—öÿKýwSùiè¿Uy¹C̲,òö_Uo&6ý:{äu"grš×‰þ‡L Ë„æe"ÇñÇò1ayèÿi06Gš×‰þUªË„æe¢È²íç„åuB¯ÿœ8?ñ_¿ö‘/ó–ë_j?ÔÇ ùÊõçê…ý?²Š-L?µ°%Ðô¿¹ÿfaî?›AßÑÿÔ:ùO9}ß{¿þ3¿ë]|?uEþ}ÐÛê-ýß’þ¥š¬«ó_ý¿~üËc+:õ£~ýÓã¿=þùßøýt³_wñþë.ê¿ìAžWÚ(!äÿöwÿ·¿üÍŸþõüÃÿú»_þõñ÷ÿûñ¿ÿôw¿~ì·þõ{ÿjtëÿAšõWñÿýfé³`÷òe{öU¾&–¥ÿµ/ý×ÿùÏý õT–Oä753ówý3é¥ôÏüüZû—™e£`/ Òü5vËÇü·ñS“th‹5‹ûý ]û‰ùÐÐ/çzïæýrú¿Òý×Pè¿Ï×qg”²Â¢üðô6ÎñBÿM~¬ÓØÏp·™ÇØÿÓ5¶óõþÙìW×4/_îZÿÒÉ¢~Y,¯¯ZÝâù:ëYÏzÖ³žõ¬g=ëYÏzÖ³žõ¬g=ëYÏzÖÿŸ_o-ªÖŸ¯/Ö¢Z:£•Ÿsøçð»w汉n”oë~àïéo²@(YÚù~âYî4³('þ´Þ· †íÁuå;GVÈå|]oËZ]Ó6¹áÔ²|ßÛâÐÆë§6^·x¾ÎzÖ³žõ¬g=ëYÏzÖ³žõ¬g=ëYÏzÖ³žõÿ.ë¥EݼջڢjüeF+?çð-Îá7J;ƼË_šîo]å/úZ”·~þ 4ùû‹!di¢û:©I“ÑW»$}½Ž¿-´ºÇY´mѪüµ#‹C¯ŸÚxÝâù:ëYÏzÖ³žõ¬g=ëYÏzÖ³žõ¬g=ëYÏzÖÿ»¬mÖŸ²ëçhQ­ ÑÊÏ9|‹søMAÛd?•P6}ìØe‡‹òÖÏ Ú}¼ÐvïËF ²¯….Ó1xÙFêx]ž`ö£hûË!žX3×dVˆ‹5^¶x¾ÎjV³šÕ¬f5«YÍjV³šÕ¬f5«YÍjV³ú?Új«'µX|}±zÒzÆ­VœÃ·8‡ïÍg_\ö>å-û[µ-Ê[?m>Ç€ÚØÖ /)Ä:7¿ÖQ¶ú¯çëµÌ.÷Ö|nòTK²—½&³¶°Xãe‹çë¬f5«YÍjV³šÕ¬f5«YÍjV³šÕ¬f5«ÿíÖzR‹Å׫'­gœÑjÅ9|‹søMó™]-°Ey6oë*}³(oý´ù6—¿ù/»Èj a“§Œ3bŒò Ìãõ:6r]YÊ<-uÁöÝÒ°Æ««Õ_¶x¾ÎjV³šÕ¬f5«YÍjV³šÕ¬f5«YÍjV³ú?Üj­'µX|}±zÒzÆ­VœÃ·8‡ß4ŸÍyù{èÞ%i>kpÅÏ(oý¬ù”¿LµŒ RšFaª›]Ä\içëµÉ¦¤+{œD[°V,«¯®VÙâù:«YÍjV³šÕ¬f5«YÍjV³šÕ¬f5«YÍêÿ€«¥žÔbñõÅêIëg´Zqßâ¾7Ÿ5ÊþŸ!Uy˜Sëäö6£¼õÓæss9¹ÜÞ¯…« r»î.w¿>‚÷nO3Ë?í…›l§üæê(Ê5nMêññI1·Åóõ}ÿñŽ´ý4j<59àgÜ88°| üÿV²WKÚײ×JÛ#[I;ÇoqNß÷>¸mOB+m¼ýæZœYÞþY•|‘'´ÙšWVjÎÉe½‰ØŽÄèeßÜc"öÚêdy݇UÜ\l3O±O,☰¼L`````````````````````|ZC{im“¥õÖ^ÚÚå™g—<çïù˜Û{÷ÊxÀ\ÒÚ}w!Ì,gð¼÷ήJÿîgc®9çç¹#1U}RÞœhIvþXíSŒ~ûIÔ™§Ø'qLX^&000000000000000000000>­¡½´¶ÉÒ{k/míò‘­K>æïyοí½wéžC»W÷3Þm~f9ƒç½w“í´E.Ú˜[.®;w=ÒvWË2±7y‹Õžbu[»ŠeŸYEXE™Ð¼N`````````````````````|VC{im“__f/míò‘­K>æïyο»O·Tï­=âæ?²á¶¼}“MV²ÛGµ6—ÚÌrOkóå)“!f—Gåo9µùpJ;û[ĶLȃ%¯öƒËõ*Ö<³Šcâ"ö ËËƧ5´—Ö6Yjsí¥­]>²uÉÇü=Ïù7·‹‡äÂ6¶ßÆV.¡¸Rg–3xÞ{g—d? ýÇÌ)çjŽsG¤—Çrµ<뽓ÞMοȈe9ÿç­y’">Äæ¶}¬ÔœFÖ+×#ÍÛ,–KqÉ_ìSÜÒ*f¹­ÞòeâuBó:ñY í¥µM–Ö\{ik—l]ò1ÏsþýÖóê’ÏùÇló{ïîŠì\R\›û§ç(›«¬,+xÚ{gïzœK.s••K>[}=£—w„^­Ê«9wûcÛÅàŽbÙÄ1±‰sbåmãUÕKk›üþf½ôj—ϼºäsþ1Ûü绤ŒÕµ±$Ù¿%|¾MÊøò-ÏšÜÏGÚóáR·,7àŸÖæÙ…±ŒTÝ1)°r9äùtýpæ‘»;Žm¢GWûmbr¹Þ‹õ°¬âœ¸ÇÄÊÛÆËÚKk›,µ¹öÒ«]>óê’ÏùÇlóÏÅyÙÞ|üA‹û&/ï\YVð¼÷®.é&0QïNsõ®ÛÚç‘ÚÜâÅ&¢OÏŸOÉ߉=[VqN܉cbåmã… é¥µM–Þ[{éÕ.[¶.Ùæó9ÿ±÷î.Ê6+ãp”ÁÕjYVð´÷>¼¼ùSö?Ò¼RsÉç&0z$¦. _}ÜÝýÖã×ZÚ±‹Á…dY?9±‰sbåmã… é¥µM~“ é¥W»|æÕ%ŸóÙæ?ôÞGtÙÏíÄÛ|âüÈR¯,+xÞ{'ÙíE¾@×»Ó\ª;Ö/z$Æ$eþ9}–*~·/Ñç]Ì.EË&Ž‰Mœ+o/kh/­m²ôÞÚK¯vÙ²uÉ6ÿ˜Ïù½÷!›®Èvâq¾˜óh.d˲‚ç½÷|sf(EäJÍuÜ]XkŸG¢<ß¾MD_]ûyï]ä]ž›8,›Ó.Ή•· Œ—5´—Ö6Yzoí¥W»lÙºd›Ìçüç¯Å<²kcIòù^‹9¾|ÕG¸k•(Á¥dYnàim^¼óå&@Ÿ¬¬<ZVm®Gbë­¢+ÇC~W°ÛV›w—ã½X½eý0çÄ8&VÞ&000000000000000000000^Ö˜½ôj“ßßV/mí²eë’mþ1Ÿó×÷:·EY½wvG´,+ðO{ï$’l<žôîVžµû\»‰ãHñׄ¬mm~îï{ïe‘wbë–Uœwâ˜Xy›ÀÀÀÀÀÀÀÀÀÀÀÀÀÀÀÀÀÀÀÀÀxYC{é:Ûdé½µ—Öü‹eë’mþ1Ÿó{ïCö½Ì×nÞd?•,Ë ž÷ÞÅå|@†_¹çÃðz$Öêr¹&b,.ý¼÷>\ó»X¯–Uœ›8'VÞ&000000000000000000000^Ø^ZÛdé½µ—^í²eë’mþ1Ÿó{ïæÒq }>þ>Kùî-Ë ž÷ÞÝ•t“}X¢vöškvÙž÷žGbk{ô•}´'Â{ïæzßı–X,«8'.Q'VÞ&000000000000000000000^ØHº{HžÏ{Wí¥W»lÙºd›Ìçü‡Þ»y_fèÕõ¹ÓÊôݲ¬àiï]£ì@.që+=W®I6ñžk×#1.Äk"úæüý³ä—Ú.&—Ë&Ž‰Mœ+o/kh/]g›,½·öÒš±l]²Í?æsþcï]•g°»<6.+¨.6˲‚ç½wq>ÜBN®êÚ5Ë-«³×#ñèRæŸ1TySçn›x¸TïÅ’-«8'îÄ1±ò6ñ²Æì¥W›,½÷쥭]¶l]²Í?æsþcï]å%›²¯Iﲂæ]®–eÏ{ïî‚îV^tíš[’WuêÚ瑘«õþš£—Í\vÛÄæŽr/¶dYÅ9q'¶dy›ÀÀÀÀÀÀÀÀÀÀÀÀÀÀÀÀÀÀÀÀÀxY£Î^ZÛä÷7ë¥W»lÙºd›Ìçü··?|ûýŸÂ-ܾÿEþçÇípá¸ù›×?‘ Æb ò>Ìïï·ßüòÛÛ÷¿Óß¾|p!L>Ì[2îme¹§µy‹.µ[H^ö>‘+5—._wÞº‰épµ\1ÒäﶉAò&Ž/-«8'6qN¬¼M`````````````````````¼°!½´¶ÉR›k/½ÚeËÖ%Ûüc>ç?<.Þ²¼?3t¿Xoŵò¬àyï}¸C†÷.ÏÍÏW®M¾€®}‰±ËÞ/çDô‡Ýÿcïå·›X\ –Uœ›8'VÞ&000000000000000000000^Ø^ZÛdé½µ—^í²eë’mþ1ŸóŸ?.Þ’;Ò-Eø~x\ü‹¶ô…ç̫˲­¹—6ý}í>[–;Þ·7ù¥@ÈÑÝØE³ì“Ög6ÈoŽm"Æ$în_bˆ»Øå/ÐÊ&†¸‹sbåmãeÙK¯6ùýmõÒÖ.[¶.Ùæó9ÿ…¾½È¦é±yâýÓÚ||ùÒnó>æcê}|ÝdYnàimÞƒëãËåºp_¹Ž/¿¶gÑ#ÉWó5S±}Ïjóq>…]Œò)®lâ˜ØÄ9±ò6ñ²†öÒÚ&Km®½ôj—-[—lóùœÿ°=˸¾ÕÙ×·4W0/\YVð¼÷>\È·PÆ@˜Wj®Ùµ¶Ö>ÄÃÏÊß&¢oökƒÇÞ;»Ãß‹µYVqN܉cbåmãe í¥µM–Þ[{éÕ.[¶.Ùæó9ÿ±÷®Î—1ŸœŸÄôîJ°,+ðO{ïæRº…£Iu/WjnÙ…õ¨»‰W§®9VkÆýCï]eÑwb¯–Uœw☈WKñú†/Ö&Kï={ik—-[—lóùœì½eð(Ï`gWd{ðè£kÞ²¬àYï}p‡¼\ôÎåJÍí°-ÕבXæZΉ˜ÆMÿlŸYËÜÄ(›­¬¬âœØÄ9±ò6ñÂFÌÖ&¿¿ÉDóg»lÙºd›Ìçü·½wr9ÝbȲÁŠ¬àp½[–<ï½³<¶Ž!‡y¥fÝ\×>Ä£ÊCìçDLãKþìyo9ïë..–M›8'VÞ&000000000000000000000^Ö˜½ôj“¥÷ž½´µË–­K¶ùÇ|ÎûÐ{)È£¯ÂÈ º Õ²¬àyï=Fü¸›îš^©Y¶j±ç½ç‘x—ò5ã!Ûìö%Ʋ‹ÍÙ²‰cbçÄÊÛÆ˳—^m²ôÞ³—¶vÙ²uÉ6ÿ˜Ïùozﱘ07X)³yÑ¥bYVð´÷ÁyÙÅ%èÞä–[rUßÚ¹ŽÄ”e çD ‡lj¾Û&zy(ýN¬É²ŠsâN¬Éò6ñªÆꥵM~³^zµË–­K¶ùÇ|Î{ûÃ÷·ßÿi¹}ÿ‹üÏÛáÂqó7¯"ø±èd‘ßßo¿ù巷刺ÿxèÌÿø}|ùÙ†ËÛ5«Þâøº‡e¹çµyv±#­GÝW®Õ•õ¸¸‰Å˺ΠùCþéãâc-%ß‹=ZVqN܉=ZÞ&000000000000000000000^Ø^ZÛd©Íµ—^í²e¹0ÏÇÃuþ1ŸóßEÞ—)‡7í½›«Ù²¬àyï]]®·Ð‚ëzåÊÕU¿Ö>Äìå͜焬íÉãâEÞº‰Í…`YÅ9±‰sbåmã… ÛW?{oí¥W»lÙºd›Ìçü狇cî/^ƒüáóÇÅ»Kí&/ß Un Ž{K–åžÖæÑ»"{ÈD×æ“ò+·æüÚ!FÄØÖ‡³²ünàÞ¾Dw1¸Ø-›èã.Ή•· ŒW5´—^mòû›Lôt¶Ë–­K>{ì‡|Ρ6o²Ã‹¼Øs\ðim£;d³ó$»ÃÈ $Z–x^›'׎[x-Ü5×ê¢}8óHL~m½¾r(®üt—•q>†]ÌòfÑ•M›8'VÞ&000000000000000000000^ÖÐ^ZÛd©Íµ—^í²eë’mþ1ŸóŸ×æ1ÌÝÍ[••}^›²é¹l²~Ì÷ŽÆæR°,7ð¼6Ÿ½|(Å¥¹ÁÌÊ=»n¿u˜Gb¶‡õW ÷À˜X\ö÷b©–Uœwb©–· Œ—5f/½Úd©Íg/mí²eë’mþ1Ÿó_¨Í¹›ùå)öÏkó溼Ã3¬—’¦ ›Ã¬,7ð´6O^ŠüPª¼“ôýf¹^Úë!C»&d»õöóMZºtõwb+–õÃœwâ˜Xy›ÀÀÀÀÀÀÀÀÀÀÀÀÀÀÀÀÀÀÀÀÀxYC{im“ßßä”ôÒ«]¶l]²Í?æsþÛã;=St!ߢ¯RAË ²+ݲ¬àyï\·pççþ,+·x¾ÓSÄ<þ¯‰ùjÏvg_k©m³ó‡eçÄ&Ή•· Œ6B¶6YzïÙK[»lÙºd›Ìçü·½÷áRšÏw×¹ÏKª®5˲‚ç½wqEž+o.Í VV®ÍÕµ‡Œ‰eî¸~NÌ™Þ?K~­¥×]¬.fË*ΉMœ+o/lH/­m²ôÞÚK¯vÙ²uÉ6ÿ˜ÏùÏ·IIÙåkùÒ6)©¹C·.éó‘ödçô•åž×æÝÕ>–¤ —[×Ü“ÜÞú<"¿(ýšØëOwçCÙÄì]N–M—¨+o/kÌ^zµÉïo«—¶vÙ²uÉ6ÿ˜Ïùo‹ë«l]2Ææ6)9»X,Ë žöÞ9ºÞn¡4tíšÇWêëqq=›m=¾rÊ¢_bÊ»˜ä•M›8'VÞ&000000000000000000000^Ö˜½ôj“¥÷ž½´µË–­K¶ùÇ|ÎûÐ{g×ý-ú¾Þü™«Ë‡eYÁóÞ»¸Po!σʕš{qÇz#¨‰eõš'܃¿³M<Ü‘îÅ,«8'îÄ1±ò6ñ²Æì¥W›,½w•^ÚÚeËÖ%Ûüc>ç¿}è½›ó²;I”]QÆ ïJ¶,+xÞ{wy»g(ã‚0×®¹eW×æçzDn*o²…Küék1ÇZjÜı–ñ%WVqN\¢N¬¼M`````````````````````¼°!½´¶Éïo2!½ôj—-[—lóùœÿöØ{ÁÅz‹þpÇ|-æ‘\K–eO{ï#ÊV(¡Ï¹rå|®]Äè×Þá+ûfO³?ôÞc-=ìb’ãWVqNlâœXy›ÀÀÀÀÀÀÀÀÀÀÀÀÀÀÀÀÀÀÀÀÀxa#Êî!³M–Þ{öÒÖ.[¶.Ùæó9ÿùk1/O†‡q÷¯ÅüÂ…1%y„ý×_(_±–Ï_ÄyŒ;)ã£;ææ1Gu>Z–l~º?-êÙû%  è_ÍÍË×Ð{™{ÇÄkB6}‰÷ö%¿‹Å¥fÙÄàwqN¬¼M`````````````````````¼¬1{éÕ&KQ?{ik—ßΞZ»ä·³Ç¾Ïçü·¨WYLhu= ^Æ`°,+xÞ{7Ù»%”î¼>\¯¹ÎbtíóˆÜMˆ×„ü¡ÿtc–q~,z»;ªeÇÄ&Ή•· Œ—5´—Ö6ùýÍzéÕ.[¶.Ùæó9ÿí±÷×·<`÷óù’\ê–eO{ïe‘¡ŒÉÙê¯\£ËkCr=Ó¸(_1tÙKf·M .·{±Ë*Ή;±ËÛÆËÚKk›,½·öÒ«]¶l]²Í?æsþÛ‡Þ{,&ÝB Ò“Ë Š;šeYÁóÞûpIöS)®ëÝi®É•¾Ö>ÄØdo˜k¢W~ú¼÷XK©÷âø“•Uœwb?,o/lH/­m²ôÞÚK¯vùÌ«K>ç³ÍûÐ{WÙR%ô šË º«Õ²¬àyïÝÜáo!—ùŠÐ›åZ\²ç½ç‘ƒÝf¤Šßík-­ìbw![VqNlâœXy›ÀÀÀÀÀÀÀÀÀÀÀÀÀÀÀÀÀÀÀÀÀxaCzim“¥÷Ö^zµË–­K¶ùÇ|Î{콫—7tJÏ]æç5Ê;@W–<í½k*=äàêÜÅeeYÐÚÅED¬N}Mô³/Ñ»]Š–MôÇ.Ή•· Œ—5´—Ö6ùýM&¤—^íò™W—|Î?f›ÿö¡÷N®øm/–Z\8ö½Xž÷ÞÙµv Ñžf_9{WÚZû<c’›:'böÖÎÇÞ;»˜wq| Á²‰1ïâœXy›ÀÀÀÀÀÀÀÀÀÀÀÀÀÀÀÀÀÀÀÀÀxYC{im“¥÷Ö^zµË–­K¶ùÇ|ÎûÐ{Wû|ï¥ïsÝ¥lYVð¼÷nÎ×!'—µ³×\üÕêÏ#1R½ŸÑû'ïµ”—o¦{±zË*Ή;±zËÛÆËÚKk›,½·öÒ«]¶l]²Í?æsþÛ‡Þ»Ë ,å=–m® Ew$˲‚§½÷‰²‹K[oÄ\¹®é.&ëˆìVê5!kK÷Ï’›8ïÅÖ-«8'îÄ1±ò6ñªÆꥵM~³^zµË–­K¶ùÇ|Î{ì½[’/5{n]ãáj´,+xÞ{g—ó-¤(ź\©ùrKzwóH ^b¿&úaû›?öÞIÊöM_­YVqNlâœXy›ÀÀÀÀÀÀÀÀÀÀÀÀÀÀÀÀÀÀÀÀÀxaCzim“¥÷Ö^zµËg^]ò9ÿ˜mþÛ‡Þ»¸TfÏæ{-[s=X–<ｫ+ò4{–kÊ•šK Çuí󈼥3í½ZïÿØ{Wù›Øäõž+›èý.Ή•· Œ—5´—Ö6Yzoí¥W»|æÕ%ŸóÙæ¿}è½»;ä=–uíÅÒ£ Þ²¬àiïݽkñÒár™Wj>dD×¾Ž´<×~NÔäÚ½}‰¡ïb?YÙD™¸DмO`````````````````````¼¨±zim“ß߬—^íò™W—|Î?f›ÿöö‡ïo¿ÿS¸…Û÷¿Èÿü¸òbLóú'R¤WÙ(åW¾¿ß~óËooßÿ6NÿñЙÿñûüòU¶Ï.ÎGÚû!²²ÜÀóÚ<;æ‡Qæƒæ+/û¶è‡3ÄÐÜѶ‰–\-w¶‰I~[p'–òŠsâN,‡åmãeÕKÏ6YjóÕKk»|æÕ%ŸóÙæ¿=>.>®ïº¹IÖÞ»¹Ü,Ë ž÷ÞÕE?îfÈq^©Y z»»y$ú(u÷5Q³ËñÎ6±¸£Ü‹-[VqN܉-[Þ&000000000000000000000^ÖÐ^ZÛdé½µ—^íò™W—|Î?f›ÿö¡÷î.Äù|w‘Ö“®T˲‚g½w’7oöy7³1·|ŒµëóÞv¤Ï]Á¯‰·`¹ÖR] B­¬Ÿ†N\¢NhÞ'000000000000000000000^×^ZÛä÷7™^zµËg^]ò9ÿ˜mþÛCï|tIv')²YAv­X–<ｓ<”bÕVßò‘]ÕíÁíH;\HÛDóöµ–žw1»-«¨—¨š÷ Œ×5¤—Ö6Yzoí¥W»|æÕ%ŸóÙæ?öÞ‡”ðá(.‡¹‚æüaYVð¼÷.òvÏ0€ w§9×ûZ»iŵ°MToO³?öÞEn{«,beeâuBó>ñª†öÒÚ&Kï­½ôj—ϼºäsþ1ÛüÇÞ»¹²öóö²‚\Ì–eÏ{ï.ïî !¹¦½æœÐµë‘žä ›×DM²¹Ën_bŠ›¼+Þ²‰2qŠkBó>ñª†öÒÚ&¿¿Y/½Úå3¯.ùœÌ6ÿ¡÷×7yûWNÊ æÆà+Ë žöÞã6Â1ä(U½\©¹\w§Gbè²Åø5ÑçMﶉÑá^¬Ý²ŠsâN+o/kh/­m²ôÞÚK¯vùÌ«K>ç³Íì½çåÍš^ÆdÕ•`YVð¼÷..åy7EsÍåܽ{‰)¹£_q|ÉøÓç½Ç‰êïÅÞ,«8'îÄ1±ò6ñ†/Ö&Kï={ik—-[—lóùœÿØ{7É¡7—æ£wÍ[–<ï½»;Ò¼›>wqY¹Ž[°V‰²·x¿&b÷ÿóç½›<æ~‰c-¡ZVqN\¢N¬¼M`````````````````````¼°!½´¶Éïo2!½ôj—-[—lóùœÿöö‡ïo¿ÿS¸…Û÷¿Èÿü¸²ùˆ¿yý)ÒóXR ²²ïï·ßüòÛÛ÷¿Óß|åÂà»´úÞ—íü>Ö\ηЫ«yÞy’ ÈW–;Ú·Çèjœoõú©j Nk_q=KYÏ°¯ºüŠ`·/Ñ·]L.Ë&Ž‰Mœ+o/kh/­m²ôíÚK¯vÙ²uÉ6ÿ˜ÏùÏûöèe#òØ“<ŸþP›ÿñûÛ?Üþß›wÞ ¶ÿýŸÿõ¡Ã·ÿ6þïooAÎßþû[ .ô[ñòBM¹ÏÙ€k”Û¼?­éÈòÌý<›]j×Ù™¶³Õõ:;Óu¶ÎuÙYMÛÙñéåëìLÛÙ*o;=ÏÎtm^¶°±³š¶³IvÎ9Ïδ-òßœggÚÎvù]Çyv¦ë¬ü•¹> MÛÙââõihÚÎv—Ëuv¦ëlðÑ•ëÃZq?¸–·ó3îç»ói;?ãv>D¯lÅýü!;]çgÜÏ7ù]Îu~Æíüø›½^ÜŠûùñ§u;?ã~¾¹P¶ó3nçÓøÿ¢íóѸŸÏîØ>ûùêêöùhÜÎg/¿‰;ÏkÜÏg¶ÏGã~¾ÊrŸq;xù•Øy^ã~>Éÿï_çgÜÏÏïK×ù·óãŒß>?ûù$¿w»ÎϸŸ/òÍê:?ã~~|WÛ>?Ûù]Û>?ûù"o¾ÎϸŸïû·¡·ó-î߈VÜÏû·¢÷óÍÕíóѸïÁõíóѸŸߢ·ÏGã~¾¹´}>¯óÑw\ŸÏŠûùìjØÎϸŸ¯®ûíüŒÛùœ¿>¿÷óÙŶŸq?_]®Ûù·óòϦëó[q?Ÿ\;¶ó3îçÇ?yóv~Æí|òû÷§÷óiÿþ´â~¾ìߟVÜÏ÷ýûÓŠÛyùMüöùhÜÏùuüu~Æý|—®ó3nçèŽíóѸŸ?ä·Ý×ù÷óMö}»Îϸ/QÞ~ž×¸Ÿ?\Ú>?ûùæŽíóÓ¸¯Á•íóÓ¸ŸÏ®mŸŸÆý|s~ûü4nçÇOIqûü4îçóþýiÅý|Ý¿?­¸ï~ÿþ´â~>ïߟVÜÏWyÛËu~Æë|òã‡í§,ûùñcHÛÎϸŸ?ˆÔíüŒûù.g®ó3nçÃ|qÌy^ã~~n°wŸq??ÿCÄëüŒÛùýÇÆó§ÈëüøÑöã¦?þ:~ÞëçÝùÓèüŸßÍÿÕ㟃ãïƒñÿ㯦üÄz òC硫¿AÆJö Ó×.œï:î®Ì_¼2ò#ì~åñÅ+Ç°%ß]Y¾ve¡û+ë¯Ìã_?ÒÝ•í‹WŽÆÇÔû•ýkW¦ñciܯŒÞë•ã__Æ·Ä12þ"È¿‘i’ šþ{ÈÃé4œé:Æ?ç~™zzÅíüø±¬úë¼Æíüø¶Ñûu^ãu>ÆCþ±hçWÜΗùcÅy^ãu~ÝŽßînþ«ßúð¶ìwÛ9~d•¤6Ù´S>»ß•ûÏü?¼lü¼1~^˜—å_qYHsÝóºøk®?EýzþW\Ç·C¿Ü¯ùjòO³ªO ýš›“o²ã_Ðä²bŽ‡¾mseÿÃ?ßþ×í_þƒoeiýÍÿæŸæ¿s ,Í…–¯ú¿ùÿþñ··ÿò×?ßþúã_ÿýÿŽov¿+Ç×¾aŒÏwüü6þ-ìÐÿGúßþ§_ºõñ§ãÛÏ¿þǸòßþõ_ÆÝýÛÿùçÿ§_¶Oâ?Ó&¬ÿ7Õ¼¥qÏ祿X”eÇÐ/öÿ÷ц¿Í¿Ùoãø±}\åŸ Ùþz|m꧷rÿ¬á‡[‘½tÃøY½è‹Ýòø»süKùÊòÿµÏ<Ìqþ5ã§À2¿]¬<~êIë»™‰ãŸ³Íoãßûxg›èåÂ;qü€²²ŠsâNlÑò6ñªÆꥵM~³^zµËg^]òÕc?d›ÿôÁC¹`ü!„*×ýŠÿÐ](¿jôéÓÿÐ?Çä‚üJó¼óCr[Yîüyß>< Ç0Ó¼RsKçoõHÌMž#<'ä9“ã~“<`±‰‡üÂeçÄ&Ή•· Œ6‚·6YúöÙK[»lÙºd›Ìçüã år,.ö1ßlMþãí•eÏ{ïùŸÔGßä¿f“+5‡bÿõ¾©ö¾¬,ÿ}ò}§~­¥§]lò°ûÊ*êÄ%êD½þ[ Œ—7b·6Yzoí¥W»|æÕ%ŸóÙæ?Î<²«XÈ]úôÏï.·¹Õ¢÷)ÎM›4Ë <­ÍÓ<#»nùyë+÷(¿ ˜·®Gb:d ™sBÞ•WZ›ó!îbýÝV6qLlâœXy›ÀÀÀÀÀÀÀÀÀÀÀÀÀÀÀÀÀÀÀÀÀxUÃzéÙ&¿¿½´¶Ë–­K¶ùÇ|Îx\|\_ä¡ï䊾 a¾TaeYÁóÞ{¾EM^.Uô-š[—ùèÝÍ#1ë×Wþùãâã| »˜å½%+›8&6qN¬¼M`````````````````````¼¬1{éÕ&Kï={ik—-Ëúxøê±ò9ÿùãâ)Ì÷|ÖC®ûôqñ4_¶cqAo ÉËW–x^›W ½ÚËgŽ¡Ÿ»¬¬#e~F×D²ÎÝ6±È«`ïÄZ-«8'îÄ1±ò6ñ²†öÒÚ&Km®½ôj—ϼºäsþ1ÛüÇÇÅ»óÇ|<ü˜¬ ‹·,+xÚ{gïR¼…±Ö1 Wj®ã–ÖÚõHKìÛ„têýþQôk-¥ß‹cÑ+ë§1'îÄ1±ò6ñ†?¬M~[½´µË–­K¶ùÇ|Îè½óȺ[yÍ{ήv˲‚ç½w2äæRWjnÝ…Õêë‘Ø’¬åœN>ÿt›”±–Vw1Ky¿²ŠsbçÄÊÛÆ 1[›,½÷쥭]¶l]²Í?æsþcï}¸,{´¹ýøXAu½Y–<ｋì@ã¹vÍ9›Ÿ¯#Õ˯‰R¬콋óe«KÙ²‰2q‰:¡yŸÀÀÀÀÀÀÀÀÀÀÀÀÀÀÀÀÀÀÀÀÀxUcõÒ³M–Þ{õÒÚ.ŸyuÉçüc¶ùÏ·ÏY6/½Ë#çŸnž›¼>3Fïò¼Å#¸P-Ë ø§µyw­ßB?äyt¹õ•«+mÝú<S´‡á5Çæò}%‰cy—xxw$Ë&Ž‰KÔ‰•· Œ—5f/½Úä÷·ÕK[»üvöÔÚ%¿=ö}>ç¿P›W×ä™ö*…þ§µùøòMvŸïî”È.ËrOkócWŒfÙ¾\®ÔÜ¢kë…¢z$ï|º&ä·駵ùåžïÄ-«8'îÄ1±ò6ñ²Æì¥W›,µù쥭]¶l]²Í?æsþÃ.+ãú.»´ißeÕ‡eYÁóÞ»¸Xçû7×ÝiŽÑõ¼Ö®GÆ׎}›¨Cúé.+ÇáJº{°¬¢Nì¢LhÞ'000000000000000000000^ÕÐ^ZÛdé½µ—^íò™W—|Î?f›ÿØ{7äìC*鱂âe[ñ•eÏ{ïîr¹…Þ¤X—µÏ,¬„µMÊ:’Ú|#èí>ïöµ–7q¬%xË*ΉKÔ‰MÄÀÀÀÀÀÀÀÀÀÀÀÀÀÀÀÀÀÀÀÀx}cöÒ^þðþ&³—žù—3¯+ÎùÇlóŸo“rTÙ|npž·IùÊ…e¤/ì¯R‚¼Z4†èŽÙø—¹åúÊrçOûö]9n¡®èg6sôÁÅõÛ„u$ÙxýšŸAºßòü}ØÅ$¿vXÙÄ1±‰sbåmãe í¥µM–¾]{éÕ.ŸyuÉçüc¶ùÏûöâeuá˜ûš^›gwÈ6äÕ5½êB°,7ð¼6?\Ë·ÐÒÚØeåÞ]Y¿ŒÐ#òˆ|j×D ɵûGà/1ú],.7Ë&F¿‹sbåmãeÙK¯6YjóÙK[»lÙºd›Ìçüjó$ÑÇ£ÈÆ1Ÿ×æUªëçÆ Tï’·,7ð¼6ï²õz Íù8o}Ïzëz$.‡m"—^›7—ú½X꽨»(š÷ ŒW5´—^Æ›õÒË8óê’ÏùÇló_¨Í‹lƒº—%üšÝ]ôByi­ÿ‰ Cî®ûÏ‹úqÃ]ŸAó#«ÉånY>²§E}#qn*ïçï(V^¾Æü°×‘\×f;+ÇC~õ°Û&w´{±–UÔ‰]” ÍûÆ«ÚKk›,E½öÒ«]>óê’ÏùÇlóŸõunØê|ÿóÚ<»n¡®Îækq¥Y–x^›Ï_F„]Ö+W¾ö¨×#²ôT®‰è‹KéÎ61˯6±8Ÿ-«8'6qN¬¼M`````````````````````¼°!½´¶ÉR›k/½ÚeËÖ%Ûüc>ç¿P›§¹ýz‹<¾P›WÙµ=Æ`O›wÙŽfe¹çµy“jäu¥1Í+Wî×o ôH=Öž9+çäÚý NM¬®—]ìò'+«¨—¨š÷ Œ×5¤—Ö6Yjsí¥W»|æÕ%ŸóÙæ¿=Tß¹y9S—š}¬ %ç‹eYÁÓÞ»WeK/o+•+WŽ®®Þ{ÉUþpMäÃ(è½Çùpìbt9Z61»¨š÷ ŒW5V/=Ûä÷·³—ÖvùÌ«K>ç³Íáqñîj¾ÅÒœïŸ?.>¾|•ÝØ‹ óy÷Vdde¹çµyv]÷†/úáhŽ]>½u=Rªëi›(AÞ¼ºÛ—˜ò.®Ë&ÊÄ%ê„æ}ãU í¥µM–Ú\{éÕ.ŸyuÉçüc¶ù‹y®\6LóõÖ¥Ó_YVð¼÷nR‡ZæÊn–ЬòŸGb®ôkB~mPâmbuGº›·¬âœ¸›·¼M`````````````````````¼¬1{éÕ&Kï={ik—-[—lóùœÿÐ{w/oE_úyË}|édYVð´÷îÁ%yÔ}scqËÝÅÕê¯#ò†Îc›¨‡ ýξÖRã½Ø»eubeBó>ñº†ôÒÚ&¿¿Y/½Úå3¯.ùœÌ6ÿ±÷N.Êî$^v[‘®E˲/KùiïÝ!O³{—ç“â–óÙê¯#ã~sØ&RGÐwûZK÷»8n¢YVQ'.Q'4ï¯kH/­m²ôÞÚK¯vùÌ«K>ç³Í¾=xîð·P¼<‰þ°=ø.ŒGw©~¾¯x/.—[ôÕúö.¿ XYîüyß^]M· ;«ûyåÌÑgÙ"F?3=’Æ=ï±È‡»Û—èû.6¹…•M›8'VÞ&000000000000000000000^ÖX½ôl“¥o_½´¶Ëg^]ò9ÿ˜mþ }û!{»„R\Œ_¨Í»+Ç-†ê’Ô懲ŸùÊrÏjóÃ{×ã-Œ[-q^©¹\ç•ëH<¢ÙšSu!ßÙ—Û.wË&Ž‰Mœ+o/j¬^zµÉïo«—¶vÙ²uÉgýÏùÇíYäú&¯ß,ÂÈ —ºeYÁóÞ;»n¡´¹qÌÍrë.öuwóH<ý¥€åØdI»mbr¹Þ‹ãüÊ*Ή;qL¬¼M`````````````````````¼¬1{éÕ&Kï={ik—-[—lóùœÿØ{çeòõBÐÃ7w4˲‚ç½wuÉO¹óʙ㒭]¤äRØ&â!oåÜík-¥Ü‹=[VqN܉=[Þ&000000000000000000000^Ø^ZÛdé½µ—^íò™W—|Î?f›ÿt›)ÒãXt-â~¶MÊịºK‹>°‚«Õ²ÜÀÓÚ\6gï·Ð›ëm^9s ]¶QŸ·¾ŽI¶;¿&RvþçÛ¤ÌÏuƒ ɲ~˜sbçÄÊÛÆ ÒKk›üþ&ÒK¯vùÌ«K>ç³Íx\ƒÒ\A“}ÅW–<ｋk² Œw¡Î+gŽ>Z«mGŽîªß&r'ÑwûcÚÅêŽ`ÙĘvqN¬¼M`````````````````````¼¬¡½´¶ÉÒ{k/½Úå3¯.ùœÌ6ÿ±÷n®ÊSÚã ÏÞ;)¼W–<í½£w^÷S sí+·*_p®]Äì¥î>'bˆ²ËnÛZºËñ^ÃÊúỉ;±zËÛÆ˳—^mòûÛꥭ]¶l]²Í?æsþCï=®ïõúÜpEVÝ-Ë üÓÞ;‰ŽæbœWj®ó=síz$¦¶^¹¹²l½òÓÞ[ÞÅîÅÖ-«8'îÄÖ-o/kh/­m²ôÞÚK¯vÙ²uÉ6ÿ˜Ïù½÷áB¹ÉžãU»ùêj°,+xÞ{y„<Œµê+=W.]ÞékŸGdÏ•T¯ Ù;¼†;ûZKó»X¯–Uœ›8'VÞ&000000000000000000000^Ø^ZÛdé½µ—^í²eë’mþ1Ÿó{ï±y;¹2»ùä]÷–eÏ{ïîJšzÔ'Å5·âŽc­}™wÓ®‰¹³¯µô¾‰c-±XVqN\¢N¬¼M`````````````````````¼°‘k“ßßd¢û³]¶l]²Í?æsþCï‚;äÝ›E¶Þ–dç»eYÁÓÞ;E×â-Ⱥv͵¹ºžU×#²×xM×DŒQ–´Û—Ú.&—Ë&Ž‰Mœ+o/kh/­m²ôÞyöÒ«]¶l]²Í?æsþcïåKÉþßE×X]l–eÏ{ïâ|¸…\å+É•škrm=Í®Gd-òNN›wf¶ûwfšxÈcîwbÉ–UœwbÉ–· Œ—5´—Ö6Yzoí¥W»lÙºd›ÌçüÇÞ»º.ï±LRdïrµ,+xÞ{wú¼›ª;¤h.M@×µÏ#1Ø>/+ûèúÏŸ÷nî(÷⸉•UœwbK–· Œ—5´—Ö6ùýÍzéÕ.[¶.Ùæó9ÿ¡÷ÎÁye–V{¬ 'WŠeYÁÓÞ;GÙŒ%´ÃUmÌWî²™É\»‘^÷P±ì]¿ßCåZK=v1ɦä+«8'6qN¬¼M`````````````````````¼°!½´¶ÉÒ{k/½ÚeËÖ%Ûüc>翽ýáûÛïÿ4ŽÜ¾ÿEþçÇípá¸ù›×?‘"=å[8º,òûûí7¿üööýoãô÷_¹P:û/¼/Ûÿø}¬+»äo¡z×æ=sqí°,wþ¼o?Ü!o.觪¹Dç×SôëH/òÑ\ãs÷÷ïê41»žw±¸,«¨—¨š÷ Œ×5¤—Ö6Yúví¥W»|æÕ%ŸóÙæ?>g^]–]Q’+]Vpxç³eYÁóÞ»¹*OÑ7¹Y»æ2… k×#­¸P·‰Ze›•Ý¾Äw±»ì-›(—¨š÷ ŒW5V/=Ûä÷·³—ÖvùÌ«K>ç³Íá9ó"ω‡ØäAõO_¾È6çÑ¥YÜÉÅdYnàim~Ù&Yö\‘+5—vÖæz$†";È\=ºvÿªÏKLa£;ºeÇÄ&Ή•· Œ—5´—Ö6Yjsí¥W»|æÕ%ŸóÙæ¿P›w×Ç’Ru>}¡6O²Jè‡|9¹±òhYnàym~ÈCó¡t×´p×ÜÒõáÌ#²ÃzËÛDïò ÓÝ61»Ãß‹µYVqN܉µYÞ&000000000000000000000^ÖÐ^ZÛd©Íµ—^íò™W—|Î?f›ÿ¼6„±$yîýîæ?©Íçþëa¬â˜/=º+Á²ÜÀóÚ¼¹¤¯6-zåÊÍeûpôHµm¢ïŸd7±ÊÎ0wb¯–UÔ‰]” ÍûÆë^·ÏeÖæÚK¯vùÌ«K>ç³Í¡6/.Ž%/ÏÁÿš—ê…Ñ'×üç}{ñ.êÃæ%Ê—(—­,wþ´o/ÁñŽ([¶È•+wç×úz$Ƽ>Õ5Ñç/vÛD/ÏÆobt¡XVqNlâœXy›ÀÀÀÀÀÀÀÀÀÀÀÀÀÀÀÀÀÀÀÀÀxaCzim“ßßdBzéÕ.ŸyuÉçüc¶ù/-Éå4Ï÷ù(}9\ï–eÏ{ïìª4ÿÑ¥ùÑ+Û®æëH¯òÈü5QŠKùξD_wñpé°l¢L\¢NhÞ'000000000000000000000^ÕÐ^:ǵ«yÑ^Zó/g^]ò9ÿ˜mþóçÌKte,É'y~ý×ÀÝ]“‡ªƒës5ºT,Ë žÊ58߇œe7¹Rs®îX…¯#½˜­Y¾äOåê]>îÅš,«¨»(š÷ Œ5V/­mòû›õÒ«]>óê’¯û!Ûü‡Þ»Î:[6×­SêáŽÃ²¬àyï]lcr×»;s´»Ó#%ËÞ*ׄîÖ²Û×ZJ¾{´¬¢Nì¢LhÞ'000000000000000000000^ט½ôl“¥÷^½´¶Ëg^]ò9ÿ˜mþóKj”gÍ{t½}¾ýH-.ô[HÁ%mÍ›«Ù²¬ÿyk^]–ýQ¢TáråÊç 2íÈáåW×Dʲ·Ën›X\K»Ø\–UÔ‰KÔ ÍûÆëÒKk›,­¹öÒ«]>óê’ÏùÇlóŸïRpËÓäÚÛ·àz²,+ðÏzïæ])·×æ“î+§èrе¯#ã)mÕËMïö%ú¸‹ÁÅnÙD™¸DмO`````````````````````¼¨±zim“ßßdBzéÕ.ŸyuÉçüc¶ù½w‹î¨óirÙ…e¬àp!Z–<ｓkǼ›æ•šSqy½ÄséA:øk¢fÙØe·/1†]Ì.7Ë&ÊÄ%ê„æ}ãU í¥µM–Þ[{éÕ.ŸyuÉçüc¶ù½÷ájÑž[WÐdq+Ë ž÷ÞÕù<î¦ÚÝiŽÝõºÖ®Gzq>m¥¹òÓ}NÚX¤¿Kµ¬¢Nì¢LhÞ'000000000000000000000^ÕÐ^ZÛdé½µ—^íò™W—|Î?f›ÿØ{7×e·ñîÊ\Aîð–eO{ïîeÃñ‚ì#.WjNQ@Ÿk_GÆZÚ&ªwG¾³m-Ýý^lŲ~:±‹2¡yŸÀÀÀÀÀÀÀÀÀÀÀÀÀÀÀÀÀÀÀÀÀxUC{im“ß߬—^íò™W—|Î?f›ÿÐ{÷è‚<ƒ¯sÙ•nYVð¼÷N²uJðeî<~³œƒKë½–ëˆÜTÞ&ds—Ÿîs2ÖRÛ.fçË*êÄ%ê„æ}ãu 饵M–Þ[{éÕ.ŸyuÉçüc¶ù½÷![‰‡£¯woöêZ³,+xÞ{Ûð·×ú¼RsÎòP·®]”CÊük¢—ã}­¥×]¬.fË*êÄ%ê„æ}ãu 饵M–Þ[{éÕ.ŸyuÉçüc¶ùooøþöû?…[¸}ÿ‹üÏÛáÂqó7¯"EúQn½¸oßßo¿ùå··ïg¿ÿx¨Ìÿø}èÍñ|wEvr)>Ècâ+ËúŸ·æÝÕ~“'ÀµðŸ1ŒëÏMRôHì²?ù5ƒ= ÿXšwÊ ï]N+š7ΟÞXy›@@@@@@@@@@@@@@@@@@@@ø;Vátû‘³ðÕÞöÌ«¥=ç³Í¡Ž®RóÊ(Gþ´–/_õš^o »X,Ë <룋² ‹ì;®Îü!Ï¡Ë…v¤ŒlÛDÊÒàïô¦¼Iv×h^Ê»7VÞ&þN-|µ§•>Z ßÕÛžyµ´çüc¶ùOûh©Ä{¾Å±’ö•>z|yÓetU÷êòaYnày]\¨·:ß )Î(û€[=Èþ$5\R´§Ÿ=]ÆéÎkaEõæù;o ¬¼M ü ³ïÕšVÚèY÷®ÖÖ¢u´6ý˜müq‹êâ›óÝÆÇ—Þ•ü³Ê÷c™Ü]*~®{FÙ»ÛºõH ÎÇm⨮ÿìÍŒ²”/p¬ÄûÕ›ç/OVÞ&þn)|µ§}“ )|Wo{æÕÒžóÙæ?Ê!¸Xçöé˜+H²KÇʲ‚§…rˆî8n=ËõráŒ!d××ãÉëHkvsšïr¿£¯¥Œ5^`r¡¯¨Þ<¿ys`åmáïV맕BY ßÕÛžyµ´çüc¶ùÏwËÞá¥êþÂnÃͲYG”-6䪬|e¹ç}ô!_`Üi)óÂC,®¬}BÖ‘’]Ý'Žg»7óÁo`q©­hÞ8¿ys`åmáïTÐÂW{Z飵ð]½í™Û­yÍ?f›ÿB\Kê‡\ðy]]9n!kç~Ñ»,Ë <¼ß0øæòܸzå˜åÐ[×#ããZm¾æRŸ= =û.Îw2®lb컨š÷ ŒW5´—Ö6ùýÍzéÕ.ŸyuÉçüc¶ùû‚Œë[¾…ÔÜ1$É¥nYVð´÷ŽÑ…xëöºÆ0þPt‡j;RÆÒ6qtyT}§ ò¹ì`-+ª7Ïßyc`åmáïTÐÂ7ΞV e-|5ÿræÕÒžóÙæ?{\àÛ-†êRüü9ì˜O·P³l˜-7PÜÑ,Ë <Âl×Ù¶¯œ³ìs¢·>DåK\Ý»òÓAÆZJ½ûaYÅ9q'Ž‰•· Œ6ÆŸ®6Yjsí¥W»|æÕ%ŸóÙæ?>‡]]”‡£›kóýŒ±»Z-Ë ž÷ÞM^3Â᪮]sjçÛ'õHôÙu¿M´C ßík-­ìbw![VqNlâœXy›ÀÀÀÀÀÀÀÀÀÀÀÀÀÀÀÀÀÀÀÀÀxaCzim“¥÷Ö^zµËg^]ò9ÿ˜mþCï¼Ëa¾vRß)/€,–eO{ï\‘ÝU¢Ü‚\©9uÙe®]ÈÝh«¿&J“'Üwûý±‹QY_ÙDìâœXy›ÀÀÀÀÀÀÀÀÀÀÀÀÀÀÀÀÀÀÀÀÀxYC{im“ßßdBzéÕ.ŸyuÉçüc¶ù½wrEßN™tE6KYYVð¼÷ήÉ/Avv‘+5·ù]ûüÓ(ï®l×DôÕ•ûNýcÞÅÃÁ²‰cbçÄÊÛÆËÚKk›,½·öÒ«]¶l]²Í?æsþcï]\•7^Ž1]cw)[–<ï½›ó²#J–­YäJ͵»fÏ{ë‘^\¯Û„æÝ6±ºœîÅê-«¨»(—ˆñê†öÒÚ&Kï­½ôj—ϼ®8ç³ÍþZÌqO³zÏþñµ˜_¹0&ÙpüÓ÷iŽuwyfriîð’£;’e¹ó§}{.Êïã«Î¦~åâåÑùù™é‘8þ°l;Z#ÿзgïJ¼[·¬âœ¸ÇÄÊÛÆ««—Ö6ùýÍzéÕ.[¶.Ùæó9ÿá9óœ\g¿³­ñp5Z–<ï½³Ëò}”GãåJÍ­¹Ø×ÝÍ#Q6Ù&¢oÏzï$'6q|µfYÅ9±‰sbåmã… é¥µM–Þ[{éÕ.[¶.Ùæó9ÿ±÷..•[èÕåù,|n®˲‚ç½wu%Í7w¦ù&Ε{tm퉮GbHòÊÏs"úär»³/q|‰M‹®–M›8'VÞ&000000000000000000000^ÖX½ôl“¥÷^½´¶Ë–­K¶ùÇ|ξ¿J>Ü!²D)Î?Ý&%wwcÁ‡‹scô#ºà-Ë <­ÍïšøÕ¥Y¸k–[÷Qo}©^žX¿&ruîö%†¾‹ÁåbÙÄ1±‰sbåmãU ë¥g›üþvöÒÚ.ŸyuÉçüc¶ù‹ëk维벂Cž+_YVð¼÷ÎÎËÃðEz{¹Rsò½»y$sWðs"Æh½wr©Þ‹å°¬âœ¸ÇÄÊÛÆ˳—^m²ôÞ³—¶vÙ²uÉ6ÿ˜Ïù½÷áºle’eLVÐä9ó•eÏ{ïꢿ…dg¹ræè½Ü‚®]Œ/qäm"vçï;u‹;ʽزeçÄ8&VÞ&000000000000000000000^ÖÐ^ZÛdé½µ—^íò™W—|Î?f›ÿØ{wdw¿ºù\©–eO{ïâ]’ÞKu/WÎ}¶]LìH:\mÛD ²Ën_kyÇ•,ë§1'6qN¬¼M`````````````````````¼°!½´¶Éïo2!½ôj—ϼºäsþ1Ûü‡Þ»Ì7pÆX]HsÙµbYVð¼÷NîhºƒÊ|VÝrwÞzo=’úz-¨åf;€?ôÞc-=ïbv1ZVQ'.Q'4ï¯kH/­m²ôÞÚK¯vùÌ«K>ç³ÍþZÌäYóPºìIòðZÌ/\kv©}þ>Ír¸Üo1÷õ,|i²À•åΟ÷íEÞC‘“+5Çñ¿>3=" JÛÄ‘$ïö%†´‹Õå`ÙD™¸DмO`````````````````````¼ª¡½´¶ÉÒ·k/½Úå3¯.ùœÌ6ÿñ9óæŠ>ûÝ»¬ ³eYÁóÞ»»^tç–yåÊò„ûzF~©eî‰~NŒ»ë÷{–_bŠ›XÇŸxË&ÊÄ)® ÍûÆ«ÚKk›üþf½ôj—ϼºäsþ1Ûü‡Þ{\ßêMÞ·â\Av9Z–<í½k’ÇËC+òD»\¹ç¹v="{»Ä}¢7Û_æ¡÷®Q¯¿k¿çÄX»åmãe í¥—ñf½ô2μºäsþ1Ûüçû«Œ B¸õäRúUÛ«èuò`yÿ‰ ÃqÈB?Ý—¥²9MhÝ5ýĪ+Á²|bÏ{úâR¾õCž¨— g e|>ëñt=S¶Gß5‡äJ¹£ <äúìmEõæù;¯7ËÛÂß­ …¯ö´R€ká»z[ËÖÒÚüc>ç?>øÝ$‡V×›wÍ[–†Ý‚Ëyèmme>.ëݲ¬ß?kÍ[t5ÞÆàªÛg Å»¸n\Ä8Ö_®‰Æ*ý}¾m`r©¬hžo»7VÞ&þN-|µ§•:Z ßÕÛZ¶–Öæó9ÿíínÿïÍ;ïýmÿßÿù_êÝpûoãÿþöäüí¿¿å\è·áô­Ö( }8¿â‘å±r=Ÿ]jÛù÷óÕu;?ãv>Íõç5îç“ì8~Ÿq?_åŸ×ù·óÙËVèçyûùYÐ_çgÜÏÏ7{^çgÜÏwùÍÃu~Æíü‘äo…ó¼Æý|qqû|4îç»Ëe;?ãv¾DW¶ÏOã~þpmûü4îç»óÛç§q;_£ìsž×¸ŸŸÿñÁu~Æý|“­ç¯ó3nçÇßÈuûü4îçÇŸnŸÆý|saû|4nç{piû|4îç³;¶ÏGã~¾ºº}>¯ó‡÷ò ";¿â~>Ë/©®ó3îç«K~;?ãv~ü?j¾>Ÿ÷óI6â¿ÎϸŸ/²Ðu~Æí|ôòßsœç5îç“ü²ê:?ã~¾È7¢ëüŒûùîJÚÎϸOQv:ÏkÜÏygÁu~Æý|ß¿?­¸Ïqÿþ´â~þØ¿?­¸Ÿo®nŸÆíü\ß>ûùñ{û|4îç›ì uŸq;/ï*Þ>ûùìêöùhÜÏW×·¿¿4nçkp~ûü4îç³üæ:?ã~¾º¼}~·óÍË8tž×¸ŸO®mŸŸÆý|u~ûü4nç»ß¿?­¸ŸOû÷§÷óeÿþ´â~¾ïߟV¼ÎåOÏ~jÜÏùMôu~Æý|— ®ó3nçCtÇõù¬¸Ÿ?\ÍÛù÷óÍõ´Ÿq;£¼‘ã<¯q?ȯ ¯ó3îç›;üv~Æí| òV’ó¼Æý|–ÿì:?ã~¾9¿}~·ó9ÈÏTçyûù¼Zq?_÷ïO+nç¿Zq?Ÿ÷ïO+îç«‹Ûç£q;?ΤíóѸŸOò†•ëüŒûùñÉöùhÜÏÏ—‚_çgÜÎ×ùò”ó¼Æýüܵí:?ã~~þ7x×ù·óû”çÏ—×ù?|ÿÉO­?þúöû?ÅõhÅüAuþÏïæÿê,¨·ñí~üå”Ç+BAž°øüÊñ·ÈXË~eúâ•u~sÚ¯Ì_»òð‡üX»_y|ñÊñmÉwW–/^9þR÷WÖ¯]Yüø—‹tweûâ•ãÇóñãõ~eÿâ•ãÓ¸_½×+ÿøý-ïŠcfül6~ö?„i”×Uý¯Z? $ù)Vãv>ŒŸþcºVÞ'Úü uMhÞ&âø6~B?'VÞ't]×ĵÎùÌÐú ¶›ÿÝö¡¤±êñOÈ6)ù ~—ï?Àÿøº0¾ºÓäï⯺püf× #ý¯¹Nþ¥vüD:ÿ’ýªëêüN)×eûë<þö–Ê6Zö?üøóíÝþå?øŽ‘Ößaññï°y@~´ÿèMiþ¬|ÝüÍ?ÿãooÿ対ýõÇ¿þûÿ½µð»Ö¿öÿ[ãÇÈñcÒø‡í¡·þï?ÿÓ¯Ý}õòdŸ#õîÿǸôßþõ_ÆíýÛÿùçÿ§_¶ã?õ/ôú7áú†ûVÆ]Ÿ?ÿýb1… C¿Øôc¶ñooÿðöÿq½[Jendstream endobj 6 0 obj << /ExtGState << >> /Font << /F1 7 0 R /F2 8 0 R /F3 9 0 R >> /ProcSet [ /PDF /Text ] >> endobj 7 0 obj << /BaseFont /ZapfDingbats /Name /F1 /Subtype /Type1 /Type /Font >> endobj 8 0 obj << /BaseFont /Helvetica /Encoding 10 0 R /Name /F2 /Subtype /Type1 /Type /Font >> endobj 9 0 obj << /BaseFont /Helvetica-Bold /Encoding 10 0 R /Name /F3 /Subtype /Type1 /Type /Font >> endobj 10 0 obj << /BaseEncoding /WinAnsiEncoding /Differences [ 45 /minus 96 /quoteleft 144 /dotlessi /grave /acute /circumflex /tilde /macron /breve /dotaccent /dieresis /.notdef /ring /cedilla /.notdef /hungarumlaut /ogonek /caron /space ] /Type /Encoding >> endobj xref 0 11 0000000000 65535 f 0000000015 00000 n 0000000064 00000 n 0000000206 00000 n 0000000291 00000 n 0000000371 00000 n 0000041342 00000 n 0000041446 00000 n 0000041529 00000 n 0000041626 00000 n 0000041728 00000 n trailer << /Info 2 0 R /Root 1 0 R /Size 11 /ID [<9ee5058f8357da803511a0a31cecbea3><9ee5058f8357da803511a0a31cecbea3>] >> startxref 41990 %%EOF gnm/vignettes/fig-LCall.pdf0000644000176000001440000012720412515147523015333 0ustar ripleyusers%PDF-1.1 %¿÷¢þ 1 0 obj << /Pages 3 0 R /Type /Catalog >> endobj 2 0 obj << /CreationDate (D:20061121121611) /Creator (R) /ModDate (D:20061121121611) /Producer (R 2.4.0) /Title (R Graphics Output) >> endobj 3 0 obj << /Count 1 /Kids [ 4 0 R ] /MediaBox [ 0 0 720 720 ] /Type /Pages >> endobj 4 0 obj << /Contents 5 0 R /Parent 3 0 R /Resources 6 0 R /Type /Page >> endobj 5 0 obj << /Length 43210 /Filter /FlateDecode >> stream xœìÝM$É•Þû}~ _’‹1ÚûËò  @š*@¸ÐÜÅ"":G˜úö׎3w Š©$q$ÿ»y<û¥EtuUöS^æÿúöŸ=b3ÖÑ[Óòák2Í>Ycãñ럎ÿrüË[ÿÿ=±ÿõþþÍŸŽÿ÷í¿þ?‡=þû›;þCÿß_Þœ¼~üÇ·ÔLóGŠÖÄt¼Ù›TÖøËñ­“e¬>‚äSÿ‹:Ù K5¥Èº1&[MÓuz!»dR½^/ÉÿàNÍJlך[ãÐÆë»Ö_Ÿãõ:ëYÏzÖ³žõ¬g=ëYÏzÖ³žõ¬g=ëYÏúϸ~¶Z¾¿Ír¶‰kœåá ßÆþv+)TcGÁ™L­k”/ý²àŒ&å#ö|ȲnŒÉ;Sç[²¦ºëõ\ŒkÚ6-ï×8´ñú¥×çx½ÎzÖ³žõ¬g=ëYÏzÖ³žõ¬g=ëYÏzÖ³þ“®—RëÃ÷·¬-äl×8Ëþ+üTpf“â‘bûAû—n²jŽò¥_œÅÔ|ÄLoiŒÉG‰- Ù6âõzv&×÷ԼݴjRXãÔúë—6^Ÿãõ:ëYÏzÖ³žõ¬g=ëYÏzÖ³žõ¬g=ëYÏúO¹^[H­ßßf 9ÛÄ5Îòp…oã {ûý÷·ßýÑîøþgù˯G2.ö°ú¤.õGvÙäv|?~óËoïé¯~ÿõ#뼕/ûW¯‹}›5Ü×=v±ø.oµ¦#…6ne=Š7É®Q>¨WulqÆ•#– _±¯cr= u«^È!èÔŸ¯÷¯Öâƒ;5k’{Ðj\ãÐÆë»Ö_Ÿãõ:ëYÏzÖ³žõ¬g=ëYÏzÖ³žõ¬g=ëYÏúϸ~¶Z¾¿Ír¶‰kœåá ßƾßo*›ÈzD¨ÜêZ’Üã:GùÒ/ Îhb=bŠz‹¬ŽÉ5“ç[r¨Rlž¯÷¯–~t¿i¿.ì©%ãâ‡6^¿´ñú¯×YÏzÖ³žõ¬g=ëYÏzÖ³žõ¬g=ëYÏzÖÒõÒBj}øþV´…œmâ‹–‡+|Wø©àÌ&–#…"uhÿÒÍ8¿FùÒ/ ÎbJ;äÔÓdÝ“³r<éØò¸ %­«×ëòF˃{j.lZ•jwŽSë¯_Úx}Ž×ë¬g=ëYÏzÖ³žõ¬g=ëYÏzÖ³žõ¬g=ë?åzm!µ>|›-äl×8Ëþ+üTp6Sê¸es<­©zýåK¿*8«3Î1»qSé¡c²Å}ä‘^÷ÝõzÿjõGÏ|¯ÖÄø •¼Æ¡×w­¿>ÇëuÖ³žõ¬g=ëYÏzÖ³žõ¬g=ëYÏzÖ³žõŸqýl!µ>|›-äl×8Ëþ+|/8k0VŸ”FÁ™L k”/ý²àŒò‡Øå=´ñ–Ƙl“ÔxKãBNIÿTþ|½¿QëÜs}¼´þuʇ6^¿´ñú¯×YÏzÖ³žõ¬g=ëYÏzÖ³žõ¬g=ëYÏzÖÒõÒBj}øþVµ…œmâgy¸Â·q…Ÿ Î,Åg AžÉÞ¿tÅ5Ê—~YpSô¹õe¼¥1&ëÖm× 9ùêçëµÊa£»{j6oZ5¡®qjýõK¯Ïñzõ¬g=ëYÏzÖ³žõ¬g=ëYÏzÖ³žõ¬gý§\¯-¤Ö‡ïo³…œmâ«–‡+|Wø©àl¦ÈcÓ£©ò¥›7!­Q¾ô«‚³9cÃxÙ˺1&äáõ²N/ääåQîçëýÞŠÓ©Yʃ–Û‡6^ßµþú¯×YÏzÖ³žõ¬g=ëYÏzÖ³žõ¬g=ëYÏzÖÆõ³…Ôúðým¶³M\ã,Wø6®ð½àlAnô”[6‹ÇëuÖ³žõ¬g=ëYÏzÖ³žõ¬g=ëYÏzÖ³žõŸt½´Z¾¿5m!g›¸ÆY®ðm\ᧂ³™Žä›¯üsWÜL¬ãYvvìÑ{ãše/»boMuGÌy­s²QP{×+ò˜¾Ø¶Dí_2<Ø—èÒ.:Ëš—Ø›8sÞ_˘]®6°ïo«Ëì9ÏþõÌßç•êŠûúÒÆ™UwLŒk–¼îŠ£q~4ßaôÜ:'×#n¾»q%;g|Ý%Ê—Øí%ö·ÅR׬âH<ˆ=1ç-ñÅ írµ•®X»ÜÙÈžóì_Ïü}^ùç®8gdzìÊ8ÃWéoç,;xÝ©£cIÆ“ŽuN¶ÿmžŸ¡Wr[¸5ËQ»}í¥ä]¬Æ¶5«8›8sÞ_Î.WX銵Ëì9ÏþõÌßç•‰îH1Ëá}Á›×,;xÙ‡þJ?'8¯4gBÞ©¾ºŸ9›ìÅ׬âH<ˆ=1ç-ñÅ írµ•ŽZ»ÜÙÈžóì_Ïü}^ùtÔ=ì¢éUG=ÆLü[ÊMÜþ#åv3¾©/²£ÜNÎ4·fùÈ^–ÛÉš$'™d9ÚZVŽ9ùxžz­W²MóŠ3‘š‰?<ºï¥…]t²Ý9ë?¾‘ØÄ‘˜ó–ÀÀÀÀÀÀÀÀÀÀÀÀÀÀÀÀÀÀÀÀør†t¹ÚÀ¾¿IBºÜÙÈžóì_Ïü}^ù”ÛÕ¤¾ä`éŸVÍÉË×R-§Ñߧd¼_³¼×Us0Íö#šÐÆÊ1'ŸÏ³>æ•6N¾¾ÙÉ&»}‰Ò®_b”Óµç¼Dwq$æ¼%00000000000000000000¾˜¡]®6°R5k—;Ùsžý뙿Ï+ÿt.t__ë‘bY·%Wi¹ç,;xÝãÝ«>ñ˜srUèÞÇ•ì›}%r›Çw{‰YZî±–5«8bOÌyK````````````````````|1C»\m`¥+Ö.w6²ç<û×3ŸWþ¹+nƵqs§~dgj\³ìàeWœ­Üz,ï&Ž½ë,­tœ'Šè•lƒ‰qKô7Yík/5í¢3®®Y?‘ØÄ‘˜ó–ÀÀÀÀÀÀÀÀÀÀÀÀÀÀÀÀÀÀÀÀør†t¹ÚÀ¾¿IBºÜÙÈžóì_Ïü}^ùŸa‘ª‰²…lŠÿùÙ›¤þ³£Ï™³¼×Usg Æe²rÌÉÖó‹z%÷­‡´%JÇ?<-¹¿îò.F9ÆcÎKì‰M‰9o Œ/fh—« ¬TÍÚåÎFöœgÿzæïóÊ?Ý–Ü×W}’^ÖÛ’«<Éoβƒ×]q1.ŒGÿe7VŽ9¹"Ç3ëÞÇ•ìÇÓ¯Dÿ’!<ØKÌÒÁ?ˆÕ®YÅ‘x{bÎ[ã‹Úåj+]±v¹³‘=çÙ¿žùû¼òÏ]q3N…çM{,Δ²fÙÁË®¸X¹w:ö ÔqG²ÎÉ6¹³zì]¯ä`åM^‰þþí»âfJÝEgœ[³~#±‰#1ç-ñå érµ}“„t¹³‘=çÙ¿žùû¼òO]qñ&Ê£ð¢iã’äð‰9Ë^wÅA^‰¹o Œ•c–¯TçÞõJQ ð+!{{´/Ѷ]Œ&º5/±'6q$æ¼%00000000000000000000¾˜¡]®6°Òk—;Ùs(ŽYó÷y埻âdJÝpÑûŠ« mͲƒ×]q16±f9±YVŽ9¹ ç$ëÞÇ•ì²IeKäúâÑx%›hÅâ׬âH<ˆ=1ç-ñÅ írµ•®X»ÜÙÈžóì_Ïü}^ù箸›ŽªIã¾âêLnk–¼ìŠ«5¡QN»=·ÎÒJ—ùX?½’C3¡n‰\îg!_{)nÜ=gý4FbGbÎ[ãË6­öýMÒåÎFöœgÿzæïóÊ?uÅÕ›z^»â4*ae¯»â`r=bìïÆ•cN6Jý=ö®W²rÐÆ•(ÕØò`_¢õ»Mˆk^bOlâHÌyK````````````````````|1C»\m`¥+Ö.w6²ç<û×3ŸWþ¹+NR*˳ìôÞçZå¡vs–¼îŠ³iíˆÅK½-+Çœœw0½+9–y×ôLÈÊ?¼¯¸¿Â.“Óš—Ø›8sÞ_ÌÐ.WX銵Ëì9ÏþõÌßç•iåHÑššeÍ™ì×,;xÙ7+·GǾW7îšÖy¿¯X¯d_Ml[Bn¦öÚK“›™Ė׬ŸÆH<ˆ=1ç-ñÅ írÏûŠg—{ÞW<çÙ¿žùûüò¾âÖçz¤f[Ý¢iaͲƒ×]q0Ù±:ãÆ=Ñ:'çÎ6ôJîWêž³”íƒ}í¥Å]ŒÆ—5«8›8sÞ_Î.×VºbíruþåœgÿzæïóÊ?wÅɤ6Î'öã¼âVåÿÌYvðº+Φù#kü8acÎÍʭѺ÷q%§db½©Ž¿íö%ö½mb1©®y‰=±‰#1ç-ñÅŒÑåÎVºâÑå®FvÍ«]ùû|査âjš=RòÆÉ9Þ:“Òše¯ºboÇÛ»¤C+ÇœluïóJýK¶-QìxÎÞñÜ7“ò£XÛšõÓ‰±'æ¼%00000000000000000000¾˜¡]®6°ïo«Ëì9ÏþõÌßç•¿wÅÞöÙ3'Šî1ÊAÅs–¼îŠƒI¡¿›&­´¬³ÜÁ\uïóJŽUÎø¹¯pöµ—Zv1o׬âHlâHÌyK````````````````````|9ûÕÀJW¬]îldÏyö¯gþ>¯ü·§®8™$ϲËrû±ì WÖ,;xÝgSå4fkj+Ç,_)ç¹÷qEN_g!¯„ôØg!_¢«»XL²k^bOlâHÌyK````````````````````|1C»\m`¥+Ö.w6²ç<û×3ŸWþÛSW\M•gÙÍóŠ}ƺfÙÁË®ØYãÒK6IWŽ9¹`òì¹õJAÞä•Èn5á÷®¸É#ýÄêÖ¬ŸÆH<ˆ=1ç-ñÅ írµ}[]îldÏyö¯gþ>¯ü·{Wì¼qú,;7î+vÑ”ºfÙÁë®8˜˜Ø¿’§WèœlÌûŠõŠ¼_¶D Roïöµ—jw1ç׬âHlâHÌyK````````````````````|9Cº\m`¥+Ö.w6²ç<û×3ŸWþÛSWœLLG ëœ Wmk–¼îŠ³ÜºSYMø˜“«1›p½’c^÷,k¢Ø»}‰Îíb11¬y‰=±‰#1ç-ñÅ írµ•®X»ÜÙÈžóì_Ïü}^ùoO]q•¾VÎ'v㜠ïL´k–¼ìŠ½5¶öwÓL¶c嘓kÆÎ3(ôJvÖIJ%²7ùÇ÷7ý£XâšõÓ‰±'æ¼%00000000000000000000¾˜¡]® ìûÛêruþåœgÿzæïóÊ»wÅÞË Ç)öËã Mqk–Ø—]q0¡õwMЖyÌIot{×+Ù·q~Æ™È}ÅÏ+î{)a£±iÍ*ŽÄ&ŽÄœ·Æ—3¤ËÕVºbírg#{γ=ó÷yå¿=uÅÉ„r¤hMvcÕX¿fÙÁë®8›bX« ºrÌIÀõîôJ뉺%ä¾jû`_¢ »XLÈk^bOlâHÌyK````````````````````|1C»\m`¥+Ö.w6²ç<û×3ŸWþÛÛï¿¿ýîîpÇ÷?Ë_~=’œ6a«ÿGÊçœÇ­ÍÎßßßüòÛãû_úËß½õÌø>¾|–Gá%ÓÆmÉÁIG=gy/«æ`•:’´å²rÌã!|ó0g½’£3Þm‰’äi}»½ªæfB|sY³~˜#ñ æ²æ-ñÅ írµ}[]îldÏyö¯gþ>¯ü·ûmÉ}}kGŠaw¢ÉqͲƒ×]qÆ<67oºÖ9yoܬÉõJ¶ÍÔ¶%äø`/Ñ›œÅV׬âH<ˆ=1ç-ñÅ írµ•®X»ÜÙÈžóì_Ïü}^ùoO]q2Áj7<\Å´´fÙÁë®8›ìX³±ºrÌÉGÓfÏ=¯4Y´%úÞ\~°¯½´¼‹EÎ\ž³Š#±‰#1ç-ñå érµ•®X»ÜÙÈžóì_Ïü}^ùŸa¢Iõˆ¹Éη#,>´0›æ~öE¨rž²<‚/Ž{®åVê¼fyç¯;êfZ8bË&æñ™9ùj꺟Y¯´ Õô•û¹Ýƒ}‰¾lb´&µ5/±'.QsÞ_ÌÐ.WØ÷·ÕåÎFöœgÿzæïóÊ £.r/µé}Ý_ÕQëÂbåÜç¿~aÒåoùŠYîÿi+Þ?©&O=”Ï:šTÖ,ŸõËV<ãcÿçæ'”è,½ÿ<ïy]iNÊð+‘êýÁ†Kô£ÏßÅf׬âH<ˆÍ®yK````````````````````|1Ã.WXiŵËì9ÏþõÌßç•ÿv?å#&ãÃv"t,¦–³I鈵É[•cN>ËóuïãJ¶Vžx%úÞÊO„î{…ü)9{Î*ŽÄ&ŽÄœ·Æ—3¤ËÕVºbírg#{γ=ó÷yå¿=uÅÕ¤8N×Ð=&'GGÏYvðº+n¦Ê&ý+éýácN¾»ºb½R³¼É+‘Šâ»}‰Þnb²&ù5/±'.QsÞ_ÌÐ.WØ÷·ÕåÎFöœgÿzæïóÊ»wÅ}}Õ®Ø3©S4±­Yvð²+NãƒrvHÑ•c–¯ç]Óz%Û["Ûuçñ­+NÞ$÷(Ö°fGâAì‰9o Œ/fh—« ¬tÅÚåÎFöœgÿzæïóÊÿüdf)Ÿû|2¥ýü¸ã4¥hå«È(¦Ú5Ëx]5g倎qß³¬s ΄y(õ¼Ò¼){¢ ÷ãª9™êw±ãšU‰M‰9o Œ/gH—« ¬TÍÚåÎFöœe¡VËš¿Ï+ÿíé¶ä*Þ“ãÓØAvróœe¯»â&e´ç^oº³–‘çñóJM&Ú-ÑÿV˃}‰Îob¶&¦5/ÑùMÔÄœ·Æ3´ËÕöýmu¹³‘=ç¤ý뙿Ï+ÿíÞ÷õE…×äfdÙA4ѯYvð²+ÎÁ8yÐ_˜õÓ9…±`ì}^©ãé|W¢ï­µ{‰ÞÄð(–¼fGâA,yÍ[ã‹Úåj+]±v¹³‘=çÙ¿žùû¼ò??¢/°}“%ïGX|ha þoY˜³ÉO_ñùÐŒœŒ­ãÆç6žå—‹Üƒ=gùÈ^—ÛYøX‹‰ã¸ 壉«ÜÖ+Í˱ W¢o0?ÚKLòÂM,Æ–5«8›8sÞ_Î.WX)·µËì9ÏþõÌßç•ÿ@¹ådçXŠÜ´ü󪹚 þKó^ïâŒk–7ðºjn¦¸ñu¼¾õ1'–ÔëJ+¦å-‘ýjþïUs36mb±&Ô5/±'.QsÞ_ÌÐ.WØ÷·ÕåÎFöœgÿzæïóÊ»Ÿ¹Ñ×;ºá0êðMHk–¼ìŠK¿àGqîÆ Í:'_åvè±÷y¥eéÓ¯D‡Rïöýh¹w1×5«8b®kÞ_ÌÐ.WX銵Ëì9ÏþõÌßç•ÿùmÉÅJ«;ð|æÆómÉ%ëFµìÜxÅä¼fy¯«æl‚HÒ¿\+Ç,I—y܈^É6ÏG!ÎDöÒ§ïö“œAý ¶¶fGâAì‰9o Œ/gX·X©šµËì9ÏþõÌßç•ÿöt[r5Aý׿â芫•;¢ç,;xÝ7i¯cÿ[Ò[ªÇ<:èæÞÇ•ñnÊ–ÈɸÇã1®½HË}Š}/Á®YÅ‘¸DMÌyK````````````````````|9Cº\m`ßß$ÑòÙÈžóì_Ïü}^ùo÷®¸:“åIzÅ´qctò¥ç,;xÙW/•wÌu>èPg¹ºçÆÞõJ¶Íd¿%r3þ‡÷÷×}ÛÅ ›žó}ÛÅ‘˜ó–ÀÀÀÀÀÀÀÀÀÀÀÀÀÀÀÀÀÀÀÀøb†v¹ÚÀJW¬]îldÏyö¯gþ>¯ü·§®8ÊfRê±q$s-&Õ5Ë^wÅYžžK6ÙŽ•cN~ºw½ÒÒ< cÎÙ÷ÃãŽkêAl~Í*ŽÄƒØüš·Æ3´ËÕVºbírg#{γ=ó÷yå~ܱ”ÏþÈÎË#õ~zÜq­r*²YaGÞ¬©mÍò^WÍͤrÄêMÑ[ªÇœ‚3aa¡W²:_‰L-ö«inû^|X³Š#q‰š˜ó–ÀÀÀÀÀÀÀÀÀÀÀÀÀÀÀÀÀÀÀÀør†t¹ÚÀ¾¿IBºÜÙÈžóì_Ïü}^ùTÍŤ¾É´ðóª¹9“Ò‘b”¿ÉˆÆÛ5ËxY57oj=¢,Ô9…`Ê<ßc^‘Ç ¶-‘½lp·/Ñ»] &Å5/±'6q$æ¼%00000000000000000000¾˜¡]®6°R5ÇÑåÎFöœgÿzæïóÊÿ¼j–î:Ùe9ÙøV5`¡éÖ~ £Ž¦êüÊ8Iº“Üšå¿î¨³q­'»¬³œñáæàóJ­¦ú-‘ÚÝ^b2É?Š5­YÅ‘x{bÎ[ã‹Úåj+µv¹³‘=çÙ¿žùû¼òßî§4·þ‚<Á/邵òæ,;xÝ7ÛûßšÞì=æ¢üM÷®Wjÿ›Ý}oáÇ]qÛáe/.¯YÅ‘8Å™˜ó–ÀÀÀÀÀÀÀÀÀÀÀÀÀÀÀÀÀÀÀÀør†t¹ÚÀ¾¿IBºÜÙÈžóì_Ïü}^ùÜÏ\LL‡Ütmã_w?ó\˜äØŸÝϬ3±ŽNZÞWçѸ°fyç¯:ê`½©ý‚ •:ûb‚Þ%¾®ÔþÚ–Hñ~tô%º¸‹ÁIJæ%Jâ5¡óžÀÀÀÀÀÀÀÀÀÀÀÀÀÀÀÀÀÀÀÀøZ†v¹ÚÀJG­]îldÏyö¯gþ>¯üO;j)½K>’œåQ>P5G9$E¯B ¶˜×,oàuÕœsýðÆꇣ³¯«¤^WZYŸÎ©˜\ì%&)ÉÄRÖ¬¢&vQ:ï Œ¯eh—« ¬TÍÚåÎFöœgÿzæïóÊ»ÕÅÁVãäÑAaØwàF;gÙÁ뮸™èû»ñëVnÃxÞŸî]¯”*ñ•HIn¸Û×^¤å>ž[׬¢&Nq&tÞ_Í.WØ÷7IH—;Ùsžý뙿Ï+ÿÛ’Ç ÌRWW÷×1ÊÑÕÿü~fçLt㨠¯ïgyç/;jçM ýStò|?Y©srbõøÌæ•þ)¶=‘²œ8½Û—hË.Úš—(‰KÔ„Î{ãkÚåj+µv¹³‘=çÙ¿žùû¼ò?ï¨5Yžðç·¨š£)òäÀ8Ÿ$èŠ+=gy¯«æ,çG'[d£²Rç¾Ñ¶ªf½"FØ)¬sœïUsíú.»f5±‹’ÐyO````````````````````|-C»\m`¥jÖ.w6²ç<û×3ŸWþéè WÕ®ØWÙ·ræÆœe¯»â& L6¯›½uÎÔ8÷®Wj49m‰Mx졯½ä¶‰}/Ö­YEMœâLè¼'00000000000000000000¾š!]®6°ïo’.w6²ç<û×3ŸWþ·%d U íŸÞ–ì áHÑÊíÔò‚iuÍò^VÍÞ›œäœqe¬Ô¹o4éóúÖ•ÒL‹["SO׸Dù—ä3˜ó%q‰šÐyO````````````````````|-C»\m`¥jÖ.w6²ç<û×3ŸWþÛÛ>þõÍö¯jý¯ÿð÷·âÖÿ¡ÿï/orn´=þã[vR3GoM·XûÑûÎYvz ÌÑEÙÞx=š”·×uÜ^7k_¯ëx½^Ƨv¾>ÇíõhbÝ^×q{½È£ ¯×u¼^¯NŽ³>_Ÿãöz”£D®×uÜ^/rÿ÷õºŽ×ëÍ®×總MˆÛë:n¯ùí‡ëu¯×í¯l5ï‰h|Þ:ï‰"?L¶„Î[¢ÿx¨ÛǼæ=倗-¡óžèªÝ:o‰þRÝ>ì5ú=¡óž(&†=¡ó–ý‡üö‘¯yOD9teKè¼'ŠÜj¿%tÞÑnÿ¦ôÄœ÷DÿwxÿÔç¼'Šü)-¡ó–HÖ¤ýSŸóžèêþ©ÏyO9"}Kè¼%²•ƒt®Äœ÷D0uÿÔç¼'Šq ·D±ò”Ò+1ç=ä'«-¡óž(ÆîŸúœ·Dµ&ìŸúœ÷D0yÿÔç¼'²U´%tÞ­«û§>ç=ÑÕýSŸóžÈògJ¶„ÎWÂ÷×ý–Xóž&…=¡óžÈò»‘[Bç-Ñÿ]wÛ§¾æ=LÌ{Bç=‘噵[Bç-!¿Ýº}êkÞA~#sKè¼'ÆïÙ:o‰0~S÷JÌyOù=×-¡óžÈ&ïŸúœ÷D“?_´%tÞòèÜýSŸóžÈ&íŸúœ÷D{ø¥sÍ[¢ÿ»¾ÿâ¹æ=‘~ù\óžh¿€®yKô×÷_B×¼'Æï²o ÷D“o\¶„Î[¢ÿ»n÷O}Î{"Ë÷<[Bç=ÑLÞ?õ9o‰ê~5]óžÈ¿š®yO´‡_M×¼%šøÕtÍ{"?üjºæ=ÑäaÍ[Bç+lW·O}Í{"ËcA¶„Î{b‚|zqÜbá­ÜcñÓu%™Öå­kI~ÒÚÖÿ¡uÎ%ùÎz_X?¶0$ù×q[ãǦ(ö¶0}ì£qý›ðþmÛ¾ðcŸëßý懯˜?öáøþã C²/ü؇ãûwvýWÿmaù؇ãû‡ãþ°~ìÃñýÃé¿®í ?öáøþ B{øŠícŽœ¯X~È5ýpþðý­ÿ’Ñ1÷[rËŽ¹ÿÇãü´÷×Ãø¯¯×åAðòŸ©óõ9^¯Op½¾ù㨹ýmË·½•0~LÄþËNÿÆVvÿwéñmÿ»ëäëì_±,÷Ÿú?¥ñcw}Vý¿ g$ÿào¿þéø/Ç¿ü;? …ùÉßÿ!yý·ti™û:Ýî?þæŸþñ·Ç?üéßþÇÿ_ÿôË¿ÿíÿôÏúØ¿Gý;Öþ]oП£û¿î+_¼{yø“<`ªÊ¯¾²ð?ýéŸ~ý·ÿù/ýíé>¶ão©æ§ùÓø[èïúú“_Öœûå÷Ø/ë‡Ïãx¦G©qD»_õ5‰àûõ_cæ?”Æ~ø~o³{z?ýÇIsG’Bgüïÿ§ç;gùþêž»èÇ?éØ|Ö4VêÜ¿zGݼ’û7m±]‰Ôˆ´Ýs'oM¾CÜÅê׬âH<ˆ=1ç-ñY í¥g›üþ6{éÕ.¯yuÉg}›ÏüýXÓèÇïñËý…9Œ$¹/mβƒ×½w”“Sc7Éʨ‡¸FùÍi}wãJN^n!º5¯ßUº÷ÞaÜ¿u‰iÜÓ¡³Š#±‰#1ç-ñ‰ 饵M–Þ[{éÙ.Ÿóì’Ïü}^ùçÞ;Ë}H¡?þÀƒ¯ò§%ç,;xÝ{—ñ§ÀR”?<)+Çœl2~½»q%§blÛÕûãÞ;?Ux‰U¨9«8›8sÞŸØ^ZÛdé½µ—žíò9Ï.ùÌßç•î½›œ —Â||Z ~œôÎǧ½ì½ÃxEZ|çÆÊ1'äм±w½’Ó8¸ôJTû✓þºœùy‰nœ¨óßÅ‘˜ó–ÀÀÀÀÀÀÀÀÀÀÀÀÀÀÀÀÀÀÀÀÀø¬Æ쥵M~[½ôl—ÏyvÉgþ>¯üSïÝ×çr¤0+&;Hò„9Ë^÷ÞAæÓ8Q]VŽ9Ù"oAßݸ’åÑQaKÈ#ÄÒƒ}‰ò$šKŒ&Õ5/±'6q$æ¼%000000000000000000000>­¡½´¶ÉÒ{k/=Ûåsž]ò™¿Ï+ÿÜ{'yDkŠ#/;¨&º5ËìËÞ»Ž“ ~¬srMÝû¸"§¶ä¼%J2­=ØKÌr̃(ÏŽ×YÅ‘xKYó–ÀÀÀÀÀÀÀÀÀÀÀÀÀÀÀÀÀÀÀÀÀø´†öÒÚ&Kï­½ôl—ÏyvÉgþ>¯üsïÝŒMãþî<î÷ŽÎd»fÙÁËÞ;Zy1zSÇÝì:'gŸ{×+9¥yÎËLÔøâ|ï¾—ÜŖ׬ŸÆH<ˆ=1ç-ñ‰ 饵M~[½ôl—ÏyvÉgþ>¯üSïÝ7ãã‘ú.ã¹–1šÒÖ,;xÝ{“Üsw¢sN6 0ö®Wrȳ÷Ÿ‰ÚW<ž~í¥–]¡ÏYÅ‘ØÄ‘˜ó–ÀÀÀÀÀÀÀÀÀÀÀÀÀÀÀÀÀÀÀÀÀøĆôÒÚ&Kï­½ôl—ÏyvÉgþ>¯üsïL GJNZmÙA1­®Yvðº÷î¤<¥³ÇÊ1'׿r™{W²¯¦Ô-Qêzw÷Þ;›w±˜׼ĞØÄ‘˜ó–ÀÀÀÀÀÀÀÀÀÀÀÀÀÀÀÀÀÀÀÀÀø´†öÒÚ&Kï­½ôl—ÏyvÉgþ>¯üsï]å—)ºyOzrÆ•5Ë^÷ÞÍÔvÈ£9½vöc–“Ä«{Wrv¦Ä-QÝ‹sNúë}o—˜¬IaÍKì‰KÔÄœ·Ƨ5´—Ö6ùýmõÒ³]>çÙ%Ÿùû¼òO½w__å<ï4ÎQé;ˆ&ä5Ë^öÞ©_§tºÙêë,w–·Ù{ëyJg«[¢ãø\ËäMŒbñkVq$Äž˜ó–ÀÀÀÀÀÀÀÀÀÀÀÀÀÀÀÀÀÀÀÀÀø´†öÒÚ&Kï­½ôl—ÏyvÉgþ>¯üsïå˜)ŽcÀeŤ´fÙÁëÞ;_ŽX¢Tõ²rÌÉe9Ä[÷>®äØù-QÜÝ^b29<ŠÍ­YÅ‘x{bÎ[ãÓÚKk›,½·öÒ³]>çÙ%Ÿùû¼òßÞ~ÿýíwt‡;¾ÿYþòë1ž¡i«ÿGø¾%éËÃñýýøÍ/¿=¾ÿ¥¿üý×[gþ‡ï¯ÆÕ#Éy-£¸ÏVîTŸ³¼×µy31÷e}8cN>¯Íç•lÓº]^}ÅÍ^b5Õobß‹³kVq$.QsÞŸØ^ZÛä÷7IH/=ÛåsÎÚ%Ÿùû¼ò¨Í‹‰rÇz31ý¼6ÏNŽOÑŽV¿¿`š_³¼—µyö&÷/W­É㙡:'Wåoã­ë•ì½ aKô/™X›÷×­ÛÅqËüœ—Ø›8sÞŸÖÐ^ZÛd©Íµ—žíò9Ï.ùÌßç•ÿymžûîܨͽÿ@mMÊGŠÑÄ2Þ@1έYÞÀëÚ<™XªIúáŒ9ù('±è[W²w&•-‘³qþÁ¾Dow1›X׼ĞØÄ‘˜ó–ÀÀÀÀÀÀÀÀÀÀÀÀÀÀÀÀÀÀÀÀÀø´†öÒÚ&Km®½ôl—Ï9k—|æïóÊ 6ršyŽÝm¨Í‹T×)z“ÆíòÅš`×,oàumÞŒ ‡|qœŠ®sêo=Ìgzê•ìÖÉí3ÑWØÇJ~‰Õ„ö(æ²fGâAì‰9o ŒOkh/­mòûÛê¥g»|ÎY»ä3ŸWþ陞}}‹ãöð6îw/ÁĶfÙÁËÞ»x¹=–lê¨üu–‚>Ì'–ê•ì³çÙ%Ÿùû¼ò?¿]¼ŽcÅSë; ÷ÛÅ?°Pîx÷ñç÷™7+gŤ˜¥Ÿï＿åÖ,ïüeßÞœ z[»>«Tgiö×±äz%»*7¾_‰r>«ôÖ··ñá<ˆ­­YÅ‘x{bÎ[ãÒKk›üþ¶zéÙ.Ÿóì’Ïü}^ù§ûÌ[0^¿éçóD[2Õ¯Yv`_öÞãVöñÆ¡ë:K³ïVï­Wú—HyKä>ÿð>ólv“quÍ*ŽÄ&ŽÄœ·Æ'6¤—Ö6Yzoí¥g»|γK>ó÷yåŸ{ïlbçˆëÁè­ëÖ,;°/{ïbŠÜEæÙ0söMn7×½ë•æW«¯s¶rˆÌn_¢m»XM(k^¢$.Q:ï ŒÏjh/­m²ôÞÚKÏvùœg—|æïóÊÿü|•–L–›Ò£å??&¥™œÆíáYjs9;ÆÛ5ËxU›'kMócA c¥ÎÁ®Â}]©í1‘qåÁ¾D_wÑÉïÌy‰’¸DMè¼'000000000000000000000>§±zim“ßßd±ôÒ³]>çÙ%Ÿùû¼ò÷ÛÅe}ÕÇoV?vLhk–¼î½£q®/Èò7Y©sð똔u¥¿›¶DJrJøn/1˜XÅþúœUÔÄ.JBç=ñY í¥µM–Þ[{éÙ.Ÿóì’Ïü}^ùŸÞ.. l×Lé¯yçZ¨G¼üõ ³‹&ÿm ƒÜ›þ³;Ûåc·~ãôÎö*'ÉÌY>ë× 1ÁŽ{ù³þSó8zfÝÙ®WZXwÍ뜣±?z¨ì¥¿×±Å5«8b‹kÞŸØ°~µÉÒðk/=Ûåsž]ò™¿Ï+ÿ|g{3^>ÏãÀ™#9gJY³ìàeïí¬‰íˆÕš0~ÿBgyÔ¨ÕûÖç•l½ü6Æ•ÈÉ´=TöRÓ.:ùm‰9ë§1›8sÞŸØ^ZÛä÷7IH/=Ûåsž]ò™¿Ï+ÿÔ{»þ¥ìè¹Ëèæ]4-­Yvðº÷&×#ö¯òX9æäƒlhì]¯ä¾…ì·Dβ¥Ý¾Dw1šà׼ĞØÄ‘˜ó–ÀÀÀÀÀÀÀÀÀÀÀÀÀÀÀÀÀÀÀÀÀø´†öÒÚ&Kï­½ôl—ÏyvÉgþ>¯üÏOtqÎä¾%_åñŸ“â’Irðy3¶Ž7ЗÅ5Ëx]›gSå䘺~S`ÌÉÛU¸Ï+ãÀ–¼%²7éñVôKôa‹InÍKì‰M‰9o ŒOkh/­m²ÔæÚKÏvùœg—|æïóÊ?ß.^M‘ãÀ´ï}ÞÉìs–¼ì½;ióK1vñ¢sòYÎB{×+Ù&9©åJ̇“Ͻw3Ñ?ŠÅ®Y?‘x{bÎ[ãÓÚKk›üþ¶zéÙ.Ÿóì’Ïü}^ùÜ.^åkgM*Õã8¯…îiáó}æ}ßM!÷v¼ó(EýœåÛ—}{,¶Œƒ]t–w¸]×+ò)z·%R}qŸ¹÷&»G±¶5«8bmkÞŸÖp£—Ö6Yúví¥g»|γK>ó÷yåÞ·÷ýkÇ¥=ÿkúv](çºäœëâû¢|¤×ï4SÜšå¿îÛ³‰±o°ÍƒÔu–/_æc@õJöÖÔ=‘½|ÉÝ^b2Õîb1¶¬YÅ‘ØÄ‘˜ó–ÀÀÀÀÀÀÀÀÀÀÀÀÀÀÀÀÀÀÀÀÀøĆôÒÚ&Kß®½ôl—ÏyvÉgþ>¯üúö(ç§çŒu¨Í« rë¸3uÔæÁšf×,oàumÞäwb*&é[³üŽÃ¼Á}^ÉÎÉ—¸¹ÉïQìö«imû^|^³Š#q‰š˜ó–ÀÀÀÀÀÀÀÀÀÀÀÀÀÀÀÀÀÀÀÀÀøÄFH«M~“D³g»|γK>ó÷y城CÎ$}ì§Gׄ(gÖÌYvð²÷^ªñXݬüu–ÓaÂ<ÒF¯dyéžÈIξÙíKtuƒ‰iÍKì‰M‰9o ŒOkh/­m²ôÞÚKÏvùœg—|æïóÊ?÷ÞQî]—Ç~Ö:vPŒ¯k–¼î½³ÜP‹‡½ë<žKZçÞÇ9 ½Ô-Q‚q?¼ß;$Ê£˜ãšU‰1Ç5o ŒOkÄÑKk›,½·öÒ³]>çÙ%Ÿùû¼òϽw1ÍãÀãØc´&–5Ë^÷ÞÍ8¹›=š:ŽT×9¹"Õ½î}\ÉýŠµ["Çû£>—XMÊb kVq$ÄÖ¼%000000000000000000000>­¡½´mòûÛê¥uþåœg—|æïóÊÿüxð¾ ÿ-Ç$ÕûONúõ”¼iã-Æ`r^³¼—µyô&Ô#f?Ñ9¹«ò×+Ù69BýJä²~ÛàV›Çqú&y´çœU‰M‰9o ŒOlH/­m²ÔæÚKÏvùœg—|æïóÊ?Ý.£ ö¿åqÃz̦¦5Ë^÷ÞÉ$¹¾šRÆÊ1'ÙûzwãJöUþv%r’-ìöµ—w1ïÖ¬âHlâHÌyK`````````````````````|bCzim“¥÷Ö^z¶Ëç<»ä3ŸWþ¹÷.rŠÔçi4ïÉ×,;xÝ{÷·‘˜ÏVÆœ\¸ö>®d[-[¢¯(?<¼¿îü.6íš—èü.ŽÄœ·Ƨ5´—Ö6ùýmõÒ³]>ç¬]ò™¿Ï+ÿócRb6¥o)óÏIé_>ËS4½ì¢¿¾Ì‡5ËxY›'gZ:bIóyœ:'—V)>¯d×ÿÖ¶D¶÷[Ñ/1¸]ô&µ5/±'6q$æ¼%000000000000000000000>­¡½´¶ÉR›k/=Ûåsž]ò™¿Ï+ÿÚ¼ÉÁàR›çv¯Í?²0X¹Ñüç}{0µ¬ß'wžMôk–wþºoï_.1÷Kã@så†õ©Ž+Ù[ü–Èý-ýðx–eËb©kVq$Äž˜ó–ÀÀÀÀÀÀÀÀÀÀÀÀÀÀÀÀÀÀÀÀÀø´†öÒÚ&Kß®½ôl—ÏyvÉgþ>¯üÓñ,©›9ÅÅ£ÓS3Ù­Yvðº÷®&„#oš6æcN®ššçÞÇ•ìƒ|É+‘­œI³Û×^r{[Y³Š#ñ ¶²æ-ñ‰ 饵M–Þ[{éÙ.Ÿóì’Ïü}^ùŸßgž²ñ}K1 ðÓÛÅû×ñqÔäyÜ(Ÿ½®6ùýMÒKÏvùœ³vÉgþ>¯üÓíâ9˜¨Çûq,y_ØÚše¯{ïqpz,Ñ8mÌÇ,_©Î£eôJöÑIJ%ú–’°/Ñ–]L&¤5/Ñ–]‰9o ŒOkh/­m²ôÞÚKÏvùœg—|æïóÊ?÷ÞY q¹¿»è›qeͲƒ×½w‘»Þc÷£ËÊ1'å®sÝû¸"-¾/[¢Œ²·/Ñç]¬&Å5/Ñç]‰9o ŒOkh/­m²ôÞÚKÏvùœg—|æïóÊ?÷ÞÍT9¼çÇåÅ›×,;xÙ{glëï¦ÿmÜ«®srœ½·^ɶÍC`f"Gy“»½DkbzKX³Š#ñ öÄœ·Æg5f/­mòûÛê¥g»|γK>ó÷yå~±s&úÇ—~x»xßKŽbókVq$Äž˜ó–ÀÀÀÀÀÀÀÀÀÀÀÀÀÀÀÀÀÀÀÀÀøĆÔÍÚ&Km®½ôl—ÏyvÉgþ>¯üÓíâ}3®InP×Þ»š×,;xÝ{9ËE~`¾»1'çLXïn\É!ÎÛåg"'S~øX̾—v±çÖ¬âHlâHÌyK`````````````````````|bCzim“¥÷Ö^z¶Ëç<»ä3ŸWþ¹÷nò‚<3ŽÞ»öÍ…5Ë^öÞÕš¬ùÌ〓K&Îw§W²¯¦æ-‘³œ‰¾Û—hý.:ãÛš—hý.ŽÄœ·Æg5f/­mòû›$¤—žíò9Ï.ùÌßç•꽫7©)”ÙÍ×dœ_³ìàuïä䕘­‰q¬sr^f}wãJŽÎļ%r5Á=Ø—èÝ.FÙôœ—Ø›8sÞŸÖÐ^ZÛdé½µ—žíò9Ï.ùÌßç•)òÌ §ËÆÂ9Ë^÷ÞÅX9Å¥˜¦+Çœü¨òuïãJ¶Ù„°%r2íǽw–›ÌÄ\Ö¬âH<ˆ=1ç-ñi í¥µM–Þ[{éÙ.Ÿóì’Ïü}^ùçÞ»ÊfR̦Žû½›3É®Yvð²÷nÖxÄTLw³ëœ\_ÙtïóJËó~÷9÷½Ùø`¯½4“Ú£XóšõÓ‰±'æ¼%000000000000000000000>­¡½´¶Éïo«—žíò9Ï.ùÌßç•ê½›7.Žû»õÉš-šÜÖ,;xÝ{Ý«—²rÌòÍùÔÊy%Ûh|ÝÙÊ¡æ»}í¥Ô]ŒÆ¦5«8›8sÞŸØ^ZÛdé½µ—žíò9Ï.ùÌßç•ÜZ.ϱÔ;Î[‘½ç,;xÝ{g“íó¨îe嘥aw«÷W²rxË•Èqórｓie‹ñqÍ*ŽÄ&ŽÄœ·Æ'6¤—Ö6Yzoí¥g»|γK>ó÷y忽ýþûÛïþèw|ÿ³üå×#—{Xý?R¤§|äPd“ßßßüòÛãû_úËß½uæøÞù*÷®§¤¯?²uò<Í9Ëx]›7Sä!ŸÑýpÆœl–'{ê[W² ‰œLúqmތ˗˜­51¬y‰=qŠ31ç-ñi í¥µM~[½ôl—ÏyvÉgþ>¯üývqY_Ü‘B4ÉDãóše¯zïl½iõˆ¥™PÆÊ1'[MÑÍ畼IuKä.ý豘òzˆ»ä´˜9/±'6q$æ¼%000000000000000000000>­¡½´¶ÉÒ{k/=Ûåsž]ò™¿Ï+ÿÜ{GÓì‘|ß“;(&¦5Ë^÷ÞÙ8¹>™¢+Çœœ]ǤÌ+Ùe“Ê–ÈÕxû`/1™ÅêÖ¬âH<ˆ=1ç-ñi í¥µM–Þ[{éÙ.Ÿóì’Ïü}^ùçÞ»ÛÆqà~œsÒƒ9®Yvðº÷n&ÈÝìUNv‘½9¹hJž{WÆ–H͸ö`_{éÌ%ö½X»fGâ51ç-ñ‰ 饵M~“„ôÒ³]>çÙ%Ÿùû¼òO½·sÆËyÞ^Î)—yžæœe/{oçM’‡|†ñÀÎcÎò•æùÞóŠÜÄ^Ý–È}~´¯½4·‹AÞÄœU‰M‰9o ŒOlH/­m²ôÞaôÒ³]>çÙ%Ÿùû¼òϽw41çXí拱~Ͳƒ×½÷ˆÄœç1uN}Cqvöz%»&Ä•è+âžk)¯;»‹Ù„ºæ%öÄ&ŽÄœ·Ƨ5´—Ö6Yzoí¥g»|ÎY»ä3ŸWþ¹÷.&§#…*ù¾owk–¼î½«Ü”.§•ÛñLN“»Nq™WZ÷ªŸ‰TMÍö%ú¶‹MŽ{™ó{bGbÎ[ãÓÚKk›üþ¶zéÙ.Ÿóì’Ïü}^ùoo¿ÿþö»?ºÃßÿ,ùõHÆ¥ÃVÿTïµÙSãñýýøÍ/¿=¾ÿ¥¿üý×[gþ‡ïãË÷X ÁÄqK»&´5ËxY›{oœ<äÓə߲r̲Ñ8ouŸWZ6Éo‰ìŒýá1)Þ™XÅþ)ÎYÅ‘x{bÎ[ãÓÚKk›,µ¹öÒ³]>çÙ%Ÿùû¼òO·‹ûhlÁÌuì éºfÙÁëÞ;™àŽñ›ãhsSÿ[›·ºÏ+­_·D*/nï{ÉåQìÔœU‰±'æ¼%000000000000000000000>±ÑóM–Þ[{i9çÙ%Ÿùû¼òϽw1^ŽïŒ“ÒL)k–¼î½«IrL[wŠ9I•¿Þ^i^V^‰¾7÷ãÞ{lz›qqÍ*ŽÄ&ŽÄœ·Æ'6¤—Ö6Yzoí¥g»|γK>ó÷yåŸzï`åFðšiãžôàåÿÌYvð²÷ÎävÄfוcN>˜<ïf×+ÙŽʯDj&¤ûmÚEo‚_ó{bGbÎ[ãÓÚKk›üþ& é¥g»|γK>ó÷yåŸ{ïþ‚=RÌ&Œæ=d9$|βƒ×½w4UNqIÊÊ1'o×5וšäð–+‘ýS¿Dw1™äּĞØÄ‘˜ó–ÀÀÀÀÀÀÀÀÀÀÀÀÀÀÀÀÀÀÀÀÀø´†öÒÚ&Kï­½ôl—ÏyvÉgþ>¯üsïMiã<ï0Î9 MþÏœe¯{ï*GˆÇMucå˜åD•2ú9¯T/·¤_‰”äÁ›»½Ä"7¥?ˆÅ®YÅ‘x{bÎ[ãÓÚKk›,½·öÒ³]>çÙ%Ÿùû¼òϽw3MžcéæY,Ñ›Ö,;xÙ{÷ˆÏG²ÁäÑêϹ¥:Ï&×+Ùö-Ä-‘ª©åÁ^¢5Ù?ŠýMÌYÅ‘xk[ó–ÀÀÀÀÀÀÀÀÀÀÀÀÀÀÀÀÀÀÀÀÀø¬Æ쥵M~[½ôl—ÏyvÉgþ>¯üSïƒqåHýoÖ$Süše¯{ï(·Ë©-MWŽ9ùjìlõõJvUnI¿9šôãÞ;ÈolbÿjuÍ*ŽÄ&ŽÄœ·Æ'6¤—Ö6Yzoí¥g»|γK>ó÷yåŸ{ïlB>’ðuì šæÖ,;xÝ{“C7^Í)+Çœ|–£ZtïãJ¶Íx·%²5ñ‡çœô×û—ØÄj|Yó{bGbÎ[ãÓÚKk›,½·öÒ³]>çÙ%Ÿùû¼òßÞ~ÿýíwt‡;¾ÿYþò둤c·‡Õÿ#EzjGvIN`ùþ~üæ—ßßÿÒ_þþëÇZ9›üo[˜ë}ácKÿ‡ïý 5“äÁ›zü‘“7ήY>²—E}²¦úqÌL¿É srÙ¸ùC¯È=ò©l‰\Lqö%º¶‹NžG:ç%öÄ&ŽÄœ·Æg5f/­mòûÛê¥g»|γK>ó÷yåŸnPï닯üsïL £çNã&úTM¬k–¼î½‹ñöˆ¥ïU[ý1'ïä$Ýû¸2NY/[¢ÚûC>—8¾èƒXãšU‰±'æ¼%000000000000000000000>­¡½´¶ÉÒ{k/=Ûåsž]ò™¿Ï+ÿÜ{7ãäÁ›Uò}ÙIA>gÙÁËÞ;[Z7msÌ9ùp¦®W䌛¶Dnë`š{ïÝdÞD'7ÏY?‘ØÄ‘˜ó–ÀÀÀÀÀÀÀÀÀÀÀÀÀÀÀÀÀÀÀÀÀøĆôÒÚ&¿¿IBzéÙ.Ÿóì’Ïü}^ù§Þ;{ܑ䀕q¾wŽ²pβƒ×½w0©±ö룳×Yù™ô1›óJ¶YNY¿ÅšúÃqö½´¸‹Ñx¿fGbGbÎ[ãÒKk›,½·öÒ³]>çÙ%Ÿùû¼òϽw’'v¦Ô¿ð8ß;W©Üç,;xÝ{gSʳ3Mó1'ïMŠsï㊭¡½´¶ÉÒ{k/=Ûåsž]ò™¿Ï+ÿÜ{W“ë‘R•ÛÂûŠ3>®Yvðº÷n¦å#¦2vvÌ9¹&GµèÞÇ•ìÆYãW¢ï-ÿð¹–ýõà7±ô ]ó{â51ç-ñi í¥µM~[½ôl—ÏyvÉgþ>¯üSï]ÆQâ)¶yOz‰&ú5Ë^öÞeKš§¸èœ|”ê~ì]¯dM [¢„»½D/ÔƒXÚšU‰±´5o ŒOkh/­m²ôÞÚKÏvùœ‹vÉgþ>¯üsïäÖr9Ï;ŒçZ–b²[³ìàuïM§tFƽê:ËåqžM®W²«cïg¢4Sâƒ}í¥ØG±Õ5«8bOÌyK`````````````````````|bCzim“¥÷Ö^z¶Ëç<»ä3ŸWþ¹÷®2ËýÝqÜq^­©vͲƒ×½w“r>¦þ´³óx‚æ|®¥^ÉÁš–¶DßRvöµ—Ú6±ïÅ•5«8—¨‰9o ŒOlH/­mòû›$¤—žíò9Ï.ùÌßç•꽫31)ºyÎI ¦µ5Ë^öÞÕ›"§¸49šEVŽ9¹²žZ9¯äMõ[¢y9’|·/ÑÖ] òs^bOlâHÌyK`````````````````````|ZC{im“¥÷Ö^z¶Ëç<»ä3ŸWþÛÛ>þõÍkí±ÿõþþVB»ã?ôÿýåM±iÿø½“æÛ:ï‰jZØ:o ïåIÛWbÎ{"™àö„Î{¢šd÷„Î["8“÷ïræ¼'¢©ûg:ç=Qåž-¡ó–ˆÎøý3óžˆ?Ó­yO”‡ŸéÖ¼%’}ø™nÍ{">üL·æ=QŒß?±9o‰Ü¿3Ú?±9ï‰þÑþ‰ÍyOôïŒöOlÎ{B¢~%tÞE7s&æ¼'ô`¾+¡óžÐ?Âx%tÞßõ^ß_‰ßÿÁw׿þóÛïþèw|ÿó¡ßP¿üÝø«^ˆò·r„þ/qÿýýýøkÞýöøþ—Ÿ/í?|ún–†.-ãg·‡¥ñcK“Mò÷ÃÒôÁ¥ýûí—æ.íÿàÓmiùØÒlû …Ç¥õƒKûBôoÿ–¶.íß)û‡¥ÞZ]ú‡ïo±ÿl*ÿ$Ôä{@sÿO§:ÿ«ê)Æ÷Õ:o‰$ÿR®Äœ·ÄTÏÄöUúÚól›ÿ»íM…þ lÖÿzMm¼‘¿KŸÀ¿¿0õÿ„*ºÐþ5ë²kòŸ<².­­ÿ7첦~ô·_ÿtü—ã_þÃügæïÿÌÆùn¾ôÿ,oã?`å+ÿãoþÛ?þöø‡?ýÛÿøïÿëŸ~ù·ã¿ýïãÿéŸ~ýØÙþ½_ÿN&ôÿ˜Iú#àÿ~Zúê#(V¾åK¶Ê÷S²ô?õ¥ÿö?ÿ¥¿CÝÊö‰üMÿU?tÍŸÏÞrÿ̯ïÓ~Y³œ–Úc¿œ?’nó™õÂ54®ùšDðýŸþMÒüó¡ÐßÎã]{÷·Ól6wøþ Lÿ%Y~}—ÿУü0yßøíû,ŒÇæØ¿WöóÏåë…þ^‹×ëýçÎÛY°S³òýÏ®õ_vç¨ßR×7M^×q{õ¬g=ëYÏzÖ³žõ¬g=ëYÏzÖ³žõ¬g=ëÿ_?[T­?ßßf‹:ÛÐ5Îòs…oã ?ý1ë`œô¼e´Iî3š£|é—í¸QÌË}'ãOë(7eÍéôBs&—ëõÜÖ0¿´A~ûùÒ’ü®ç‡¦¯Ÿš¾®ãö:ëYÏzÖ³žõ¬g=ëYÏzÖ³žõ¬g=ëYÏzÖÿŸX/-ªÖŸïoY[ÔÙ†®q–Ÿ+|Wø© ÍÒûŒ×£:åyÎQ¾ôë§?ÉÍæ^þØÎè”u Vþ¤Øز^hYþháùº>ÞjwÏ]´°iUþÐɇ¦¯Ÿš¾®ãö:ëYÏzÖ³žõ¬g=ëYÏzÖ³žõ¬g=ëYÏzÖÿŸX/-ªÖŸrhãhQgºÆY~®ðm\ᧂ¶ýíÝMÇ–Þñ}ŠZJ‹Š÷—¥H ð˜ f3ð\]XPÏØÌ÷wœ8™QY¬!eÌ𿸟Š¿Îª+‘­‡©H«6²r”Gñr’ÇŒò¥Ÿ´e,xçå/}ŸF9z9 e·—Xål–;1—•Uwb.+o/kŒ^z¶Éo³—^íòÊ«K^ó×|Ì?ôÞ}“#²‹±rö¸sÁĶ²\ÁÓÞÛy©Ò] ÆëNÍÍ«ù|Å'k¢='|°«÷¿ôÞΙTïÅšVVqL܉5­¼M`````````````````````¼¬1zéÙ&Kï=zéÕ.¯¼ºä5ÍÇücï ã~ð¦÷{g“ëÊrÏ{ïd¢Ów£;5·íÝW¤Å÷áœð®ÉÑ+»}^K)»˜+«8&6qL̼M`````````````````````¼°áÂj“¥÷½ôj—W^]òš¿æcþ±÷.r”¸kIŽ —+h¦–•å ž÷ÞÕd{s¹_r;5—²vÎW¼o㉜kÂÛdÒ—ï÷î[ó.6ùfVqLl☘y›ÀÀÀÀÀÀÀÀÀÀÀÀÀÀÀÀÀÀÀÀÀxaCzim“¥÷Ö^z¶Ë+¯.yÍ_ó1ÿÐ{{k’ÞµFóîƒ\ÜÌrO{oïLi7‹ë²SsŽÇéåúŠ÷Q:øs¢VÓ¾x¿w_wi½‰~å%º´‹cbæmãe í¥µMþx[½ôl—<»äcþš×ücïL‘óÂíxnf¿‚,ë3Ë<ï½£| —’ÉqìÔ\âjÌç+Þ6㶠o)îÎ>Åw±ÿÄ­¼Ä>±‰cbæmãe í¥µM–Þ[{éÙ.¯¼ºä5ÍÇücïMm}ÝÎ'kúfb\Y®àyï]“ÞÊë²sæbÚê½õ•Væs=g®ÕÄû³Ã—Xä"ïÄjWVQ'vQ&4ï¯jh/­m²ôÞÚKÏvùȳK>æ¯yÍzûùýí§_ÝÍÝÞ“>ß’ÜrnoV"œ¿y7®ìýãöÃ?ÞÞïËïŸ/ù/ïo.ô÷Ro.5ãFm¼ÜW>³¼§µyp&È—³r¡²SslròËxëúŠTøÁoÅgïì%öìïÅÖVVqL܉}bæmã… é¥µMþx[½ôl—<»äcþš×üÃíâ!/‡¬s†dª_Y®À>í½£IQ/YwjŽXïN_Éuþ±ÁÌ©Ê“æ¯yÍývñàMê—d³ñª†öÒa´ÉÒ·k/­ù#Ï.ù˜¿æ5ÿxŸy39Ý\*ÆkŒ}Ý®,Wð´÷ŽÖ4?ÞM'³ÌœÆíãÚç+µÈéçD.rRúnŸ¢¯»èLÊ+/Q&NQ'4ï/j¬^z´ÉoG/­íò‘g—|Ì_óšÿ†ûÌ«©ý’Z6ínï_¾Ê*Ñøñ<јLh+Ëx^›G9¹\®Ov™Yv™(0_)}¢m¹ÊUìöƒ‰å^,ieubeBó>ñª†öÒÚ&Km®½ôl—<»äcþš×üÃíâ1ëorßzÄÄjR]Y®àyï]L°7'ýGªÏì‹< T¯]_)Á¸°M¤bÊ}§~^KÎ÷b‹+«¨»(š÷ Œ×5¤—Ö6Yzoí¥g»|äÙ%ó×¼æ{ïf¼Þ_טœ)ee¹‚§½w²&¶›³^ž(úq[9ØñàO¹öùJõrÌ9‘ü:@üÚ{79;}| 3맡§¨š÷ Œ×5¤—Ö6ùãM&¤—žíò‘g—|Ì_óšè½û`”»´«o²¯û°‹EZñ™—(§¨š÷ ŒW5f/=Údé½g/íV¯­yvÉÇü5¯ùÇÞ»š"w;'yle¿‚ìL+Ë<í½³•j\žžêØyäu€È|¥z¹–s"y¾x€Hj&ú{±Ø•õÓЉ]” ÍûÆ«ÚKû6™½´æ?Ž<»äcþš×ü§·Ÿßß~úÕÝÜíý7ùáó-—nöfõ'²Áú[ òãûÇí‡?~¼½ÿÞWß?_*ó_ÞÇWorw™›ÌQΟY®ß>m̓`Íɺlѹf¬Vâë•þ 9¿MÄþ~½Àþ™¹;°¶Õëw^m+oß© …¯ö´RGká;{Û#Ï–ö˜¿æ5ÿpvNÆå› EN)‘+(¦¸•å žÊÙÄxs6˜8žA9³«Ò¨ëµë+±oªÛD &¹QN¦Ú],Æ–•UÔ‰SÔ ÍûÆëÒKk›,½·öÒ³]>òì’ùk^ó½w5!§ýhÞ‹•;¨g–+xÞ{7ù͙ǥèl•ž]/]_éo&åm"ô¿´;ú¼”ÖN°_‰Ï3ª7ÖOOfÞ&¾[A _íi?ÞdB ßÙÛy¶´Çü5¯ù‡B¹8“bŸwr³\A4¶­,Wð´P.ÞT“'D¶Q†ÏìÚ:xc½«Éi›ˆ~½ÿK£Ü×]ÝÅ Ÿœy‰2qŠ:¡yŸÀÀÀÀÀÀÀÀÀÀÀÀÀÀÀÀÀÀÀÀÀxUC{im“¥÷Ö^z¶ËGž]ò1Íkþ±÷Ž²à|0E›÷b|]Y®àyïåLy7^wjvÞ”y'õ|%v2n!=¹“º$ʽ˜ãÊ*êÄ.Ê„æ}ãU í¥µM–Þ»Œ^z¶ËGž]ò1Íkþ±÷.¦É“½ ãàìjM,+Ë<ï½›q­¿›²Z}ÍNŸ 9®]_‰Ö¤°M„`\¹³—Xå6÷;±†•UÔ‰]” ÍûÆ«eôÒÚ&¼­^z¶ËGž]ò1ÍkþÓÛÏïo?ýênîöþ›üðù–ŒK7{³úÙÐÿ’«îýþqûáoï¿÷Õ÷Ï—Êü—÷þÕqrw’¯Ö¯¿“óÊrýO[óÚ/¨ÞJ6eþ[=ÎHÑ\ŒòÎWÙ°ËËOá<½þõüŒêõÍ3oß­àܬi¥ŒuïlmW\íš¾æ5þp vr¯µÔãI¯/›šV–/ÿ¼LN&•[kÆk =¢³ÎÄyù|Åe©ÛÏ çMüâÉ#ýRZÜÀl¼›Q½±¾yc`æmá»Fá;zZ©“g᫽í‘gK{Ì_óš,”‹ :çæÝÕÍW–+x^(W9îºTÓôÒGìïí¸òñ‚œ.RÒ9 §jÇûÓBNÏùÍs—çüî™· „ïT}¯Ö´o³î­íÛÑþŽm¼íð}^ãerߞ먯ƒ—/ß‚ñaeùòOËäæLK·þ£ÞU­qÜÖ<χž¯„j|Û&|Ïl¼=´É}=¸ ô&µ—Üî™· „ïTÐÂW{Z©“µð½í‘gK{Ì_óš,”ƒ©r…4Åã ²‰~e¹‚ç…r2.ÞJ›çdh”Û ó,Âç+Á›\¶ ßßþ— åh’½KQ½±~畺ò6€€€€€€€€€€€€€€€€€€€ð ZøjO+…²¾³·=òliùk^ó…r1V¡°r³\A3Ù­,Wð¼P®&„[­ãŽÛŒ}{óóÊÇ .Ž 8œoÆ–;ù¼’Üî¼VfTo¬ßy}`æmá»lž5­ÔÉ£î­íŠ«£]Ó׼Ưe²·Öx9-£êÙÉÞzSíÊò埕ÉÞ:“ü­Z)³eãˆÎ“êظ^éïÄåmÂS¾Ô&Ë¥Ôº^öiTo¬oÞ˜y›@@@@@@@@@@@@@@@@@@@@øn)|µ§ýx“ )|go{äÙÒó×¼æ?=ÊÁÄ0Ž·hZ(÷Ÿ´•å žÊÑw«Í476Žèl56ÍK×WB–¢Ï ßLúr¡¥?ÁdBšqy}}óÆÀÌÛÂw*há«=­ÊZøÎÞöȳ¥=æ¯yÍz(”³Ércq6¥+hÆ••å žÊÅ4{«ÎTíÂ5Öõ ¿ù‚‹Õ{8WÆ“o}r1>o^•CA4.ÏçÝ3oß©0ú^­i¥NuïlmW\íš¾æ5þé¡Ln¦ºq7r}·ó&ä•åË?-“3¶ÝZ–eãˆÎvw1_ ãìsÂ[yáN/К˜îÀfTo¬ßy}`æmáûfá«=íÇÛ*|go{äÙÒó×¼æ?½ýüþöÓ¯ý•ÛûoòÃç[2.ÝìÍêOdƒM·¤€~ÿ¸ýðÇ·÷ßûêûçoÙçb‘Û¢/ïKì_ÞûezÚ¹lBo<™”V–7þ¼ÇŽòNúgXÇñånê0úùJpÆÖmÂg9d§LŽw`ó3ª7Öï¼>0ó6€€€€€€€€€€€€€€€€€€€ðÝ RøjO+=¶¾³·=òliùk^ó_ï±ç<§fjü†>:'·C7“õ TSâÊòž÷ÑÅÄr+ë|­?éÐ\¨ÒÃRÐÇt'/o|ª§Ws3ª7Ö7o ̼M |·‚k³¦•6zÔ½³µ]qu´kúš×ø§ë!®™PoÎ9ãG]î<;pfùòOËäþòx¢wEklÑäùA}ÁÅ(uû1 7ô÷gIŸžõ›çdNãòúúæ™· „ïSоWkÚ7haµ¶+®Žöh‡/yº–ÉÞËÒ=§q"µORsÏ,_þy™LM·ÖŒ/cãˆru.óé+ÁË ÓçDÿ %ÞÑ'èÝFëŒËón÷ÆÀÌÛÂw*há«=­ÔÉZøÎÞöȳ¥=æ¯yÍý¤ ïä¶èÒ·µ¯”!_&Ë­ÏMom®rδF¹úçet16JËí´Å֘׾ù‚|!œÎ5)ßwyyY˜¸{ýÝjTo¬ßy¹¬¼M |§‚¶½£¦•*ZÛ^mmW\íš¾æ5þéáÎæjš<óÏÎgg’]Y¾üÓ29X9y£¸ùXErõ|¬¢¾ Q´ñ{¦½»“ו4é¾w¯æõcëw^˜y›@@@@@@@@@@@@@@@@@@@@øN…Ñ÷jMûñ6ëÞÙÚ®¸:Ú5}ÍküÓµLÞ8¹_¸H%-_>šÜV–/ÿ¼L&Ž;¶“nQkó4éùJ”=ÛD¿ üÅ'øõK)u£±iFõÆúæ™· „ïV…ïèi¥Nž…¯ö¶Gž-í1ÍkþÓC¡œä.h缩zÅÔº²\ÁóB9›lo5š2žN¨ÑYk켫z¾"C§mÂUÓÊ}^J_8Á"gSkTo¬oÞ˜y›@@@@@@@@@@@@@@@@@@@@øn)|µ§•BY ßÙÛy¶´Çü5¯ù¯Ÿ»¢<ÂÏÅjÜ7¼ªI~ÜЬ§yDglYYÞÀó>º™Òn¥ÊÃýäèl5ažI=_IVŽë8'‚•;²wúû×<ÀhM 3.¯¯Ÿž̼M |§‚¾ÚÓ~¼­Âwö¶Gž-í1ÍkþÓõ羿È1I–+ˆÆç•å ì³B9zÓ¤&Ž³@f–³6fÛ>_餫ۄ\ÒýÑΧâ.“ýÊK”‰SÔ ÍûÆ«ÚKk›,½·öÒ³]>òì’ùk^óŸzïhšœ¾1þ"WPLL+Ë<ï½³qÒÀ;SËØ©ÙãVï­¯ToÊ>Ñß´ wö“Iá^¬neubeBó>ñª†öÒÚ&Kï­½ôl—<»äcþš×ü§‡Þ»Ê³]ð³›OÖ串\ÁóÞ»™  ¼Õk×ìíÙ{ë+yœ©}NÄpíÔÏk)~ûµX»²Š:qˆsBó>ñº†ôÒÚ&¼É„ôÒ³]>òì’ùk^ó_?w;Ž;ë3Ñ}ýØí䌗SVš‰ãœ–L +Ëõ?mÍ“7)?Èã—3gÂ|Èå|¥_‰ÍÛDÿË—ïï×ÒÜ.99|fuâuBó>ñºÆè¥G›,­ù쥵]>òì’ùk^óŸ®w‹§(·“¹Ç))©ëW–+xÞ{'Sâ­9ùD6Ž(Å ó€—ùJ«RÀŸéáI–'èìf‘4.¯¯oÞ˜y›@@@@@@@@@@@@@@@@@@@@øN-|µ§•BY ßÙÛy¶´Çü5¯ùO…r1YŽ/IòTÊ~ÙïV–+x^(WÓÂM6IãYš3‡,Y¯]_É£5?'’¾¼=6ÊÕø¶‹Í¤²òeâuBó>ñª†öÒÚ&¼­^z¶ËGž]ò1ÍkþëǤlj¿¤ÒŒ÷_?¤ùGMîÇÈÁ„¶²¼§µ¹±"g³äy&ù̾­Û¬×+ýÃðq›H‰wö‰õ^,yeubeBó>ñª†öÒÚ&Km®½ôl—<»äcþš×ü×kó¾ÁÖ[rg÷×[óh¬Ü®W×7g“êÊrýÏ[ó$¢8ïåÐqÙ©9f“Wk®¯ä*—uNÈ—ürkM.÷bK+«¨»(š÷ Œ×5¤—Ö6YZsí¥g»|äWK®ó׼濡5ã˜oMø–Ú¼È=érÈw«ã 4SÊÊòž×æÕ$Û?ŒþŽõÃÑ,§»¬G_)Eû>'ú;*÷ç¿,±˜šw±ÉCgVQ'NQ'4ï¯kŒ^z´ÉR›Ï^ZÛå#Ï.ù˜¿æ5ÿéz6y±r º+nœ}~óÅ›–W–+xÚ{÷‘,'ÐTSGå?s¬Çƒ:õo£‰ûDsÒäïö)Ú´‹Þ¿òûÄ&Ž‰™· Œ—5´—Ö6ùãM&¤—žíò‘g—|Ì_óšÿôÐ{“õl;ŽI)YnŸY®àyïM•×¯.;5§(·›ëµW¼ rçù9Q“É÷—Ÿ¢»˜Lr+/±Ol☘y›ÀÀÀÀÀÀÀÀÀÀÀÀÀÀÀÀÀÀÀÀÀxYC{im“¥÷Ö^z¶ËGž]ò1ÍkþÓCïM‘{°ûWwœ—&?™Y®àyï]-CŽúî4Ë[X×>^ñ¶“öœðÖJý½ÛK,&†{±Ø•UwbŸ˜y›ÀÀÀÀÀÀÀÀÀÀÀÀÀÀÀÀÀÀÀÀÀxYC{im“¥÷n£—žíòÊ«K^ó×|Ìzè½ÇAÞ®5ãÇT/ƒ3Ë<í½«3^Ïh±ã^õ™k3q^»¾âC3ÙŸÞµ'ŬVïÄÚVVqL܉}bæmãUÙKk›üñ¶zéÙ.¯¼ºä5ÍÇü§kï]ƒqrvšî¬IŸY®àyïMŒãD•ª;G–Æ=γÉç+¾™Z¶ ïŒkwöy-ÕíbÿjueÇÄ&Ž‰™· Œ6\Ym²ôÞ³—ÖvùȳK>æ¯yÍzûùýí§_ûUÝÞ“>ß’”boV"Ez´7Ÿ¬¸ï·þøñöþ{_~ÿ|éÌyï|6AnéΦiµ_Ms+Ëx^›“õVw?¨9³k&Í›áç+É™˜·‰èówö)Z»‹UÎv™y‰2qŠ:¡yŸÀÀÀÀÀÀÀÀÀÀÀÀÀÀÀÀÀÀÀÀÀxUC{im“¥6×^z¶ËGž]ò1ÍkþÓÃíâ}!Ý|ˆ¦Œâ¾õve¹‚§½w³¦ú›ëâ8Ú\³<ÒÓÎʾ"'”ûm"¶u€ø¥÷îë®íâøf^bŸØÄ11ó6ñªÆ쥵Mþx[½ôl—<»äcþš×ü§kïÝ÷9º$ÎÊ[’gvÎ,Wð¼÷¹<ªzÙ9²·y=Ps½’ú%Æm"úÕû_{ï`B¹sZYÅ1q'ö‰™· Œ—5´—Ö6Yzoí¥g»|äÙ%ó×¼æ?=ôÞÉ´póÁ«Í{5±®,Wð¼÷.òäL¹{½èNÍrgù|,æ|¥¬‡~Îœô‘·ÇÞ;›”ïÅWVQ'vQ&ÊùQ Œ7´—Ö6Yzoí¥g»|äÙ%ó×¼æ?=ôÞMDñ>/Í{°Îä²²\Á³Þ;Xk‚ÜÍnõnö™åÎòèÆÎõJlr›÷9ãѯ½w“¼‰ÎØ°²~cbÇÄÌÛÆ ί6ùãM&¤—žíò‘g—|Ì_óšÿt齃õ&Èé$Ö„4® ššW–+xÞ{“êÍ¥£1×\ûµëÎùŠIžÌyLxçMúRï-×Òâ.FãýÊ*Ž‰M3o/l·Údé½G/½Úå•W—¼æ¯ù˜콓‰­Ï[ãó¸‚*ƒ3Ë<ï½³<úÒÕ G³ÈΑ½urt‹^»¾Ça.ç„«&ÝŸ¡rŠ.ìb‘²~æ%º°‹cbæmãe í¥µM–Þ[{éÙ.yvÉÇü5¯ùOo?¿¿ýôk¿ŠÛûoòÃç[2.ÝìÍêO¤H/ý¢‹3±ÜÞ?n?üñãíý÷¾üþùÒ™ÿò>¾|–[¸‹¬ô7àœtú3Ëx^›7Ó¤ÀϦùñÖG–Í£Ÿo]_ Ö4·Mx/íùnŸbð›è¬Évå%¿‰:1ó6ñ²†öÒÚ&¼­^z¶ËGž]ò1Íkþávñ¾¿ÊMÞuå⢠Î,Wð´÷vAjvW’>PsåL™×®¯øXäžøcÂû Mþn/Ñ›äîÅÒVVqL܉¥­¼M`````````````````````¼¬1zéÙ&Kï=zéÕ.¯¼ºä5ÍÇücïÝ/FîÁvú`ÎàŠÉne¹‚ç½w6!v¹™PÇΑ}¿ ¹s½«œy~N„$;vû¼–bïÅVWVqL܉}bæmã… é¥µM–Þ[{éÙ.yvÉÇü5¯ùÇÞ»Jöý ÛqP‹·¦Ú•å ž÷ÞÍ$9O¥™<Ÿ¹“æ/úŠÙävNxŸåKìöy-µmb¿WVVqLœ¢N̼M`````````````````````¼°áÓj“?Þd¢Ú£]^yuÉkþšù‡ÞÛ;ãè¹ë¸'ÝÓÚÊrO{o9JÅ÷wM'ƒk–RÜ<çd¾£É~›8ßÝ¥÷îë¶îb0!¯¼D[wqL̼M`````````````````````¼¬¡½´¶ÉÒ{k/=Ûå#Ï.ù˜¿æ5ÿéíïnÿ÷Íkímÿñ¿ÿçK ínÿ¥ÿï÷7'ë·ÿú–qí–­<™R.tTØå:ï—5¥(7¯Õ(wŸ«#m«£“?VG:W˸®µªi[ r"̱:Ò¶Zä©£ÇêHçjµr¸úZÕ´­9‚æXi[Ír£þ±:Ò¶Úä3ŽÕ‘ÎÕäÿôµªi[ÍÆŸŸ†¦mµ™˜ÏÕ‘ÎUg½â¾Þä?48×GÜÖ7þüÈfÜד<Öõ\q_¯r~ý¹>â¶Þÿn-ç7ã¾ÞZ¶õ÷õþ·yÞÖGÜÖƒ3aû|4îëýÍíóѸ¯yfì¹>ⶭݬkÜ×£ücx®¸¯ùÏ2Îõ·õdåø¢c]ã¾ä4ÿs}Ä}=Ë©GçúˆÛz_±Ûç§q_ò§fçúˆûz–_mÎõ÷õfòöùiÜÖ‹—?n;Ö5îëÙØíóÓ¸¯·ý—¡·õê÷_ˆfÜ×ÓþKÑŒûz5eû|4nëÍ™¶}>÷õdÜöùhÜ׫ Ûç£ñ\—§§óó™q_rœÕ¹>â¾^L³ÛúˆÛºsÆžŸßŒûz4¾në#îëE¦|®¸­{+ä{¬kÜ׃©i[q_/ÆÆm}Äm=Øýק÷õ°ÿú4㾞÷_ŸfÜ×ÛþëÓŒÛzôòÓc]㾞åϵÏõ÷õ&¿ÿŸë#nëÉË·ǺÆ}=™²}>÷õjZØÖGÜÖ³7nûü4îëÉ„íóÓ¸¯W“¶ÏOã¶^œ|[r¬kÜ×£©Ûç§q_¯òŸ¡ë#nëÕ¿}~÷õ¸ÿú4ã¾^ö_ŸfÜÖ›Ý}šq_û¯O3îëÅøíóÑx®Û¿ Ù¾ËÒ¸¯÷oCê¶>⾞å±.çúˆûúx²ø¹>â¶îÆXŽuûú8©î\q_ÿEß¹>ⶾÛx|y®ÿüþ…oM?ÿíí§_ý|Žûøntüð—ñ£¾Ðìôûÿ›ò$w×¼“‡¹ucÿ¤_ɾ1|ÛFçÆ/MûÎø;c’oa÷éwöo`s¼Û™¿m§ïÿG§ûåwF'úÙwÖoÜÙ¿ïßPï;Û·í ýÛR¿ïôÖêÎ_Þßú/‰}D~a’ÿ„V“ü}\ôߨ.Ëa|;;Ò¹*÷ÅcUÓ¹ê‚ׇ-éòŒÛzÿEÍŸøŒçºïÿPæt¬Ï¸­÷÷eOÆm½DýσçºÆs}¾Ùµ¾½÷þOÏñÑnç_¶¹C+¿áöoËêødÿ’ïÿù7·õïfªþò—ø'¶¹þ»´wºÏÿ™}ýw—þ»‡ì³b›<ʪS;þ¾ù3Ûúïõ>ŒmæÍßÂô"ÿÌG)¿27½ÈºþÎîÿb~Ûæòþ—ϽýÛ?ý¿>†ùO”¿þåõïé(eãßËå«þýÿëï¼ý§¿ýõö·Ïÿü¯ÿ§+ð—¾íW¡þMsÿ¦°ÿ«]Ò¿þç_ÿáó7½õþÓþݲ‹U¾)•ÿ­ïü—þ§þîþåÿã¿þÃÛ'ñÿSQÌöõ7—·Ðßóñ½î+ʹ¯Vnœÿ0ÝÇ5¸;zV¾ÉÜÁþ=FõÆú×fÞ&¾Oa¾ÚÓ~¼­Âwö¶Gž-í1Ík~þi›Uþíýó-ÉÕ؛՟ȆþWËúCQ¿÷å÷Ï—rZþÔÍã측j<-ú$ª5³¼ç}ôøó«þeÚøC-rgÚü3ßõJ)òÇ}çDÌra;½À 75œ`’??Ó¨ÞXß¼10ó6€€€€€€€€€€€€€€€€€€€ðÝ RøjO+}´¾³·=òliùk^ó×óV£ÏÆ·›ëLÿ1„¯²qf¹‚ç…òøÄ[”]6ŽèÜq˜ìz%GùÏ½Ï ùmïav^J Xå?ÊÒ¨ÞXß¼10ó6€€€€€€€€€€€€€€€€€€€ðÝ RøjO+…²¾³·=òliùk^ó…r“CmäŽä6.1øqAç7ÐÉ)^—çüî™· „ïS˜…¯ö´o«ð½í‘gK{Ì_óš(”ûþ\nNž2Ž`ãTú™å žÊã1TÍÉÃ[dãˆò°¥yvózEn·öÛD,«r¾ÊANù8Á(“Æå·{c`æmá;´ðÕžV e-|go{äÙÒó×¼æ ådªéœæ3ýB•§OÏ,Wð¼P.ruOs–#:åzôÒõ•ê䤋s"ùëñÌ ÌrÍ;XÊŒêõ;¯̼M |§‚¾ÚÓJ¡¬…ïìm<[Úcþš×üc¡ÜŒMãŽä4ÊÉve¹‚§…r´&ø[ Æ&Z£³í8Cy¾’‹©i›HýÍ}ùÈ‹¾µÝ-ϨÅX¿óúÀÌÛÂw+Há«=íÇÛ*|go{äÙÒó×¼æ åس<„/È…ÈDSÚÊrÏ å`’TÛÙ„Q†ÏìÆ×>_ÉÉÔºMôK.÷öy-µìb4.­¬¢Nœ¢NhÞ'000000000000000000000^×^ZÛdé½µ—žíò‘g—|Ì_óš콓‰ò¬À(xÈÓêÊrÏ{ïlŠ½µ².}DysmÝH­¯ä¼hÍýŠÛOæèë6o`1!θ<›wo ̼M |§‚¾ÚÓJ¡¬…ïìm<[Úcþš×üc¡\M–gFSÇiÔÉWV–+x^(7SÛ­ØqdÈmÆqÇöÁ~i˜¬IaÆåõõÓÓ™· „ïTÐÂW{Ú·UøÎÞöÈQ[Úcþš×üC¡Ü÷W;ž5èµPŽ&ä•å žÊ©¿0„XtãˆÎgÓôØõJu&»m")ÑwzÞÄx?£zcýÎë3oß© …¯ö´R(ká;{Û#Ï–ö˜¿æ5ÿéíç÷·Ÿ~u7w{ÿM~ø|Krʆ½Yý‰l°ñæm5¥ÞÞ?n?üñãíý÷¾üþùRFÿò>¾|wåQƒV+óbRZYÞÀó>:_nÎ%SÆ1Ø3û:z8Þº¾R¢q~›ˆíÉ‘)™îÅæVVQ'vQ&4ï¯jh/­m²ÔæÚKÏvùȳK>æ¯yÍCmålW«©ùjójœž+£¼lM‰+Ëx^›7óÍÙjÒ8ÓdæàŽ›Ôç+ýJ¼Û&ú—L÷öû%ûMì×âìÊ*êÄ!Î ÍûÆëÒKk›üñ&ÒKÏvùÈ},éq&:Íkþáø’ìäi2_ô ‚i~e¹‚§½wö&Kïmc§æ˜MœG¯è+ÞZcÓ6Qƒ4ý»}ŠÖíb0¾­¼Ä>±‰cbæmãe í¥µM–Þ[{éÙ.yvÉÇü5¯ùÇÞ;š$O¢tr¯¸\A‘ó¾g–+xÞ{'Sc7Éd;vjý+ÍV__ñ.¬w§%H¹¿Û§èí.Ž‹žy‰}bÇÄÌÛÆËÚKk›,½·öÒ³]>òì’ùk^ó½w‘îÙ¥$/÷+(VÎS™Y®àyïÝŒ *£ÓWnRåëµW¼¦Åm¢¿É|ôø« í^ÌeeÇÄØ'fÞ&000000000000000000000^ÖÐ^ZÛä·ÕKÏvùȳK>æ¯yÍ?ôÞ}ÿR.{“Æ9'%˜ØV–+xÚ{/G§H‹ŸÆµÏÃñîæ+ÍÎþ˜(qór齋3©Þ‹5­¬¢Nì¢LhÞ'000000000000000000000^ÕÐ^ZÛdé½µ—žíò‘g—|Ì_óšì½£qr¶•ãRä ²Éue¹‚ç½w2QOqч~Κ)«÷¯x›L+Û„~ÉÝ>¯¥”]ÌÆÆ•U›8&fÞ&000000000000000000000^ØpaµÉÒ{k/=Ûå#ëØÇü5¯ùÇÞ»˜ Ï±´z¿w3µ¬,Wð¼÷®r\‹ Å4Ý©99y¦^ûxÅÛ`jØ&J0>ÞÙçµ´¼‹Íø°²ŠcbÇÄÌÛÆ £—m²ôÞ³—ÖvùȳK>æ¯yÍ?ôÞÕÊ&Nn ×Xƒ±ye¹‚§½wu¦´qgxwŠ¯ìÞ{¾Ò’IûDÍr“ûnŸ¢K»èMô+/Q&NQ'4ï¯jh/­mòÇÛê¥g»|äÙ%ó×¼æ?½ýüþöÓ¯ý•ÛûoòÃç[’K°7«?‘"½Ä›sÕ¸x{ÿ¸ýðÇ·÷ßûòûçKgþËûøòEnò.&kmžedfyöimåžv9ì¼jΛÉóføùJ-rGû9‘û—ørmMˆ»˜ä"f^¢Lœ¢NhÞ'000000000000000000000^ÕÐ^ZÛd©Íµ—žíò‘g—|Ì_óšÿzm^½iáæZ11}CmžM•{Ä‹Iã$˜ÚLŒ+Ëx^›Wãòmœ•®b 9õë÷Òë+ÞÊ6Ñ/0·;{‰ý‹†{±Ú•UwbŸ˜y›ÀÀÀÀÀÀÀÀÀÀÀÀÀÀÀÀÀÀÀÀÀxYC{im“¥6×^z¶ËGž]ò1Íkþᔕf•§hFiµû´¾1¬,Wð´÷n΄4þÀŽw7³œK>Ÿ:_é_BŸ*:s¿¶ðÅÞ»ÛðïÄÖVVQ'vQ&4ï¯kH/­mòÇÛê¥g»|äÙ%ó×¼æ¿ávñfB_vÍÄüõÛÅ[0^7or¹¼dª_YÞ€}Z›Gyrì¹þ¡ÀÌ9¯›Áç+ÞZSÓ6Ñß‘Kwöƒiv“œ,3³ŠcbÇÄÌÛÆ ¾¬6Yjsí¥g»|äÙ%ó×¼æoÏR»R乂f¬[Y®àyï]L‘?ÊW’šsY×>^ñ®ï·‰þ%ç å×Þ»Ûv±Ê]ó3/±Ol☘y›ÀÀÀÀÀÀÀÀÀÀÀÀÀÀÀÀÀÀÀÀÀxYC{im“¥÷Ö^z¶ËGž]ò1Íkþ±÷n&ËÑ%Á9ß<Ù¾nW–+xÖ{'kMó79€%§±Ss¦ËÎùŠ¼9ý˜*þþ^òSôuœQ>óûÄ&Ž‰™· Œ5V/­mòÇÛê¥g»|äÙ%ó×¼æ¿~LJ«¦öKòÙøøÕcRäËW9<›¤µy’ãYf–7ð¼6ÆõËHÉÔ2vj.u=rt¾â}3-Ÿr3¼»·—L,÷bI+«8&îÄ>1ó6ñ²†öÒÚ&Km®½ôl—W^]òš¿æcþz»x²ã¨sß™ÆT)¼g–+xÞ{ä˜8nd¿­\ãºÑ|¾â}Ð#XVvþYïåO îÄWVqL܉}bæmã… ëW›,½÷è¥W»¼òê’×ü5ó½w3^ŽY錕+pΔ²²\ÁÓÞÛYå˜*çËNÍ5Œçyʵë+>&ë9!×Ó}^KM»èŒ +ë§1&6qL̼M`````````````````````¼°!½´¶Éo2!½ôl—W^]òš¿æcþ¡÷–A½ÛêD¹1|f¹‚ç½w0¹ŽhÚÑÙÏÜüq7»¾âC“3Ð ï£\ònŸ¢»Mð+/±Ol☘y›ÀÀÀÀÀÀÀÀÀÀÀÀÀÀÀÀÀÀÀÀÀxYcôÒ³M–Þ{ôÒ«]^yuÉkþšùÇÞ;™$ç¤d9e\® W–+xÞ{g!]m&ëΑ儕fçµë+Ñɽçç„÷&ä;û}ØÅb’[y‰>ì☘y›ÀÀÀÀÀÀÀÀÀÀÀÀÀÀÀÀÀÀÀÀÀxYcöÒ£M–Þ{öÒÚ.yvÉÇü5¯ù¯>SŠôÚ/)eÓÊ׋ٿ|‘cVª ãxpïä'3ËxZ›{+í¿«Öx?vjne=Ps¾â‹3Õ>Óܽjóf¢¿‹]Y?Ì1q'ö‰™· Œ—5F/=Ûä·ÙK¯vyåÕ%¯ùk>ænïû›Üä]æñà>šäW–+°O{ï ’kÖTÝ9óƵë+¾d¹'þ˜ð!­gÚûÞÛ{“ݽXÛÊ*Ž‰;±O̼M`````````````````````¼¬1zéÙ&Kï=~²Úå•W—¼æ¯ù˜콓<ÓËù,ãs_Lq+Ë<ï½³‰q¼§ïndoÏ£Íç+u< óœMÊýÝ>¯¥Ú],Æ–•U›8&fÞ&000000000000000000000^Ø^ZÛdé½µ—žíò‘g—|Ì_óš콫 òÌ&t¿‚0ÎS™Y®àyïÝL–C`’ܲ.×>ò8eÞ«>_Iy<Òó˜ðÁ¤û3TÎkimûµø¼²Šcâubæmã… é¥µMþx“ é¥g»|äÙ%ó×¼æzïàL’ãÅ­ÌËDy"çÌrO{ïàMÕ;Ïõðó•ÛñîôŸÊlõg–7ïìStuƒ‰iå%ö‰M3o/kŒ^z¶ÉÒ{^zµËoGO­]òÛÑcßçcþ±÷Ž¦ÈùàN w¹‚b|]Y®àyïu7W²Éã„ÍÞºóÚõ•bkÛDHrI»½Ä$‡«Ü‰9®¬â˜¸ûÄÌÛÆËÚKk›,½·öÒ³]>rÐ.ù˜¿æ5ÿØ{Óü¸¿»ŒÆ£5±¬,Wð¼÷nr‘r'yÒkŸ¹žïN_É~½;ÍÑIý½ÛK¬&å{±†•UÔ‰]” ÍûÆ«ÚKk›üñ¶zéÙ.yvÉÇü5¯ùOo?¿¿ýô«»¹ÛûoòÃç[2.ÝìÍêOdCÿ‹oNÎ$yÿ¸ýðÇ·÷ßûòûçKgþË{ÿòÎ8¹É»§µy09¯,oàim½ õæòødçÈòáÌë•MiÛDÈò—Ý^¢“¼‰ýKú•U›8&fÞ&000000000000000000000^ؽôh“¥6Ÿ½´¶ËGž]ò1ÍkþávñMÇ`ÖYÜÇljZY®àyïL’›á‹iãfxÍÞfiþõÚõ•äŒÛ„o¦}¹÷Žr¶Ë&fãÝÊ*Ž‰M3o/lH/­m²ôÞÚKÏvùȳK>æ¯yÍývñ8¨é‹—+¹Ü.þûüú}æãtïÓ<–òì’ùk^ó÷™÷ý¹Þ¼óÒj÷+HÁø°²\ÁÓÞ;9ÓR7M‚‘#{Û¿À¼v}EÞy>'ú_â}§~ŠÁí¢—slf^bŸØÄ11ó6ñ²†öÒÚ&Kï­½ôl—œ´K>æ¯yÍ?öÞÁÔró.™ÆdýÊrÏ{ï$G¶¸æå€Ù9²4ìež¯2_ÉÉd»MkZ¸³—¥î¿K]YÅ1q'ö‰™· Œ—5´—Ö6Yzoí¥g»|äÙ%ó×¼æ{ïbl¾yåFs¹‚f²[Y®àyï]M7W³Tõ²sdo‹I«³×WZ4Õn±_R¼³ÏkÉí^leeÇÄØ'fÞ&000000000000000000000^Ø^ZÛdé½µ—žíò‘g—|Ì_óšè½³5^ÎóΦy¹‚ìåKÏ,Wð´÷ÎÎ$9ÅÅš<îU×,'¤8½S|½Rª©~›Ð[Ôwû¼–Zw±SyeÇÄ&Ž‰™· Œ6¤—Ö6ùãM&¤—žíò‘g—|Ì_óš콃‰áæCgÆY,9™ÖV–+xÞ{GSÜhñ‹^ûÈò•Úºv}%—ù´Ð™S¿¤zgŸ¢-»8[™y‰}bÇÄÌÛÆËÚKk›,½·öÒ³]>òì’ùk^ó½w6Ùß¼/&è56ãÊÊrÏ{ïbší年Fþ6³wãQzíúJÿK'ωh¥þÞíSôy«Ü¤>óûÄ&Ž‰™· Œ—5´—Ö6Yzoí¥g»|äÙ%ó×¼æ{ï&7v{ï¤ïWP¼ ye¹‚§½wqÆÊ).Á¸qŠ‹fo«ãÚç+ÉÉ!æçDˆ¦å;{‰ÖÄt/–°²ŠcâN,aåmãUÙKk›üñ¶zéÙ.yvÉÇü5¯ùOo?¿¿ýô«»¹ÛûoòÃç[’ÞÞ¬þD6Ø~ÑiœÀòþqûáoï¿÷å÷Ï—Îü—÷þåGî]–ãÄå $“ÒÊòž×æQîHwÕ›¢ÎÈÞö¯1oŸ¯ä,pNÈ#;Ë½Ä çÇ܉ͯ¬â˜¸ûÄÌÛÆ ÒKk›,µ¹öÒ³]>òì’ùk^ó·‹—,¾÷~ÞÒ^ª)qe¹‚ç½w1±t9Í#^fvÞ„õÇúJÊrBù9!•þ½}^K »Xs+«¨§¨š÷ Œ×5\[m²ôÞÚKÏvùȳK>æ¯yÍývñ’Æy⥘¾ávñfBí|3yœóRl›YÞÀÓÚ¼Z“å '¦ìÙÛ&¤·>_ MŽ’9'‚•G‡îö)Z¿‹ý'må%ö‰M3o¯jÌ^ZÛä7™^z¶ËGž]ò1Íkþjó:Ž#ÏÞ¤o8e¥ö1½G¼OËHrŠùÌòž×æÁT9Í%›àÆNÍ­G¯ë+>zÙyLxßL¾·OÑ»]Œ&Ö•—Ø'6qL̼M`````````````````````¼¬1zéÙ&Km>zéÕ.¯¼ºä5ÍÇü×kóêÆiæÝMñjódŠ<Ã3I£-o Ê{›YÞÀóÚ¼È=÷®Úqýmfïœ)ópòùJr〙c"“Û½Äl¢½sYYÅ1q'æ²ò6ñ²†öÒÚ&Km®½ôl—<»äcþš×üÃáä}“‡pZSÆíâmæ¯yÍ?ôÞÍ'á´¦‡¶(…ùÌrÏ{ï`¢»99•EßÝÈòøO»zo}%69qýœðýMñtñ~-¥îb46­¬â˜ØÄ11ó6ñ†ôÒÚ&Kï­½ôl—<»äcþš×ücïL§h&Sµy/¦Ö•å ž÷ÞÙd¹›½˜6îf×, ódôùJ®ó •™C’›Üwû¼–Vv±Hý?³ŠcbÇÄÌÛÆ ÒKk›,½·öÒ³]>òì’ùk^ó_?¼EiÌ}÷fõxðVMòã8ð(µdëŒ-+Ëx^›7SÚ͵ ×/o}äñøÏõáè+±o·‰àäöÝ>E—O1[kbXy‰.Ÿâœ˜y›ÀÀÀÀÀÀÀÀÀÀÀÀÀÀÀÀÀÀÀÀÀxYC{im“?ÞV/=Ûå#Ï.ù˜¿æ5½]\ö9º¤è-íÙFãóÊrÏzïl½iõæj59Œ#{[Óë•r¼;Í¡™ö¥ãÁe=Ä] &û•—Ø'6qL̼M`````````````````````¼¬¡½´¶ÉÒ{k/=Ûå#Ï.ù˜¿æ5ÿØ{G¹w]ŽE)e\A11­,Wð¼÷ÎÆ•›ËI.YvŽì;˜ü¼v}%;!Ή×Íð×Þ;™îÅêVVqL܉}bæmãe í¥µM–Þ[{éÙ.yvÉÇü5¯ùÇÞ»ÛÆ1*nôÞΚW–+xÞ{79yÅUg¬¶úš[“ÎõÚÇ+rØy©ç„œ¡’ï?¯¥øMì×bíÊ*Ž‰SÔ‰™· Œ6l[mòÇ›Läx´Ë+¯.yÍ_ó1ÿÐ{;g|¹yçMsã ‚©ae¹‚§½wIi±‰cbæmãe í¥µM–¢^{éÙ.yvÉÇü5¯ùÇÔ‹òr{×è­ñne¹‚ç½w5-Ü\ &jc®¹¹óÆ+>9;æ˜ð.˜”ïìSôm›Ieå%ö‰M3o/kÌ^z´ÉoG/­íòÊ«K^ó×|Ì?ôÞ}•»´Ë¼‰ÞÚÊrO{oï“cgühäoGnGï­¯ø茋ç„÷EîKßí%:ë½XòÊ*Ž‰;±ä•· Œ—5F/=Ûdé½G/½Úå•W—¼æ¯ù˜ì½£œŽîZ0M¯ ›TW–+xÞ{'ô—::û™k‘ Òk¯xçLmç„·Až6ºÛçµär/¶´²ŠcâNliåmã… é¥µM–Þ[{éÙ.¯¼ºä5ÍÇü×$÷aœ`¼qåëǃûb¼×cQÊxÍ”²²¼çµy•;Ê]òpNÙ©¹VéÆõ­WätöbÏ o›ñ_|Žg¿–šw±ÉŸ̬â˜ØÄ11ó6ñ†ôÒÚ&Km®½ôl—W^]òš¿æcþávñ`Mtãöð:niÞ´¼²\ÁÓÞ;8“ÛÍ¥j|;5—bܼ^_‘#W²;'¼ÙíS´i½ ~å%ö‰M3o/kh/­mòÇ›LH/=Ûå•W—¼æ¯ù˜콃ÉòèN'wËd¹i|f¹‚ç½w”jÜÅ0¨9sŠ&ÌÎ^_ñ¾õœ›áóý#7OÑÇ]L&¹•—èã.Ž‰™· Œ—5´—Ö6Yzoí¥g»¼òê’×ü5ó½w6EïÒöãxð0ΟY®À>í½«±¥ËÖÄòì’ùk^ó½w òøKW:3îIïWüÊrÏ{ïhb¼¹àç-3§þ–fg?_iÉø}¢ZÓ¾x¾·Üïv±µº²Š:qŠ:¡yŸÀÀÀÀÀÀÀÀÀÀÀÀÀÀÀÀÀÀÀÀÀx]ÕÕ&Kï­½ôl—<»äcþš×ücïM»´£©Ú¼WÓÜÊrÏ{ïbrèï&ÊA*²Ssr&–yíúJ‡·œ¥‰/žïÝ×û—ØÄj|Yy‰2qŠ:¡yŸÀÀÀÀÀÀÀÀÀÀÀÀÀÀÀÀÀÀÀÀÀxUC{im“¥÷Ö^z¶ËGž]ò1Íkþ±÷n&Éyá^Ù¯ ÿÄÙ•å žöÞÉÊ1åÎg¹dÙ©9Úã|ïùJ&úm"åÕû_zï¾îÚ.:9fæ%ÊÄ)ê„æ}ãEÙKk›üñ¶zéÙ.yvÉg}ÉkþÓÛÏïo?ýênîöþ›üðù–äÁ˜öfõ'R¤—:nÏñöþqûáoï¿÷å÷Ï—Îü—÷ñå‹nÒÆå)ßV–7ð¼6ÆöËðÕØ0vjŽÙØU›ë+5K³NȇóÅãÁS0¡Ü‹9­¬¢Nì¢LhÞ'000000000000000000000^ÕÐ^ZÛd©Íµ—žíò‘g—|Ì_óšÿzmžœirIÕ”ü µy2-t>È1çòª‰ueyÏkób¼½9׌ÓÂ]s(rܹ¾u}%G9èåœÐ?(Øí%f“ò½XãÊ*êÄ.Ê„æ}ãU í¥µM–Ú\{éÙ.yvÉÇü5¯ùo¨Íû_ú%•&¾^›÷÷"G«T“ƃC³3¹¬,oàimž­ r+¾5u|83‡$o}¾RÒxëÇDÿŒÒýì«6o’7ÑÉO̬¦Nœ¢NhÞ'000000000000000000000^×^ZÛä7™^z¶ËGž]ò1ÍkþápòìMÃÎãæ¯yÍ?öÞÉD9µ%ʽârÕØ´²\ÁóÞ;›RnÎz¹Ù©Ù;Öµë+¹ﶉÖáì×Þ;v±È“Ag^¢Lœ¢NhÞ'000000000000000000000^ÕÐ^ZÛdé½µ—žíò‘g—|Ì_óš콫ÉuœkâÇñàÅW–+xÞ{7Ór7ã/ríšÃ(ÚõÚõ•LrÛDjë¹›×Þ»™à7±X“íÊK”‰Cœš÷ ŒW5´—Ö6ùãmõÒ³]>òì’ùk^ó½wß_åìq¸\A4ѯ,Wð´÷îéR—û%ëNͱhc¾^iýZÜ6Q’)÷ú½¼í;±´•UÔ‰]” ÍûÆ«ÚKk›,½·öÒ³]>òì’ùk^ó½w_{²½ ãÁœ¥˜ìV–+xÞ{gä—ñ$OÙ©9:ÓÖµW¼õ&Úm¢6)÷wû¼–bïÅVWVqL܉}bæmã… é¥µM–Þ[{éÙ.yvÉÇü5¯ùÇÞ»J–ž;ó½«5Õ®,Wð¼÷nòðMç³<¬R®]sÊÇ-úŠwÕ¤¼MÔ´N/¿öÞÕÔ¶‰ýZ\YYÅ1qŠ:1ó6ñ†ôÒÚ&¼É„ôÒ³]>òì’ùk^óŸÞ~~ûéWws·÷ßä‡Ï·$‡Ø›ÕŸH‘û%Ålb¹½Ü~øãÇÛûï}ùýó·llýëÙë¾û®ý—÷~ÙÎÄ82×ÇiÖ`Z[YÞøÓº½z¹“ÝEkl;5§dBÐL_ñ>G5®BþR·÷u[w1˜W^bŸØÄ11ó6ñ²†öÒÚ&KÝ®½ôl—<»äcþš×ü×ëöjårŸœüåÏÔíºÑÙlÒCßþËûÛßÝþï[ã¶oØ~üïÿùR~»ÛéÿûýÍÉúí¿¾EïŒk·lçÑìutç庬Ϙ¢Ü®¯ëÑ„º­¸¯“ʶ>â¶Æõë÷õþ@ÜÖGÜ׋<1õ\q[ï7®kÜ×Çóçúˆûz–ÿRá\q_oò':çúˆÛz ò·à±®q_Ïò`Ös}Ä}½™˜·õ·õ<þƒ…c]㾞ä?º8×GÜ×›üÓr®¸­/~=Ö5îëIžP{®¸¯Wù2Îõ·õþPÙ>?ûzÿéöùhÜ׫qÛç£q[o΄íóѸ¯G“¶ÏGã¾^LÙ>çz²}åü|fÜ×£qn[q_/ò Þs}Äm½ÿƒÏÏgÆ}=È/ çúˆûz–óŸÎõ·uoå¿“9Ö5îëAþ,ï\q_Ïò à¹>â¾ÞLÛúˆÛzð¦nŸŸÆ}=ËQXçúˆûzÛ}šq[~ÿõiÆ}=í¿>͸¯WS¶ÏGã¶Þ!oÛç£q_8×GÜ׫üÎq®¸­ggÒöùhÜ×£)Ûç£q_/¦miÜÖ‹3vûü4îëýwºíóÓ¸¯ùsãs}Äm]~ÛÚ>?ûz0uûü4îëEàq®¸­7»ÿú4ã¾ö_ŸfÜ×óþëÓŒûzÛ}šñ\ÏÖËOß?5îëYþ„ÿ\q_oò-ɹ>â¶î¼Içç3㾞L‰Ûúˆûz5-lë#nëÞw~~3îëI˜r®¸¯Wù~ä\q[ýíû ûz4uûü4îëÕØíóÓ¸­GgüöùiÜ×ãþëÓŒûzÙ}šq[OvÿõiÆ}=î¿>͸¯¹?â\q[ï+aû|4îëAn9×GÜ×û7$Ûç£q_Ù9×GÜÖËxͱ®q_göë#îëã¿m<×GÜÖ÷o(ï/Ïõþíòãw­ŸÿÖ¿öó[áñêøá/ãG}!Ê_Ê­ÿrßÿï”ï†]óN¾!þúNùOBóÝÎð;ËøÅiß¿mg²I¾­Ýw¦oÜÙ¿¡ÍñngþÆýÿêt¿³|ÛÎlû¿Ô„»õwöoÏû·×ûÎö;û7¦~ßé­ÕýßmbÿU±Ï´q³Tÿ&L£ï¿Wøù]Úu Èw±·u×u ù˜yŸHãû¿sBó>Q³ü|NhÞ&än.ùo–×ÄÌûD¿Ëžš÷ }oçÄù^Ç¿`Îsûÿ²}°¡Û—íß+Ï[ºþ’ïÿOø·÷¹þMbÿËØÿÔÆþÝQÕÿûþâÿÔÆ:þµN6Ú?³Ï÷+íÿº7þ~ùSûú…ößÆdߟzƒ¾û2ÿí;¯¿9û¿xßBÞFóþ—ϽýÛ?ý¿Ì…ù…¿þc1^é‡0þµ[¾îßÿðÿãí?ýí¯·¿}þçý?ý—–¿äôm¿ ôï}­tòw´Pÿó¯ÿðùÛÞ}±ò]¯üß[ôSûo}ë¿üó?õ·÷/ÿûÿõþØ>Œÿ¯Bÿ®Ÿ¿K¼åþ®oZÿX18'C¬ék^ãŸÞþîíÿ~/|£endstream endobj 6 0 obj << /ExtGState << >> /Font << /F1 7 0 R /F2 8 0 R /F3 9 0 R >> /ProcSet [ /PDF /Text ] >> endobj 7 0 obj << /BaseFont /ZapfDingbats /Name /F1 /Subtype /Type1 /Type /Font >> endobj 8 0 obj << /BaseFont /Helvetica /Encoding 10 0 R /Name /F2 /Subtype /Type1 /Type /Font >> endobj 9 0 obj << /BaseFont /Helvetica-Bold /Encoding 10 0 R /Name /F3 /Subtype /Type1 /Type /Font >> endobj 10 0 obj << /BaseEncoding /WinAnsiEncoding /Differences [ 45 /minus 96 /quoteleft 144 /dotlessi /grave /acute /circumflex /tilde /macron /breve /dotaccent /dieresis /.notdef /ring /cedilla /.notdef /hungarumlaut /ogonek /caron /space ] /Type /Encoding >> endobj xref 0 11 0000000000 65535 f 0000000015 00000 n 0000000064 00000 n 0000000206 00000 n 0000000291 00000 n 0000000371 00000 n 0000043654 00000 n 0000043758 00000 n 0000043841 00000 n 0000043938 00000 n 0000044040 00000 n trailer << /Info 2 0 R /Root 1 0 R /Size 11 /ID [<65afc89bde22018dca44f46b641d2111><65afc89bde22018dca44f46b641d2111>] >> startxref 44302 %%EOF gnm/MD50000644000176000001440000002441712515671462011414 0ustar ripleyusers7f541872b75ba736c841e471b7a9bd18 *DESCRIPTION 8285accfa7f7d4838e714e69afdde143 *NAMESPACE 854496a25c0c74492a5047b936d55ab5 *R/Const.R 15bae4b18a9b5a81677347c080f12cb4 *R/Diag.R 5fee39db2b681eff9a6b1b0cd336eb94 *R/Dref.R e20fafcb9c40fa3fb4ad864ce638fd28 *R/DrefWeights.R 7a7e2f355094ea09b3727919a2a810f7 *R/Exp.R 1163bbf6c25fb51810d0219643d31fb4 *R/Inv.R 1be94a5f1cbc953c9547efcd219250f9 *R/Log.R 948e44dbfd7e9b44638fa6e804f5623f *R/Logistic.R 32ce6aac8288253199fd161a3750edc6 *R/Logit.R 3820f8ae17dd6e16bd3f7b3ff78c0c8a *R/MPinv.R ce0a897eb161f67a8b3e1a9758dd6e7b *R/Mult.R 56b99626a0fabc857f5b5740a7b16719 *R/MultHomog.R cd7fd3c31dd817fcdd8fd533926279a7 *R/Raise.R 3ec65ae6027cf15ef47da35b620c6910 *R/Symm.R 70d6c78e132bb23c8b44a683a2fb0dfb *R/Topo.R d187978d3dbfc4b4a9a33a4ba70b6099 *R/add1.gnm.R 3fa8820d54cbd3ccc84bf37f583d1104 *R/addterm.gnm.R c11c021ce659dcaa94158059983e0498 *R/alias.gnm.R 728df721820db99459a313ce15e43d1e *R/anova.gnm.R ce592c42122cc711bda928ee403e4105 *R/asGnm.R 3133d8a14d3f6df07bdc093409b8d664 *R/asGnm.default.R 3f343550a46023f76d92fb43dfe18e81 *R/asGnm.glm.R fbe0395a6fe899b3412606d88153ad12 *R/asGnm.lm.R 954badc596f77b0aac6615e41c214717 *R/boxcox.gnm.R 75b9c6027d9e86f0a44bb0c16458e873 *R/checkCall.R c83e266015ec278c98d8ce1702077b61 *R/checkEstimable.R e02ab67e5b1cf204e8b2df3b72389415 *R/cholInv.R f8239fe470d0d8e3a74c1c5e32c2b070 *R/coef.gnm.R 0534ab467e95d9cfe46c1df4552df0b8 *R/confint.gnm.R 410e30e1d2d6166c40fdcb52517d7475 *R/confint.profile.gnm.R c41c1536e1866d8037032da16de62f47 *R/cooks.distance.gnm.R aae1beb67cc6062c2f0d1fb4f28c2238 *R/dfbeta.gnm.R 31254d1c7c6eed6abe766b71208b8722 *R/dfbetas.gnm.R b04dfc82a13bbe887f45cec3cb542908 *R/drop1.gnm.R 597fb498683d628c7d52f2e288289ab1 *R/dropterm.gnm.R 7b226f6597ab38664e89de3b319c11ca *R/dummy.coef.gnm.R 35aa210357604f393bc2994a25821b94 *R/effects.gnm.R 654710a33bb547685fdea664e3c03e9e *R/exitInfo.R 64bc4cfd43ac4b8089f877316a8f2382 *R/expandCategorical.R 5e4aaa614ef38eafe07f57eba879d957 *R/fitted.gnm.R deb576177f9224df45d283a696d41964 *R/getContrasts.R 461cc8f39f38d35edd430e870328aa3b *R/getData.R 8467f3c8cf017738180c198ab1f5e69e *R/glm.fit.e.R 512c1d6b73395b508ad1095c9783453d *R/gnm-defunct.R ff5127b7e56965a0cb90046cb589ad1b *R/gnm.R f44fffffc6fbd2d5b92e215ffc98a93e *R/gnmFit.R 4d9338e33215facd25e8b7cf4ea90987 *R/gnmStart.R 48d5a59edd38a099def6ef5069e025d9 *R/gnmTerms.R 6bd1e1c1fa38d380a9daf2ce8af23653 *R/gnmTools.R b6be5ae7a90f3cdc0e9d054505d1dfdb *R/grp.sum.R 68992380696b028e9c7a6cd24b52b2bc *R/hashSplit.R 0211542e7701cda87f26f69d4ed516c1 *R/hatvalues.gnm.R 6a8b167dd634bd1533f51ff2cf781e84 *R/influence.gnm.R c77000f134b8096b3f55c9efb977e73b *R/instances.R fd6cbaee5c3d909a31889dba229741e4 *R/kappa.gnm.R 0120485b96379a3d741ef386f1dedebb *R/labels.gnm.R 876d01bb73a00985db1c3bb8ab8a3206 *R/logtrans.gnm.R d26bf73a737729cdd09482fa6fa7bd37 *R/meanResiduals.R af43d37bfd33c2c1325eaa7fef761c78 *R/model.frame.gnm.R 838e4f634c956834ff3209838b95f074 *R/model.matrix.gnm.R a0f246dfc591ea337f018f72100c9ded *R/naToZero.R cf3704b714dc11e3c39e96b45572103a *R/nonlinTerms.R 7f89fd4a6221fa64df48b8f86ca80ec3 *R/ofInterest.R 26ebbe121dbe55779aac6423bf10be9b *R/ofInterestReplacement.R 2b3a80421fb27a5f9f1a1536f6d0bcdb *R/parameters.R 584f74898b8de13acb948d1d6714fb30 *R/pickCoef.R f3b56eba9b769ac6d41b67cdf9e16ea1 *R/plot.gnm.R fc2b02c97a5bf22b3f8a7c00ba5cb250 *R/plot.profile.gnm.R ad13f0645e87f20b9788a099602f04fd *R/pprod.R 50c8ed0dc792d486adf8defb3311e48f *R/prattle.R 598b4b565837607e911a67c46a296450 *R/predict.gnm.R 3b21b08dbeb848c8c5fc202f68d32818 *R/print.coef.gnm.R 385e40991edec5085c915233e68e21bd *R/print.gnm.R 8840fb853d3edafef7b1a1602ed0750e *R/print.meanResiduals.R dea050614f63bef7044cd199b2af1296 *R/print.profile.gnm.R 9689be0a5036c00d99b0287a5a2ff422 *R/print.summary.gnm.R 0db04b53db8222364c02cfada6a4ee98 *R/print.vcov.gnm.R 2e07923ea518e4d62738e7c88fb49dac *R/profile.gnm.R 07123ec72cd3e255f1556af97c10f8cc *R/proj.gnm.R 6397c01c19d1dc9e2e53ba447c3afa8f *R/psum.R 0c101c57f953af46125d1dd658294483 *R/quick.glm.fit.R c7ef02916b6b62a2e5b5a8a5a9b55fce *R/quickRank.R 1cf01ac34302023ab3483815fc765645 *R/residSVD.R 87a178b9b723b19f6d6ce19d0e2a1b8c *R/residuals.gnm.R 279be6c011ff254bcd7bd7428d3c90b9 *R/rstandard.gnm.R 99830d4e74dfa4f0a1c539135e2e7370 *R/rstudent.gnm.R a27976daa2bfdb60515406590e19296d *R/se.R 150226136bf1ca5996f68cd68866fc05 *R/sumExpression.R 5ddfa1e7782f0cdcbd1e5bbdd04e215c *R/summary.gnm.R d146f76866e65641b6a8b7d7fa3b73f9 *R/summary.meanResiduals.R 6b5ac9ba80f30a49a5b7901b3f4dce0e *R/termPredictors.R 2bdbd1e59eaf31da95f953b9f6bd44d1 *R/termPredictors.default.R 0fcea4dfca5685734a3c594d853b905d *R/termPredictors.gnm.R 14c1377b257b1c7f36caf0b1d8e130dd *R/unlistOneLevel.R 55823d141ade47c4e32470b9242a846e *R/update.gnm.R 939d8938906a946a477745a3adc6c5c3 *R/updateLinear.R 192d90bae21e8ed3bd4c37997ee522fc *R/variable.names.gnm.R abc1b5f31d4068a6c821fe0715166cdc *R/vcov.gnm.R 0ba281b4a6492dc7e28cafb62b442056 *R/wedderburn.R af58d5d6c6349e1353f1c8811f5abaec *R/weighted.MM.R 051baec2bdd0bce8410af20d4448deac *R/weights.gnm.R 23ff83b70cc72eeda326b5be80db4eee *R/zzz.R 77576c3062f975b5b3a9b23361ddc402 *build/vignette.rds 8f72b925841b1c376f4b13c6c488c2e1 *data/House2001.rda 4962afb549958d7229ff8799ff2788f1 *data/backPain.rda cc42f2dc6c3f880c131254cb094b2daa *data/barley.rda b0decec2aca96e108ea4ae279e17a064 *data/barleyHeights.rda 3f60b1c199cff48c2ee9463787bd2e16 *data/cautres.rda 5391431659176b8181276c2c7b2ac8b3 *data/erikson.rda f972851cdb782ea8a1a975e6c63c3964 *data/friend.rda ce883333b4f83e251e5208df56cbbd92 *data/mentalHealth.rda 362c4f2e32d6ed80cf78ff31b91b7847 *data/voting.rda 83c098ebe1bf7a9e1846ced7800ab139 *data/wheat.rda 6745ad3f2df2e1fe373743ba4589bb6b *data/yaish.rda c8c477e8cc92a00190fc07f155cc1df8 *demo/00Index c9d7e30a137b9f9fb6311ecf82d81842 *demo/gnm.R b32f42d7df5a203bf363c5d922fe4a00 *inst/CITATION 4d79ba3e19786b63f4850486a0bee94b *inst/NEWS 7ca3b3eb71c300a9339a3535f3524bb9 *inst/doc/gnmOverview.R 1fcfca4a717ed1f39f6adeaf1e29fd50 *inst/doc/gnmOverview.Rnw d49612f473fef1bef5107586c0c2bc63 *inst/doc/gnmOverview.pdf 85e992e91d7cba65da2a6d599bd31413 *man/Const.Rd 0452668f00f9a1c18625804512d13eb9 *man/Diag.Rd 0358eb8f515685aa7cc613e41ed1b887 *man/Dref.Rd d474d3d982cec7798cee2d76e547bfb6 *man/Exp.Rd f01b62d1ceb9a304e5874125cba7cac4 *man/House2001.Rd 264892ca68e2b7c62ed161035af01d9e *man/Inv.Rd bfb3220d153ffb0a9ca78878149ff28a *man/MPinv.Rd b56090e88095215fdbe1c862b98117f5 *man/Mult.Rd 858863f066143f20bc98e248e4e06823 *man/MultHomog.Rd b21bfdfc325cb5d59f52b544ce8b3895 *man/Symm.Rd db0d8f974ec0b38432efd34acc97e010 *man/Topo.Rd 5ca24d22d50cb807d73839ffef9fa2c0 *man/anova.gnm.Rd 3709acc93ded6573df4aff6e7ac299b1 *man/asGnm.Rd 8006f41596e0d8de7b3dc6f6edee539b *man/backPain.Rd 854f6de396cf1435127151a005fd4d7b *man/barley.Rd 1847ccfd274aefeb23138f23da285af0 *man/barleyHeights.Rd f1a7a3da20cfc54cba892c3efc96f530 *man/cautres.Rd e46716e760d814ec8c984d473727a612 *man/checkEstimable.Rd 257ea8bf7503a245bb8645e0ba4ec28d *man/confint.gnm.Rd ef625b756b5fe3f6afcd3097801bc7f8 *man/erikson.Rd 5efe06dbc64e8e8eeab5ae5260a6c7fa *man/exitInfo.Rd fbb83ca880d5b19176f31bd8dd3a6533 *man/expandCategorical.Rd 761ca574a0deec5c8d07683806029efa *man/friend.Rd a17df861fced8a5d29e31e42e5c307d6 *man/getContrasts.Rd 5db92406797fbf7005362f638dd72983 *man/gnm-defunct.Rd 208e6ad06c53ef31dc053af2dcb55031 *man/gnm-package.Rd dfa1ec5e22574631c4dccf09430d508d *man/gnm.Rd ed561d469b59eb78fee8cfc94423c5b4 *man/instances.Rd 5875a3c18e4d92e9499a72c4b7443c18 *man/meanResiduals.Rd 7f020321e45f8862fbff8f8db391d263 *man/mentalHealth.Rd fae4fd1c5b87ac311be846bd0818c1e1 *man/model.matrix.gnm.Rd ad5b6dd7b9d4dd661661923a2f417f11 *man/nonlin.function.Rd db68c7ef8303a1271b07fb418f22f924 *man/ofInterest.Rd 16e981f6532b4ef22679cd20b6c83b58 *man/parameters.Rd 8da13b4f7b46f1e0e8c2d7ed44ac81fd *man/pickCoef.Rd dfcffdf6846ac9ca365a9e35dd676cb2 *man/plot.gnm.Rd 9e07000e50faedda6d0828cb889a35f2 *man/predict.gnm.Rd b3be0e8d4776bdbdbe754a8c5b07ce31 *man/profile.gnm.Rd 90144720cb986b40e1e35096ef02c176 *man/residSVD.Rd 45c87642034e559ee9fa7dcc7ab2a598 *man/se.Rd 56b06f50038ae7d971d25d7da2863b98 *man/summary.gnm.Rd badbc1338de4cc070474541ba8db74cb *man/termPredictors.Rd e5b04d40dd15a8b7eabebdbb3f89f5df *man/vcov.gnm.Rd e2b7330e553d7fa198f4ee866547e047 *man/voting.Rd 86e2c54eede728c71f53fffe588b4747 *man/wedderburn.Rd 45f6173b89f29140a32b451db068a05f *man/wheat.Rd 382958f7c88a849929c32db5d54cfc92 *man/yaish.Rd 2fa4c7011c2bc0f7449ae151d5cc44ae *src/Makevars f753abc9263a88a84d9d8eedf26539cb *src/gnm.c b92e86ec2f01009050df417383270ef6 *tests/RC.R 0589e813e77d626804ae98597ea44c47 *tests/RC.Rout.save faf03c52907dea1838c30ddce2b0d2e3 *tests/RChomog.R 7965b49f82c7759d175235cc80e1d48d *tests/RChomog.Rout.save 17c484a0c636b36f59edd112aeae85a5 *tests/biplot.R 0147d61b8d1f158038cc3bdc34334172 *tests/biplot.Rout.save 2280df7fc46271c138423b133ff5449e *tests/bwt.R 9cbb5fc90f1685c460ca65b0fe359114 *tests/bwt.Rout.save ad8debb998906face4985522aa5c8908 *tests/diagonalRef.R bf5154bd4e82a56162474086fa7a17e1 *tests/diagonalRef.Rout.save 345f20b9559c77d318a31e04dd7ca078 *tests/doubleUnidiff.R e838ff265ee9ea12990ca6943fceaa3e *tests/doubleUnidiff.Rout.save f8b8a561028ba76d1345b2fdb87a9cb9 *tests/gammi.R 9766951409b2fcf5d3e4a65ed93c42ea *tests/gammi.Rout.save bd79a07aee68af4417c254182544a98e *tests/logexcess.R 338a6e857242b151e71a8dc5d39af3b7 *tests/logexcess.Rout.save f814163cf4c30fad2d284a018dccd06a *tests/logistic.R af8dd43de08dd55a455ea3396ce35535 *tests/logistic.Rout.save d4641f76c5ea09c63a1aa20828b7517b *tests/nls.R e32a611401ddfd56f36dbd5d82acc8db *tests/nls.Rout.save 9e7211b8ffd69b16266bbc7c504cfdb5 *tests/stereotype.R 27a490d8983ae6def3d38568ae3640ec *tests/stereotype.Rout.save 4dc08204f1123841fa4fefa1ed8fdd77 *tests/unidiff.R a56a88e91859c8d20b4acc25e868936e *tests/unidiff.Rout.save 8501b0386c1efafa0a4f934f6388da37 *vignettes/fig-Effect_plots.pdf fd726d14883e8a94f45664a8ce062803 *vignettes/fig-LCall.pdf 596d106d3518e1cb77dbe23745d2f32a *vignettes/fig-LCover45.pdf e77867e5284795f1ed6d089899067d9c *vignettes/fig-LCqvplot.pdf ed765957742f3d6e102227a3ae699b04 *vignettes/fig-deaths1921-1940.pdf 39c2eaa3b5904faca6c419c0cab7b70e *vignettes/fig-profilePlot.pdf d39bcd84e80e26b411e58a29a7ebf113 *vignettes/gnm.bib 1fcfca4a717ed1f39f6adeaf1e29fd50 *vignettes/gnmOverview.Rnw f7bb932423c663aba2113c8ebb840cdc *vignettes/screenshot1.png 49e2d977c91fc67eceac382b56480625 *vignettes/screenshot2.png 91bb1eb0a9bef5dae03f658159ad9ac1 *vignettes/screenshot3.png gnm/build/0000755000176000001440000000000012515664670012176 5ustar ripleyusersgnm/build/vignette.rds0000644000176000001440000000040512515664670014534 0ustar ripleyusers‹}QËnÂ0tЂ„@âÚÃ~A> 7$DU Š8pµð†ZuÖ‘µ§~ya "®TK»~ììxÆÞ…¡ˆBÎ/£9§!ÇŒ#±ñ<=P¾>¢;j¬“”j¯¼xAB'þFdÉhBé · M š }†mÀfP} 0)rÿ)øß}…ʼò¤'‡…Ï/Å+ˆGÞ$s,=†‡%Hª9þõú;ÈÓ½©÷›©·‹©ëÁÉëy|ůںŽ·‡¶˜x¥ vÚ¶ººm¢ÍrÕ.ƒÔ×ù‡¿çqält>'͇þp:ñðcodé?ÆXÉJ&™ãþF÷—š4ognm/DESCRIPTION0000644000176000001440000000254212515671462012605 0ustar ripleyusersPackage: gnm Title: Generalized Nonlinear Models Version: 1.0-8 Date: 2015-04-21 Authors@R: c(person("Heather", "Turner", role = c("aut", "cre"), email = "ht@heatherturner.net"), person("David", "Firth", role = "aut"), person("Brian", "Ripley", role = "ctb"), person("Bill", "Venables", role = "ctb"), person(c("Douglas", "M."), "Bates", role = "ctb"), person("Martin", "Maechler", role = "ctb")) Description: Functions to specify and fit generalized nonlinear models, including models with multiplicative interaction terms such as the UNIDIFF model from sociology and the AMMI model from crop science, and many others. Over-parameterized representations of models are used throughout; functions are provided for inference on estimable parameter combinations, as well as standard methods for diagnostics etc. URL: http://go.warwick.ac.uk/gnm License: GPL-2 | GPL-3 Depends: R (>= 2.3.0) Imports: MASS, stats, graphics, Matrix, nnet, qvcalc (>= 0.8-3), relimp Suggests: vcdExtra LazyData: yes NeedsCompilation: yes Packaged: 2015-04-22 09:28:24 UTC; hturner Author: Heather Turner [aut, cre], David Firth [aut], Brian Ripley [ctb], Bill Venables [ctb], Douglas M. Bates [ctb], Martin Maechler [ctb] Maintainer: Heather Turner Repository: CRAN Date/Publication: 2015-04-22 12:08:50 gnm/man/0000755000176000001440000000000012515147523011643 5ustar ripleyusersgnm/man/Symm.Rd0000644000176000001440000000122512515147523013057 0ustar ripleyusers\name{Symm} \alias{Symm} \title{ Symmetric Interaction of Factors } \description{ \code{Symm} codes the symmetric interaction of factors having the same set of levels, for use in regression models of symmetry or quasi-symmetry. } \usage{ Symm(...) } \arguments{ \item{\dots}{ one or more factors. } } \value{ A factor whose levels index the symmetric interaction of all factors supplied as input. } \note{\code{Symm} relies on the gtools package from CRAN} \author{ David Firth and Heather Turner } \seealso{ \code{\link{Diag}}} \examples{ row <- gl(4, 4, 16) col <- gl(4, 1, 16) symm4by4 <- Symm(row, col) matrix(symm4by4, 4, 4) } \keyword{ models } gnm/man/Const.Rd0000644000176000001440000000206612515147523013224 0ustar ripleyusers\name{Const} \alias{Const} \title{ Specify a Constant in a "nonlin" Function Predictor } \description{ A symbolic wrapper to specify a constant in the predictor of a \code{"nonlin"} function. } \usage{ Const(const) } \arguments{ \item{const}{ a numeric value. } } \value{ A call to \code{rep} used to create a variable representing the constant in the model frame. } \note{ \code{Const} may only be used in the predictor of a \code{"nonlin"} function. Use \code{offset} to specify a constant in the model formula. } \author{ Heather Turner } \seealso{\code{\link{gnm}}, \code{\link{formula}}, \code{\link{offset}}} \examples{ ## One way to fit the logistic function without conditional ## linearity as in ?nls library(gnm) set.seed(1) DNase1 <- subset(DNase, Run == 1) test <- gnm(density ~ -1 + Mult(1, Inv(Const(1) + Exp(Mult(1 + offset(-log(conc)), Inv(1))))), start = c(NA, 0, 1), data = DNase1, trace = TRUE) coef(test) } \keyword{ models } \keyword{ regression } \keyword{ nonlinear } gnm/man/barleyHeights.Rd0000644000176000001440000000326712515147523014734 0ustar ripleyusers\name{barleyHeights} \alias{barleyHeights} \docType{data} \title{ Heights of Barley Plants } \description{ Average heights for 15 genotypes of barley recorded over 9 years. } \usage{barleyHeights} \format{ A data frame with 135 observations on the following 3 variables. \describe{ \item{\code{height}}{average height over 4 replicates (cm)} \item{\code{year}}{a factor with 9 levels \code{1974} to \code{1982}} \item{\code{genotype}}{a factor with 15 levels \code{1:15}} } } \source{ Aastveit, A. H. \& Martens, H. (1986). ANOVA interactions interpreted by partial least squares regression. \emph{Biometrics}, \bold{42}, 829--844. } \references{ Chadoeuf, J \& Denis, J B (1991). Asymptotic variances for the multiplicative interaction model. \emph{J. App. Stat.} \bold{18(3)}, 331--353. } \examples{ set.seed(1) ## Fit AMMI-1 model barleyModel <- gnm(height ~ year + genotype + Mult(year, genotype), data = barleyHeights) ## Get row and column scores with se's gamma <- getContrasts(barleyModel, pickCoef(barleyModel, "[.]y"), ref = "mean", scaleWeights = "unit") delta <- getContrasts(barleyModel, pickCoef(barleyModel, "[.]g"), ref = "mean", scaleWeights = "unit") ## Corresponding CI's similar to Chadoeuf & Denis (1991) Table 8 ## (allowing for change in sign) gamma[[2]][,1] + (gamma[[2]][,2]) \%o\% c(-1.96, 1.96) delta[[2]][,1] + (delta[[2]][,2]) \%o\% c(-1.96, 1.96) ## Multiplier of row and column scores height <- matrix(scale(barleyHeights$height, scale = FALSE), 15, 9) R <- height - outer(rowMeans(height), colMeans(height), "+") svd(R)$d[1] } \keyword{datasets} gnm/man/House2001.Rd0000644000176000001440000001351112515147523013521 0ustar ripleyusers\name{House2001} \alias{House2001} \docType{data} \title{ Data on twenty roll calls in the US House of Representatives, 2001 } \description{ The voting record of every representative in the 2001 House, on 20 roll calls selected by \emph{Americans for Democratic Action}. Each row is the record of one representative; the first column records the representative's registered party allegiance. } \usage{House2001} \format{ A data frame with 439 observations on the following 21 variables. \describe{ \item{\code{party}}{a factor with levels \code{D} \code{I} \code{N} \code{R}} \item{\code{HR333.BankruptcyOverhaul.Yes}}{a numeric vector} \item{\code{SJRes6.ErgonomicsRuleDisapproval.No}}{a numeric vector} \item{\code{HR3.IncomeTaxReduction.No}}{a numeric vector} \item{\code{HR6.MarriageTaxReduction.Yes}}{a numeric vector} \item{\code{HR8.EstateTaxRelief.Yes}}{a numeric vector} \item{\code{HR503.FetalProtection.No}}{a numeric vector} \item{\code{HR1.SchoolVouchers.No}}{a numeric vector} \item{\code{HR1836.TaxCutReconciliationBill.No}}{a numeric vector} \item{\code{HR2356.CampaignFinanceReform.No}}{a numeric vector} \item{\code{HJRes36.FlagDesecration.No}}{a numeric vector} \item{\code{HR7.FaithBasedInitiative.Yes}}{a numeric vector} \item{\code{HJRes50.ChinaNormalizedTradeRelations.Yes}}{a numeric vector} \item{\code{HR4.ANWRDrillingBan.Yes}}{a numeric vector} \item{\code{HR2563.PatientsRightsHMOLiability.No}}{a numeric vector} \item{\code{HR2563.PatientsBillOfRights.No}}{a numeric vector} \item{\code{HR2944.DomesticPartnerBenefits.No}}{a numeric vector} \item{\code{HR2586.USMilitaryPersonnelOverseasAbortions.Yes}}{a numeric vector} \item{\code{HR2975.AntiTerrorismAuthority.No}}{a numeric vector} \item{\code{HR3090.EconomicStimulus.No}}{a numeric vector} \item{\code{HR3000.TradePromotionAuthorityFastTrack.No}}{a numeric vector} } } \details{ Coding of the votes is as described in ADA (2002). } \source{ Originally printed in ADA (2002). Kindly supplied in electronic format by Jan deLeeuw, who used the data to illustrate methods developed in deLeeuw (2006). } \references{ Americans for Democratic Action, ADA (2002). 2001 voting record: Shattered promise of liberal progress. \emph{ADA Today} \bold{57}(1), 1--17. deLeeuw, J (2006). Principal component analysis of binary data by iterated singular value decomposition. \emph{Computational Statistics and Data Analysis} \bold{50}, 21--39. } \examples{ \dontrun{ ## This example takes some time to run! summary(House2001) ## Put the votes in a matrix, and discard members with too many NAs etc: House2001m <- as.matrix(House2001[-1]) informative <- apply(House2001m, 1, function(row){ valid <- !is.na(row) validSum <- if (any(valid)) sum(row[valid]) else 0 nValid <- sum(valid) uninformative <- (validSum == nValid) || (validSum == 0) || (nValid < 10) !uninformative}) House2001m <- House2001m[informative, ] ## Make a vector of colours, blue for Republican and red for Democrat: parties <- House2001$party[informative] partyColors <- rep("black", length(parties)) partyColors <- ifelse(parties == "D", "red", partyColors) partyColors <- ifelse(parties == "R", "blue", partyColors) ## Expand the data for statistical modelling: House2001v <- as.vector(House2001m) House2001f <- data.frame(member = rownames(House2001m), party = parties, rollCall = factor(rep((1:20), rep(nrow(House2001m), 20))), vote = House2001v) ## Now fit an "empty" model, in which all members vote identically: baseModel <- glm(vote ~ -1 + rollCall, family = binomial, data = House2001f) ## From this, get starting values for a one-dimensional multiplicative term: Start <- residSVD(baseModel, rollCall, member) ## ## Now fit the logistic model with one multiplicative term. ## For the response variable, instead of vote=0,1 we use 0.03 and 0.97, ## corresponding approximately to a bias-reducing adjustment of p/(2n), ## where p is the number of parameters and n the number of observations. ## voteAdj <- 0.5 + 0.94*(House2001f$vote - 0.5) House2001model1 <- gnm(voteAdj ~ Mult(rollCall, member), eliminate = rollCall, family = binomial, data = House2001f, na.action = na.exclude, trace = TRUE, tolerance = 1e-03, start = -Start) ## Deviance is 2234.847, df = 5574 ## ## Plot the members' positions as estimated in the model: ## memberParameters <- pickCoef(House2001model1, "member") plot(coef(House2001model1)[memberParameters], col = partyColors, xlab = "Alphabetical index (Abercrombie 1 to Young 301)", ylab = "Member's relative position, one-dimensional model") ## Can do the same thing with two dimensions, but gnm takes around 40 ## slow iterations to converge (there are more than 600 parameters): Start2 <- residSVD(baseModel, rollCall, member, d = 2) House2001model2 <- gnm( voteAdj ~ instances(Mult(rollCall - 1, member - 1), 2), eliminate = rollCall, family = binomial, data = House2001f, na.action = na.exclude, trace = TRUE, tolerance = 1e-03, start = Start2, lsMethod = "qr") ## Deviance is 1545.166, df = 5257 ## memberParameters1 <- pickCoef(House2001model2, "1).member") memberParameters2 <- pickCoef(House2001model2, "2).member") plot(coef(House2001model2)[memberParameters1], coef(House2001model2)[memberParameters2], col = partyColors, xlab = "Dimension 1", ylab = "Dimension 2", main = "House2001 data: Member positions, 2-dimensional model") ## ## The second dimension is mainly due to rollCall 12, which does not ## correlate well with the rest -- look at the coefficients of ## House2001model1, or at the 12th row of cormat <- cor(na.omit(House2001m)) } } \keyword{datasets} gnm/man/pickCoef.Rd0000644000176000001440000000555712515147523013671 0ustar ripleyusers\name{pickCoef} \alias{pickCoef} \title{ Get Indices or Values of Selected Model Coefficients } \description{ Get the indices or values of a subset of non-eliminated coefficients selected via a Tk dialog or by pattern matching. } \usage{ pickCoef(object, pattern = NULL, value = FALSE, ...) } \arguments{ \item{object}{ a model object. } \item{pattern}{ character string containing a regular expression or (with \code{fixed = TRUE}) a pattern to be matched exactly. If missing, a Tk dialog will open for coefficient selection. } \item{value}{ if \code{FALSE}, a named vector of indices, otherwise the value of the selected coefficients. } \item{\dots}{ arguments to pass on to \link[relimp]{pickFrom} if \code{pattern} is missing, otherwise \code{grep}. In particular, \code{fixed = TRUE} specifies that \code{pattern} is a string to be matched as is.} } \value{ If \code{value = FALSE} (the default), a named vector of indices, otherwise the values of the selected coefficients. If no coefficients are selected the returned value will be \code{NULL}. } \author{ Heather Turner } \seealso{ \code{\link{regexp}}, \code{\link{grep}}, \code{\link[relimp]{pickFrom}}, \code{\link{ofInterest}}} \examples{ set.seed(1) ### Extract indices for use with ofInterest ## fit the "UNIDIFF" mobility model across education levels unidiff <- gnm(Freq ~ educ*orig + educ*dest + Mult(Exp(educ), orig:dest), family = poisson, data = yaish, subset = (dest != 7)) ## set coefficients in first constituent multiplier as 'ofInterest' ## using regular expression ofInterest(unidiff) <- pickCoef(unidiff, "[.]educ") ## summarise model, only showing coefficients of interest summary(unidiff) ## get contrasts of these coefficients getContrasts(unidiff, ofInterest(unidiff)) ### Extract coefficients to use as starting values ## fit diagonal reference model with constant weights set.seed(1) ## reconstruct counts voting Labour/non-Labour count <- with(voting, percentage/100 * total) yvar <- cbind(count, voting$total - count) classMobility <- gnm(yvar ~ -1 + Dref(origin, destination), family = binomial, data = voting) ## create factors indicating movement in and out of salariat (class 1) upward <- with(voting, origin != 1 & destination == 1) downward <- with(voting, origin == 1 & destination != 1) ## extract diagonal effects from first model to use as starting values diagCoef <- pickCoef(classMobility, "Dref(., .)", fixed = TRUE, value = TRUE) ## fit separate weights for the "socially mobile" groups ## -- there are now 3 parameters for each weight socialMobility <- gnm(yvar ~ -1 + Dref(origin, destination, delta = ~ 1 + downward + upward), family = binomial, data = voting, start = c(rep(NA, 6), diagCoef)) } \keyword{ models } gnm/man/termPredictors.Rd0000644000176000001440000000314612515147523015144 0ustar ripleyusers\name{termPredictors} \alias{termPredictors} \title{ Extract Term Contributions to Predictor } \description{ \code{termPredictors} is a generic function which extracts the contribution of each term to the predictor from a fitted model object. } \usage{ termPredictors(object, ...) } \arguments{ \item{object}{ a fitted model object. } \item{\dots}{ additional arguments for method functions. } } \details{ The default method assumes that the predictor is linear and calculates the contribution of each term from the model matrix and fitted coefficients. A method is also available for \code{\link{gnm}} objects. } \value{ A matrix with the additive components of the predictor in labelled columns. } \author{ Heather Turner } \seealso{ \code{\link{gnm}} } \examples{ ## Linear model G <- gl(4, 6) x <- 1:24 y <- rnorm(24, 0, 1) lmGx <- lm(y ~ G + x) contrib <- termPredictors(lmGx) contrib all.equal(as.numeric(rowSums(contrib)), as.numeric(lmGx$fitted)) #TRUE ## Generalized linear model y <- cbind(rbinom(24, 10, 0.5), rep(10, 24)) glmGx <- glm(y ~ G + x, family = binomial) contrib <- termPredictors(glmGx) contrib all.equal(as.numeric(rowSums(contrib)), as.numeric(glmGx$linear.predictors)) #TRUE ## Generalized nonlinear model A <- gl(4, 6) B <- gl(6, 1, 24) y <- cbind(rbinom(24, 10, 0.5), rep(10, 24)) set.seed(1) gnmAB <- gnm(y ~ A + B + Mult(A, B), family = binomial) contrib <- termPredictors(gnmAB) contrib all.equal(as.numeric(rowSums(contrib)), as.numeric(gnmAB$predictors)) #TRUE } \keyword{ models } \keyword{ regression } gnm/man/confint.gnm.Rd0000644000176000001440000001070212515147523014352 0ustar ripleyusers\name{confint.gnm} \alias{confint.gnm} \alias{confint.profile.gnm} \title{ Compute Confidence Intervals of Parameters in a Generalized Nonlinear Model } \description{ Computes confidence intervals for one or more parameters in a generalized nonlinear model, based on the profiled deviance. } \usage{ \method{confint}{gnm}(object, parm = ofInterest(object), level = 0.95, trace = FALSE, ...) \method{confint}{profile.gnm}(object, parm = names(object), level = 0.95, ...) } \arguments{ \item{object}{ an object of class \code{"gnm"} or \code{"profile.gnm"}} \item{parm}{ (optional) either a numeric vector of indices or a character vector of names, specifying the parameters for which confidence intervals are to be estimated. If \code{parm} is missing, confidence intervals are found for all parameters.} \item{level}{ the confidence level required. } \item{trace}{ a logical value indicating whether profiling should be traced. } \item{\dots}{ arguments passed to or from other methods } } \details{ These are methods for the generic function \code{confint} in the \code{base} package. For \code{"gnm"} objects, \code{profile.gnm} is first called to profile the deviance over each parameter specified by \code{parm}, or over all parameters in the model if \code{parm} is missing. The method for \code{"profile.gnm"} objects is then called, which interpolates the deviance profiles to estimate the limits of the confidence interval for each parameter, see \code{\link{profile.gnm}} for more details. If a \code{"profile.gnm"} object is passed directly to \code{confint}, parametrs specified by \code{parm} must be a subset of the profiled parameters. For unidentified parameters a confidence interval cannot be calculated and the limits will be returned as \code{NA}. If the deviance curve has an asymptote and a limit of the confidence interval cannot be reached, the limit will be returned as \code{-Inf} or \code{Inf} appropriately. If the range of the profile does not extend far enough to estimate a limit of the confidence interval, the limit will be returned as \code{NA}. In such cases, it may be desirable create a profile object directly, see \code{\link{profile.gnm}} for more details. } \value{ A matrix (or vector) with columns giving lower and upper confidence limits for each parameter. These will be labelled as (1-level)/2 and 1 - (1-level)/2 in \% (by default 2.5\% and 97.5\%). } \author{ Modification of \code{\link[MASS]{confint.glm}} by W. N. Venables and B. D. Ripley. Adapted for \code{"gnm"} objects by Heather Turner. } \seealso{ \code{\link{profile.gnm}}, \code{\link{gnm}}, \code{\link[MASS]{profile.glm}}} \examples{ ### Example in which profiling doesn't take too long count <- with(voting, percentage/100 * total) yvar <- cbind(count, voting$total - count) classMobility <- gnm(yvar ~ -1 + Dref(origin, destination), constrain = "delta1", family = binomial, data = voting) ## profile diagonal effects confint(classMobility, parm = 3:7, trace = TRUE) \dontrun{ ### Profiling takes much longer here, but example more interesting! unidiff <- gnm(Freq ~ educ*orig + educ*dest + Mult(Exp(educ), orig:dest), ofInterest = "[.]educ", constrain = "[.]educ1", family = poisson, data = yaish, subset = (dest != 7)) ## Letting 'confint' compute profile confint(unidiff, trace = TRUE) ## 2.5 \% 97.5 \% ## Mult(Exp(.), orig:dest).educ1 NA NA ## Mult(Exp(.), orig:dest).educ2 -0.5978901 0.1022447 ## Mult(Exp(.), orig:dest).educ3 -1.4836854 -0.2362378 ## Mult(Exp(.), orig:dest).educ4 -2.5792398 -0.2953420 ## Mult(Exp(.), orig:dest).educ5 -Inf -0.7007616 ## Creating profile object first with user-specified stepsize prof <- profile(unidiff, trace = TRUE, stepsize = 0.1) confint(prof, ofInterest(unidiff)[2:5]) ## 2.5 \% 97.5 \% ## Mult(Exp(.), orig:dest).educ2 -0.5978324 0.1022441 ## Mult(Exp(.), orig:dest).educ3 -1.4834753 -0.2362138 ## Mult(Exp(.), orig:dest).educ4 NA -0.2950790 ## Mult(Exp(.), orig:dest).educ5 NA NA ## For 95\% confidence interval, need to estimate parameters for which ## z = +/- 1.96. Profile has not gone far enough for last two parameters range(prof[[4]]$z) ## -1.566601 2.408650 range(prof[[5]]$z) ## -0.5751376 1.1989487 } } \keyword{ models } \keyword{ nonlinear } gnm/man/expandCategorical.Rd0000644000176000001440000000745012515147523015555 0ustar ripleyusers\name{expandCategorical} \alias{expandCategorical} \title{ Expand Data Frame by Re-expressing Categorical Data as Counts } \description{ Expands the rows of a data frame by re-expressing observations of a categorical variable specified by \code{catvar}, such that the column(s) corresponding to \code{catvar} are replaced by a factor specifying the possible categories for each observation and a vector of 0/1 counts over these categories. %Expands the rows of a data frame containing a categorical variable %\code{catvar} with \eqn{c} possible categories, such that each %observation of \code{catvar} is represented by \eqn{c} 0/1 counts and %all other variables are replicated appropriately. } \usage{ expandCategorical(data, catvar, sep = ".", countvar = "count", idvar = "id", as.ordered = FALSE, group = TRUE) } \arguments{ \item{data}{ a data frame. } \item{catvar}{ a character vector specifying factors in \code{data} whose interaction will form the basis of the expansion. } \item{sep}{ a character string used to separate the concatenated values of \code{catvar} in the name of the new interaction factor. } \item{countvar}{ (optional) a character string to be used for the name of the new count variable. } \item{idvar}{ (optional) a character string to be used for the name of the new factor identifying the original rows (cases). } \item{as.ordered}{ logical - whether the new interaction factor should be of class \code{"ordered"}.} \item{group}{logical: whether or not to group individuals with common values over all covariates. } } \details{ Each row of the data frame is replicated \eqn{c} times, where \eqn{c} is the number of levels of the interaction of the factors specified by \code{catvar}. In the expanded data frame, the columns specified by \code{catvar} are replaced by a factor specifying the \eqn{r} possible categories for each case, named by the concatenated values of \code{catvar} separated by \code{sep}. The ordering of factor levels will be preserved in the creation of the new factor, but this factor will not be of class \code{"ordered"} unless the argument \code{as.ordered = TRUE}. A variable with name \code{countvar} is added to the data frame which is equal to 1 for the observed category in each case and 0 elsewhere. Finally a factor with name \code{idvar} is added to index the cases. } \value{ The expanded data frame as described in Details. } \author{ Heather Turner } \note{ Re-expressing categorical data in this way allows a multinomial response to be modelled as a poisson response, see examples. } \seealso{ \code{\link{gnm}}, \code{\link[nnet]{multinom}}, \code{\link{reshape}}, \code{\link[catspec:mcl]{mclgen}} } \references{ Anderson, J. A. (1984) Regression and Ordered Categorical Variables. \emph{J. R. Statist. Soc. B}, \bold{46(1)}, 1-30. } \examples{ ### Example from help(multinom, package = "nnet") library(MASS) example(birthwt) library(nnet) bwt.mu <- multinom(low ~ ., data = bwt) ## Equivalent using gnm - include unestimable main effects in model so ## that interactions with low0 automatically set to zero, else could use ## 'constrain' argument. bwtLong <- expandCategorical(bwt, "low", group = FALSE) bwt.po <- gnm(count ~ low*(. - id), eliminate = id, data = bwtLong, family = "poisson") summary(bwt.po) # same deviance; df reflect extra id parameters ### Example from ?backPain set.seed(1) summary(backPain) backPainLong <- expandCategorical(backPain, "pain") ## Fit models described in Table 5 of Anderson (1984) noRelationship <- gnm(count ~ pain, eliminate = id, family = "poisson", data = backPainLong) oneDimensional <- update(noRelationship, ~ . + Mult(pain, x1 + x2 + x3)) } \keyword{ manip } \keyword{ models } gnm/man/summary.gnm.Rd0000755000176000001440000001261412515147523014416 0ustar ripleyusers\name{summary.gnm} \alias{summary.gnm} \alias{print.summary.gnm} \title{ Summarize Generalized Nonlinear Model Fits } \description{ \code{summary} method for objects of class \code{"gnm"} } \usage{ \method{summary}{gnm}(object, dispersion = NULL, correlation = FALSE, symbolic.cor = FALSE, with.eliminate = FALSE, ...) \method{print}{summary.gnm}(x, digits = max(3, getOption("digits") - 3), signif.stars = getOption("show.signif.stars"), symbolic.cor = x$symbolic.cor, ...) } \arguments{ \item{object}{ an object of class \code{"gnm"}. } \item{x}{ an object of class \code{"summary.gnm"}. } \item{dispersion}{ the dispersion parameter for the fitting family. By default it is obtained from \code{object}. } \item{correlation}{ logical: if \code{TRUE}, the correlation matrix of the estimated parameters is returned. } \item{digits}{ the number of siginificant digits to use when printing. } \item{symbolic.cor}{ logical: if \code{TRUE}, the correlations are printed in a symbolic form rather than numbers (see \code{symnum}). } \item{signif.stars}{ logical. If \code{TRUE}, "significance stars" are printed for each coefficient. } \item{with.eliminate}{ Logical. If \code{TRUE}, any eliminated coefficients are included in the summary. } \item{\dots}{ further arguments passed to or from other methods. } } \details{ \code{print.summary.gnm} prints the original call to \code{gnm}; a summary of the deviance residuals from the model fit; the coefficients of the model; the residual deviance; the Akaike's Information Criterion value, and the number of main iterations performed. Standard errors, z-values and p-values are printed alongside the coefficients, with "significance stars" if \code{signif.stars} is \code{TRUE}. When the \code{"summary.gnm"} object has a \code{"correlation"} component, the lower triangle of this matrix is also printed, to two decimal places (or symbolically); to see the full matrix of correlations print \code{summary(object, correlation = TRUE)$correlation} directly. The standard errors returned by \code{summary.gnm} are scaled by \code{sqrt(dispersion)}. If the dispersion is not specified, it is taken as \code{1} for the \code{binomial} and \code{Poisson} families, and otherwise estimated by the residual Chi-squared statistic divided by the residual degrees of freedom. For coefficients that have been constrained or are not estimable, the standard error is returned as \code{NA}. } \value{ \code{summary.gnm} returns an object of class \code{"summary.gnm"}, which is a list with components \item{call }{ the \code{"call"} component from object. } \item{ofInterest }{ the \code{"ofInterest"} component from object. } \item{family }{ the \code{"family"} component from object. } \item{deviance }{ the \code{"deviance"} component from object. } \item{aic }{ the \code{"aic"} component from object. } \item{df.residual }{ the \code{"df.residual"} component from object. } \item{iter }{ the \code{"iter"} component from object. } \item{deviance.resid }{ the deviance residuals, see \code{\link{residuals.glm}}. } \item{coefficients }{ the matrix of coefficients, standard errors, z-values and p-values. } \item{elim.coefs }{ if \code{with.eliminate = TRUE} a matrix of eliminated coefficients, standard errors, z-values and p-values. } \item{dispersion }{ either the supplied argument or the estimated dispersion if the latter is \code{NULL}. } \item{df}{ a 3-vector of the rank of the model; the number of residual degrees of freedom, and number of unconstrained coefficients. } \item{cov.scaled }{ the estimated covariance matrix scaled by \code{dispersion} (see \code{\link{vcov.gnm}} for more details). } \item{correlation }{ (only if \code{correlation} is \code{TRUE}) the estimated correlations of the estimated coefficients. } \item{symbolic.cor }{ (only if \code{correlation} is \code{TRUE}) the value of the argument \code{symbolic.cor}. } } \note{ The \code{gnm} class includes generalized linear models, and it should be noted that \code{summary.gnm} differs from \code{\link{summary.glm}} in that it does not omit coefficients which are \code{NA} from the objects it returns. (Such coefficients are \code{NA} since they have been fixed at \code{0} either by use of the \code{constrain} argument to \code{gnm} or by a convention to handle linear aliasing). } \author{ Modification of \code{\link{summary.glm}} by the R Core Team. Adapted for \code{"gnm"} objects by Heather Turner. } \seealso{ \code{\link{gnm}}, \code{\link{summary}}} \examples{ ### First example from ?Dref set.seed(1) ## reconstruct counts voting Labour/non-Labour count <- with(voting, percentage/100 * total) yvar <- cbind(count, voting$total - count) ## fit diagonal reference model with constant weights classMobility <- gnm(yvar ~ -1 + Dref(origin, destination), family = binomial, data = voting) ## summarize results - note diagonal weights are over-parameterised summary(classMobility) ## refit setting first weight to zero (as DrefWeights() does) classMobility <- gnm(yvar ~ -1 + Dref(origin, destination), family = binomial, data = voting, constrain = "delta1") summary(classMobility) } \keyword{ models } \keyword{ regression } \keyword{ nonlinear } gnm/man/anova.gnm.Rd0000644000176000001440000000655612515147523014032 0ustar ripleyusers\name{anova.gnm} \alias{anova.gnm} \title{ Analysis of Deviance for Generalized Nonlinear Models } \description{ Compute an analysis of deviance table for one or more generalized nonlinear models } \usage{ \method{anova}{gnm}(object, ..., dispersion = NULL, test = NULL) } \arguments{ \item{object}{ an object of class \code{gnm} } \item{\dots}{ additional objects of class \code{gnm} or \code{glm}} \item{dispersion}{ the dispersion parameter for the fitting family. By default it is derived from \code{object} } \item{test}{ (optional) a character string, (partially) matching one of \code{"Chisq"}, \code{"F"}, or \code{"Cp"}. See \code{\link{stat.anova}}. } } \details{ Specifying a single object gives a sequential analysis of deviance table for that fit. The rows of the table show the reduction in the residual deviance and the current residual deviance as each term in the formula is added in turn. If more than one object is specified, the rows of the table show the residual deviance of the current model and the change in the residual deviance from the previous model. (This only makes statistical sense if the models are nested.) It is conventional to list the models from smallest to largest, but this is up to the user. If \code{test} is specified, the table will include test statistics and/or p values for the reduction in deviance. For models with known dispersion (e.g., binomial and Poisson fits) the chi-squared test is most appropriate, and for those with dispersion estimated by moments (e.g., 'gaussian', 'quasibinomial' and 'quasipoisson' fits) the F test is most appropriate. Mallows' Cp statistic is the residual deviance plus twice the estimate of \eqn{\sigma^2}{sigma^2} times the residual degrees of freedom, which is closely related to AIC (and a multiple of it if the dispersion is known). } \value{ An object of class \code{"anova"} inheriting from class \code{"data.frame"}. } \author{ Modification of \code{\link{anova.glm}} by the R Core Team. Adapted for \code{"gnm"} objects by Heather Turner. } \section{Warning }{ The comparison between two or more models will only be valid if they are fitted to the same dataset. This may be a problem if there are missing values and R's default of \code{na.action = na.omit} is used; an error will be given in this case. } \seealso{ \code{\link{gnm}}, \code{\link{anova}}} \examples{ set.seed(1) ## Fit a uniform association model separating diagonal effects Rscore <- scale(as.numeric(row(occupationalStatus)), scale = FALSE) Cscore <- scale(as.numeric(col(occupationalStatus)), scale = FALSE) Uniform <- glm(Freq ~ origin + destination + Diag(origin, destination) + Rscore:Cscore, family = poisson, data = occupationalStatus) ## Fit an association model with homogeneous row-column effects RChomog <- gnm(Freq ~ origin + destination + Diag(origin, destination) + MultHomog(origin, destination), family = poisson, data = occupationalStatus) ## Fit an association model with separate row and column effects RC <- gnm(Freq ~ origin + destination + Diag(origin, destination) + Mult(origin, destination), family = poisson, data = occupationalStatus) anova(RC, test = "Chisq") anova(Uniform, RChomog, RC, test = "Chisq") } \keyword{ models } gnm/man/Exp.Rd0000644000176000001440000000317012515147523012667 0ustar ripleyusers\name{Exp} \alias{Exp} \title{ Specify the Exponential of a Predictor in a gnm Model Formula } \description{ A function of class \code{"nonlin"} to specify the exponential of a predictor in the formula argument to \code{\link{gnm}}. } \usage{ Exp(expression, inst = NULL) } \arguments{ \item{expression}{ a symbolic expression representing the (possibly nonlinear) predictor. } \item{inst}{ (optional) an integer specifying the instance number of the term. } } \details{ The \code{expression} argument is interpreted as the right hand side of a formula in an object of class \code{"formula"}, except that an intercept term is not added by default. Any function of class \code{"nonlin"} may be used in addition to the usual operators and functions. } \value{ A list with the components required of a \code{"nonlin"} function: \item{ predictors }{the \code{expression} argument passed to \code{Exp}} \item{ term }{a function to create a deparsed mathematical expression of the term, given a label for the predictor.} \item{ call }{the call to use as a prefix for parameter labels. } } \author{ Heather Turner and David Firth } \seealso{ \code{\link{gnm}}, \code{\link{formula}}, \code{\link{nonlin.function}}} \examples{ set.seed(1) ## Using 'Mult' with 'Exp' to constrain the first constituent multiplier ## to be non-negative ## Fit the "UNIDIFF" mobility model across education levels unidiff <- gnm(Freq ~ educ*orig + educ*dest + Mult(Exp(educ), orig:dest), family = poisson, data = yaish, subset = (dest != 7)) } \keyword{ models } \keyword{ regression } \keyword{ nonlinear } gnm/man/instances.Rd0000644000176000001440000000263612515147523014130 0ustar ripleyusers\name{instances} \alias{instances} \title{ Specify Multiple Instances of a Nonlinear Term in a gnm Model Formula } \description{ A symbolic wrapper, for use in the formula argument to \code{\link{gnm}}, to specify multiple instances of a term specified by a function with an \code{inst} argument. } \usage{ instances(term, instances = 1) } \arguments{ \item{term}{ a call to a function with an inst argument, which specifies some term. } \item{instances}{ the desired number of instances of the term. } } \value{ A deparsed expression representing the summation of \code{term} specified with \code{inst = 1}, \code{inst = 2}, ..., \code{inst = instances}, which is used to create an expanded formula. } \author{ Heather Turner} \seealso{\code{\link{gnm}}, \code{\link{formula}}, \code{\link{nonlin.function}}, \code{\link{Mult}}, \code{\link{MultHomog}} } \examples{ \dontrun{ ## (this example can take quite a while to run) ## ## Fitting two instances of a multiplicative interaction (i.e. a ## two-component interaction) yield.scaled <- wheat$yield * sqrt(3/1000) treatment <- factor(paste(wheat$tillage, wheat$summerCrop, wheat$manure, wheat$N, sep = "")) bilinear2 <- gnm(yield.scaled ~ year + treatment + instances(Mult(year, treatment), 2), family = gaussian, data = wheat) } } \keyword{ models } \keyword{ regression } \keyword{ nonlinear } gnm/man/wheat.Rd0000644000176000001440000001063512515147523013247 0ustar ripleyusers\name{wheat} \alias{wheat} \docType{data} \title{ Wheat Yields from Mexican Field Trials } \description{ Data from a 10-year experiment at the CIMMYT experimental station located in the Yaqui Valley near Ciudad Obregon, Sonora, Mexico --- factorial design using 24 treatments in all. In each of the 10 years the experiment was arranged in a randomized complete block design with three replicates. } \usage{wheat} \format{ A data frame with 240 observations on the following 33 variables. \describe{ \item{yield}{numeric, mean yield in kg/ha for 3 replicates} \item{year}{a factor with levels \code{1988:1997}} \item{tillage}{a factor with levels \code{T} \code{t}} \item{summerCrop}{a factor with levels \code{S} \code{s}} \item{manure}{a factor with levels \code{M} \code{m}} \item{N}{a factor with levels \code{0} \code{N} \code{n}} \item{MTD}{numeric, mean max temp sheltered (deg C) in December} \item{MTJ}{same for January} \item{MTF}{same for February} \item{MTM}{same for March} \item{MTA}{same for April} \item{mTD}{numeric, mean min temp sheltered (deg C) in December} \item{mTJ}{same for January} \item{mTF}{same for February} \item{mTM}{same for March} \item{mTA}{same for April} \item{mTUD}{numeric, mean min temp unsheltered (deg C)in December} \item{mTUJ}{same for January} \item{mTUF}{same for February} \item{mTUM}{same for March} \item{mTUA}{same for April} \item{PRD}{numeric, total precipitation (mm) in December} \item{PRJ}{same for January} \item{PRF}{same for February} \item{PRM}{same for March} \item{SHD}{numeric, mean sun hours in December} \item{SHJ}{same for January} \item{SHF}{same for February} \item{EVD}{numeric, total evaporation (mm) in December} \item{EVJ}{same for January} \item{EVF}{same for February} \item{EVM}{same for March} \item{EVA}{same for April} } } \source{ Tables A1 and A3 of Vargas, M, Crossa, J, van Eeuwijk, F, Sayre, K D and Reynolds, M P (2001). Interpreting treatment by environment interaction in agronomy trials. \emph{Agronomy Journal} \bold{93}, 949--960. } \examples{ set.seed(1) ## Scale yields to reproduce analyses reported in Vargas et al (2001) yield.scaled <- wheat$yield * sqrt(3/1000) ## Reproduce (up to error caused by rounding) Table 1 of Vargas et al (2001) aov(yield.scaled ~ year*tillage*summerCrop*manure*N, data = wheat) treatment <- interaction(wheat$tillage, wheat$summerCrop, wheat$manure, wheat$N, sep = "") mainEffects <- lm(yield.scaled ~ year + treatment, data = wheat) svdStart <- residSVD(mainEffects, year, treatment, 3) bilinear1 <- update(asGnm(mainEffects), . ~ . + Mult(year, treatment), start = c(coef(mainEffects), svdStart[,1])) bilinear2 <- update(bilinear1, . ~ . + Mult(year, treatment, inst = 2), start = c(coef(bilinear1), svdStart[,2])) bilinear3 <- update(bilinear2, . ~ . + Mult(year, treatment, inst = 3), start = c(coef(bilinear2), svdStart[,3])) anova(mainEffects, bilinear1, bilinear2, bilinear3) ## Examine the extent to which, say, mTF explains the first bilinear term bilinear1mTF <- gnm(yield.scaled ~ year + treatment + Mult(1 + mTF, treatment), family = gaussian, data = wheat) anova(mainEffects, bilinear1mTF, bilinear1) ## How to get the standard SVD representation of an AMMI-n model ## ## We'll work with the AMMI-2 model, which here is called "bilinear2" ## ## First, extract the contributions of the 5 terms in the model: ## wheat.terms <- termPredictors(bilinear2) ## ## That's a matrix, whose 4th and 5th columns are the interaction terms ## ## Combine those two interaction terms, to get the total estimated ## interaction effect: ## wheat.interaction <- wheat.terms[, 4] + wheat.terms[, 5] ## ## That's a vector, so we need to re-shape it as a 24 by 10 matrix ## ready for calculating the SVD: ## wheat.interaction <- matrix(wheat.interaction, 24, 10) ## ## Now we can compute the SVD: ## wheat.interaction.SVD <- svd(wheat.interaction) ## ## Only the first two singular values are nonzero, as expected ## (since this is an AMMI-2 model, the interaction has rank 2) ## ## So the result object can be simplified by re-calculating the SVD with ## just two dimensions: ## wheat.interaction.SVD <- svd(wheat.interaction, nu = 2, nv = 2) } \keyword{datasets} gnm/man/gnm-defunct.Rd0000644000176000001440000000145112515147523014342 0ustar ripleyusers\name{gnm-defunct} \alias{gnm-defunct} \alias{Nonlin} \alias{getModelFrame} \alias{qrSolve} \title{Defunct Functions in gnm Package} \description{ The functions listed here are no longer part of gnm as they are not needed any more. } \usage{ Nonlin(functionCall) getModelFrame() qrSolve(A, b, rank = NULL, ...) } \details{ \code{Nonlin} is not needed any more as the plug-in architecture has been replaced by functions of class \code{"nonlin"}, see \code{\link{nonlin.function}}. \code{getModelFrame} was designed to work from within a plug-in function so is no longer needed. \code{qrSolve} was a function to solve the linear system Ax = b by two applications of QR decomposition. Alternative methods were found to be more robust. } \seealso{\code{\link{.Defunct}}} \keyword{internal} gnm/man/plot.gnm.Rd0000644000176000001440000001132412515147523013671 0ustar ripleyusers\name{plot.gnm} \alias{plot.gnm} \title{ Plot Diagnostics for a gnm Object } \description{ Five plots are available: a plot of residuals against fitted values, a Scale-Location plot of \eqn{\sqrt{| residuals |}}{sqrt{| residuals |}} against fitted values, a Normal Q-Q plot, a plot of Cook's distances versus row labels, and a plot of residuals against leverages. By default, all except the fourth are produced. } \usage{ \method{plot}{gnm}(x, which = c(1:3, 5), caption = c("Residuals vs Fitted", "Normal Q-Q", "Scale-Location", "Cook's distance", "Residuals vs Leverage"), panel = if (add.smooth) panel.smooth else points, sub.caption = NULL, main = "", ask = prod(par("mfcol")) < length(which) && dev.interactive(), ..., id.n = 3, labels.id = names(residuals(x)), cex.id = 0.75, qqline = TRUE, cook.levels = c(0.5, 1), add.smooth = getOption("add.smooth"), label.pos = c(4, 2), cex.caption = 1) } \arguments{ \item{x}{ a \code{"gnm"} object. } \item{which}{ a subset of the numbers 1:5 specifying which plots to produce (out of those listed in Description section). } \item{caption}{ captions to appear above the plots. } \item{panel}{ panel function. The useful alternative to \code{points}, \code{panel.smooth} can be chosen by \code{add.smooth = TRUE}. } \item{sub.caption}{ common title - above figures if there are multiple; used as \code{sub} (s.\code{title}) otherwise. If \code{NULL}, as by default, a possible shortened version of \code{deparse(x$call)} is used. } \item{main}{ title to each plot - in addition to the above \code{caption}. } \item{ask}{ logical; if \code{TRUE}, the user is asked before each plot, see \code{par(ask = .)}.} \item{\dots}{ other parameters to be passed through to plotting functions. } \item{id.n}{ number of points to be labelled in each plot starting with the most extreme. } \item{labels.id}{ vector of labels, from which the labels for extreme points will be chosen. \code{NULL} uses observation numbers. } \item{cex.id}{ magnification of point labels. } \item{qqline}{ logical indicating if a \code{qqline()} should be added to the normal Q-Q plot.} \item{cook.levels}{ levels of Cook's distance at which to draw contours. } \item{add.smooth}{ logical indicating if a smoother should be added to most plots; see also \code{panel} above.} \item{label.pos}{ positioning of labels, for the left half and right half of the graph respectively, for plots 1-3. } \item{cex.caption}{ controls the size of 'caption'. } } \details{ \code{sub.caption} - by default the function call - is shown as a subtitle (under the x-axis title) on each plot when plots are on separate pages, or as a subtitle in the outer margin (if any) when there are multiple plots per page. The "Scale-Location" plot, also called "Spread-Location" or "S-L" plot, takes the square root of the absolute residuals in order to diminish skewness (\eqn{\sqrt{| E |}}{sqrt{| E |}} is much less skewed than \eqn{| E |} for Gaussian zero-mean \eqn{E}). The S-L, the Q-Q, and the Residual-Leverage plot, use \emph{standardized} residuals which have identical variance (under the hypothesis). They are given as \eqn{R[i] / (s*\sqrt(1 - h_{ii}))}{R[i] / (s*sqrt(1 - h.ii))} where \eqn{h_{ii}}{h.ii} are the diagonal entries of the hat matrix, \code{influence()$hat}, see also \code{\link{hat}}. The Residual-Leverage plot shows contours of equal Cook's distance, for values of \code{cook.levels} (by default 0.5 and 1) and omits cases with leverage one. If the leverages are constant, as typically in a balanced \code{aov} situation, the plot uses factor level combinations instead of the leverages for the x-axis. } \author{ Modification of \code{\link{plot.lm}} by the R Core Team. Adapted for \code{"gnm"} objects by Heather Turner. } \seealso{ \code{\link{gnm}}, \code{\link{plot.lm}} } \examples{ set.seed(1) ## Fit an association model with homogeneous row-column effects RChomog <- gnm(Freq ~ origin + destination + Diag(origin, destination) + MultHomog(origin, destination), family = poisson, data = occupationalStatus) ## Plot model diagnostics plot(RChomog) ## Put 4 plots on 1 page; allow room for printing model formula in outer margin: par(mfrow = c(2, 2), oma = c(0, 0, 3, 0)) title <- paste(deparse(RChomog$formula, width.cutoff = 50), collapse = "\n") plot(RChomog, sub.caption = title) ## Fit smoother curves plot(RChomog, sub.caption = title, panel = panel.smooth) plot(RChomog, sub.caption = title, panel = function(x,y) panel.smooth(x, y, span = 1)) } \keyword{ models } \keyword{ regression } \keyword{ nonlinear } \keyword{ hplot } gnm/man/barley.Rd0000644000176000001440000000464312515147523013417 0ustar ripleyusers\name{barley} \alias{barley} \docType{data} \title{ Jenkyn's Data on Leaf-blotch on Barley } \description{ Incidence of \emph{R. secalis} on the leaves of ten varieties of barley grown at nine sites. } \usage{barley} \format{ A data frame with 90 observations on the following 3 variables. \describe{ \item{y}{the proportion of leaf affected (values in [0,1])} \item{site}{a factor with 9 levels \code{A} to \code{I}} \item{variety}{a factor with 10 levels \code{c(1:9, "X")}} } } \note{ This dataset was used in Wedderburn's original paper (1974) on quasi-likelihood. } \source{ Originally in an unpublished Aberystwyth PhD thesis by J F Jenkyn. } \references{ Gabriel, K R (1998). Generalised bilinear regression. \emph{Biometrika} \bold{85}, 689--700. McCullagh, P and Nelder, J A (1989) \emph{Generalized Linear Models} (2nd ed). Chapman and Hall. Wedderburn, R W M (1974). Quasilikelihood functions, generalized linear models and the Gauss-Newton method. \emph{Biometrika} \bold{61}, 439--47. } \examples{ set.seed(1) ### Fit Wedderburn's logit model with variance proportional to [mu(1-mu)]^2 logitModel <- glm(y ~ site + variety, family = wedderburn, data = barley) fit <- fitted(logitModel) print(sum((barley$y - fit)^2 / (fit * (1-fit))^2)) ## Agrees with the chi-squared value reported in McCullagh and Nelder ## (1989, p331), which differs slightly from Wedderburn's reported value. ### Fit the biplot model as in Gabriel (1998, p694) biplotModel <- gnm(y ~ -1 + instances(Mult(site, variety), 2), family = wedderburn, data = barley) barleySVD <- svd(matrix(biplotModel$predictors, 10, 9)) A <- sweep(barleySVD$v, 2, sqrt(barleySVD$d), "*")[, 1:2] B <- sweep(barleySVD$u, 2, sqrt(barleySVD$d), "*")[, 1:2] ## These are essentially A and B as in Gabriel (1998, p694), from which ## the biplot is made by plot(rbind(A, B), pch = c(levels(barley$site), levels(barley$variety))) ## Fit the double-additive model as in Gabriel (1998, p697) variety.binary <- factor(match(barley$variety, c(2,3,6), nomatch = 0) > 0, labels = c("rest", "2,3,6")) doubleAdditive <- gnm(y ~ variety + Mult(site, variety.binary), family = wedderburn, data = barley) ## It is unclear why Gabriel's chi-squared statistics differ slightly ## from the ones produced in these fits. Possibly Gabriel adjusted the ## data somehow prior to fitting? } \keyword{datasets} gnm/man/Mult.Rd0000644000176000001440000000700612515147523013056 0ustar ripleyusers\name{Multiplicative interaction} \alias{Mult} \title{Specify a Product of Predictors in a gnm Model Formula} \description{ A function of class \code{"nonlin"} to specify a multiplicative interaction in the formula argument to \code{\link{gnm}}. } \usage{ Mult(..., inst = NULL) } \arguments{ \item{\dots}{a comma-separated list of two or more symbolic expressions representing the constituent multipliers in the interaction.} \item{inst}{a positive integer specifying the instance number of the term.} } \details{ \code{Mult} specifies instances of a multiplicative interaction, i.e. a product of the form \deqn{m_1 m_2 ... m_n,} where the constituent multipliers \eqn{m_1, m_2, ..., m_n} are linear or nonlinear predictors. Models for the constituent multipliers are specified symbolically as unspecified arguments to \code{Mult}. These symbolic expressions are interpreted in the same way as the right hand side of a formula in an object of class \code{"formula"}, except that an intercept term is not added by default. Offsets can be added to constituent multipliers, using \code{offset}. The family of multiplicative interaction models include row-column association models for contingency tables (e.g., Agresti, 2002, Sec 9.6), log-multiplicative or UNIDIFF models (Erikson and Goldthorpe, 1992; Xie, 1992), and GAMMI models (van Eeuwijk, 1995). } \value{ A list with the required components of a \code{"nonlin"} function: \item{ predictors }{ the expressions passed to \code{Mult}} \item{ term }{ a function to create a deparsed mathematical expression of the term, given labels for the predictors.} \item{ call }{ the call to use as a prefix for parameter labels. } } \references{ Agresti, A (2002). \emph{Categorical Data Analysis} (2nd ed.) New York: Wiley. Erikson, R and Goldthorpe, J H (1992). \emph{The Constant Flux}. Oxford: Clarendon Press. van Eeuwijk, F A (1995). Multiplicative interaction in generalized linear models. \emph{Biometrics} \bold{51}, 1017-1032. Vargas, M, Crossa, J, van Eeuwijk, F, Sayre, K D and Reynolds, M P (2001). Interpreting treatment by environment interaction in agronomy trials. \emph{Agronomy Journal} \bold{93}, 949--960. Xie, Y (1992). The log-multiplicative layer effect model for comparing mobility tables. \emph{American Sociological Review} \bold{57}, 380-395. } \author{Heather Turner} \seealso{\code{\link{gnm}}, \code{\link{formula}}, \code{\link{instances}}, \code{\link{nonlin.function}}, \code{\link{MultHomog}} } \examples{ set.seed(1) ## Using 'Mult' with 'Exp' to constrain the first constituent multiplier ## to be non-negative ## Fit the "UNIDIFF" mobility model across education levels unidiff <- gnm(Freq ~ educ*orig + educ*dest + Mult(Exp(educ), orig:dest), family = poisson, data = yaish, subset = (dest != 7)) \dontrun{ ## (this example can take quite a while to run) ## ## Fitting two instances of a multiplicative interaction (i.e. a ## two-component interaction)) yield.scaled <- wheat$yield * sqrt(3/1000) treatment <- factor(paste(wheat$tillage, wheat$summerCrop, wheat$manure, wheat$N, sep = "")) bilinear2 <- gnm(yield.scaled ~ year + treatment + instances(Mult(year, treatment), 2), family = gaussian, data = wheat) formula(bilinear2) ## yield.scaled ~ year + treatment + Mult(year, treatment, inst = 1) + ## Mult(year, treatment, inst = 2) } } \keyword{ models } \keyword{ regression } \keyword{ nonlinear } gnm/man/ofInterest.Rd0000644000176000001440000000431712515147523014261 0ustar ripleyusers\name{ofInterest} \alias{ofInterest} \alias{ofInterest<-} \title{ Coefficents of Interest in a Generalized Nonlinear Model } \description{ Retrieve or set the \code{"ofInterest"} component of a \code{"gnm"} (generalized nonlinear model) object. } \usage{ ofInterest(object) ofInterest(object) <- value } \arguments{ \item{object}{ an object of class \code{"gnm"}. } \item{value}{ a numeric vector of indices specifying the subset of (non-eliminated) coefficients of interest, or \code{NULL} to specify that all non-eliminated coefficients are of interest. } } \details{ The \code{"ofInterest"} component of a \code{"gnm"} object is a named numeric vector of indices specifying a subset of the non-eliminated coefficients which are of specific interest. If the \code{"ofInterest"} component is non-NULL, printed summaries of the model only show the coefficients of interest. In addition methods for \code{"gnm"} objects which may be applied to a subset of the parameters are by default applied to the coefficients of interest. These functions provide a way of extracting and replacing the \code{"ofInterest"} component. The replacement function prints the replacement value to show which parameters have been specified by \code{value}. } \value{ A named vector of indices, or \code{NULL}. } \author{ Heather Turner } \note{ Regular expression matching is performed using \code{grep} with default settings. } \seealso{ \code{\link{grep}}, \code{\link{gnm}}, \code{\link{se}}, \code{\link{getContrasts}},\code{\link{profile.gnm}}, \code{\link{confint.gnm}}} \examples{ set.seed(1) ## Fit the "UNIDIFF" mobility model across education levels unidiff <- gnm(Freq ~ educ*orig + educ*dest + Mult(Exp(educ), orig:dest), ofInterest = "[.]educ", family = poisson, data = yaish, subset = (dest != 7)) ofInterest(unidiff) ## Get all of the contrasts with educ1 in the UNIDIFF multipliers getContrasts(unidiff, ofInterest(unidiff)) ## Get estimate and se for the contrast between educ4 and educ5 in the ## UNIDIFF multiplier mycontrast <- numeric(length(coef(unidiff))) mycontrast[ofInterest(unidiff)[4:5]] <- c(1, -1) se(unidiff, mycontrast) } \keyword{ models } gnm/man/profile.gnm.Rd0000644000176000001440000001734212515147523014361 0ustar ripleyusers\name{profile.gnm} \alias{profile.gnm} \alias{plot.profile.gnm} \title{ Profile Deviance for Parameters in a Generalized Nonlinear Model } \description{ For one or more parameters in a generalized nonlinear model, profile the deviance over a range of values about the fitted estimate. } \usage{ \method{profile}{gnm}(fitted, which = ofInterest(fitted), alpha = 0.05, maxsteps = 10, stepsize = NULL, trace = FALSE, ...) } \arguments{ \item{fitted}{ an object of class \code{"gnm"}. } \item{which}{ (optional) either a numeric vector of indices or a character vector of names, specifying the parameters over which the deviance is to be profiled. If missing, the deviance is profiled over all parameters. } \item{alpha}{ the significance level of the z statistic, indicating the range that the profile must cover (see details). } \item{maxsteps}{ the maximum number of steps to take either side of the fitted parameter. } \item{stepsize}{ (optional) a numeric vector of length two, specifying the size of steps to take when profiling down and up respectively, or a single number specifying the step size in both directions. If missing, the step sizes will be determined automatically. } \item{trace}{ logical, indicating whether profiling should be traced. } \item{\dots}{ further arguments. } } \details{ This is a method for the generic function \code{profile} in the \code{base} package. For a given parameter, the deviance is profiled by constraining that parameter to certain values either side of its estimate in the fitted model and refitting the model. For each updated model, the following "z statistic" is computed \deqn{z(\theta) = (\theta - \hat{\theta}) * \sqrt{\frac{D_{theta} - D_{\hat{theta}}}{\delta}}}{ z(theta) = (theta - theta.hat) * sqrt((D_theta - D_theta.hat)/delta)} where \eqn{\theta}{theta} is the constrained value of the parameter; \eqn{\hat{\theta}}{theta.hat} is the original fitted value; \eqn{D_{\theta}}{D_theta} is the deviance when the parameter is equal to \eqn{\theta}{theta}, and \eqn{\delta}{delta} is the dispersion parameter. When the deviance is quadratic in \eqn{\theta}{theta}, z will be linear in \eqn{\theta}{theta}. Therefore departures from quadratic behaviour can easily be identified by plotting z against \eqn{\theta}{theta} using \code{plot.profile.gnm}. \code{confint.profile.gnm} estimates confidence intervals for the parameters by interpolating the deviance profiles and identifying the parameter values at which z is equal to the relevant percentage points of the normal distribution. The \code{alpha} argument to \code{profile.gnm} specifies the significance level of z which must be covered by the profile. In particular, the profiling in a given direction will stop when \code{maxsteps} is reached or two steps have been taken in which \deqn{z(\theta) > (\theta - \hat{\theta}) * z_{(1 - \alpha)/2}}{ z(theta) > (theta - theta.hat) * z_{(1 - alpha)/2}} By default, the stepsize is \deqn{z_{(1 - \alpha)/2} * s_{\hat{\theta}}}{ z_{(1 - alpha)/2} * s_theta.hat} where \eqn{s_{\hat{\theta}}}{s_theta.hat} is the standard error of \eqn{\hat{\theta}}{theta.hat}. Strong assymetry is detected and the stepsize is adjusted accordingly, to try to ensure that the range determined by \code{alpha} is adequately covered. \code{profile.gnm} will also attempt to detect if the deviance is asymptotic such that the desired significance level cannot be reached. Each profile has an attribute \code{"asymptote"}, a two-length logical vector specifying whether an asymptote has been detected in either direction. For unidentified parameters the profile will be \code{NA}, as such parameters cannot be profiled. } \value{ A list of profiles, with one named component for each parameter profiled. Each profile is a data.frame: the first column, "z", contains the z statistics and the second column "par.vals" contains a matrix of parameter values, with one column for each parameter in the model. The list has two attributes: "original.fit" containing \code{fitted} and "summary" containing \code{summary(fitted)}. } \references{ Chambers, J. M. and Hastie, T. J. (1992) \emph{Statistical Models in S} } \author{ Modification of \code{\link[MASS]{profile.glm}} from the MASS package. Originally D. M. Bates and W. N. Venables, ported to R by B. D. Ripley, adapted for \code{"gnm"} objects by Heather Turner. } \seealso{ \code{\link{confint.gnm}}, \code{\link{gnm}}, \code{\link[MASS]{profile.glm}}, \code{\link{ofInterest}} } \examples{ set.seed(1) ### Example in which deviance is near quadratic count <- with(voting, percentage/100 * total) yvar <- cbind(count, voting$total - count) classMobility <- gnm(yvar ~ -1 + Dref(origin, destination), constrain = "delta1", family = binomial, data = voting) prof <- profile(classMobility, trace = TRUE) plot(prof) ## confint similar to MLE +/- 1.96*s.e. confint(prof, trace = TRUE) coefData <- se(classMobility) cbind(coefData[1] - 1.96 * coefData[2], coefData[1] + 1.96 * coefData[2]) \dontrun{ ### These examples take longer to run ### Another near quadratic example RChomog <- gnm(Freq ~ origin + destination + Diag(origin, destination) + MultHomog(origin, destination), ofInterest = "MultHomog", constrain = "MultHomog.*1", family = poisson, data = occupationalStatus) prof <- profile(RChomog, trace = TRUE) plot(prof) ## confint similar to MLE +/- 1.96*s.e. confint(prof) coefData <- se(RChomog) cbind(coefData[1] - 1.96 * coefData[2], coefData[1] + 1.96 * coefData[2]) ## Another near quadratic example, with more complex constraints count <- with(voting, percentage/100 * total) yvar <- cbind(count, voting$total - count) classMobility <- gnm(yvar ~ -1 + Dref(origin, destination), family = binomial, data = voting) wts <- prop.table(exp(coef(classMobility))[1:2]) classMobility <- update(classMobility, constrain = "delta1", constrainTo = log(wts[1])) sum(exp(parameters(classMobility))[1:2]) #=1 prof <- profile(classMobility, trace = TRUE) plot(prof) ## confint similar to MLE +/- 1.96*s.e. confint(prof, trace = TRUE) coefData <- se(classMobility) cbind(coefData[1] - 1.96 * coefData[2], coefData[1] + 1.96 * coefData[2]) ### An example showing asymptotic deviance unidiff <- gnm(Freq ~ educ*orig + educ*dest + Mult(Exp(educ), orig:dest), ofInterest = "[.]educ", constrain = "[.]educ1", family = poisson, data = yaish, subset = (dest != 7)) prof <- profile(unidiff, trace = TRUE) plot(prof) ## clearly not quadratic for Mult1.Factor1.educ4 or Mult1.Factor1.educ5! confint(prof) ## 2.5 \% 97.5 \% ## Mult(Exp(.), orig:dest).educ1 NA NA ## Mult(Exp(.), orig:dest).educ2 -0.5978901 0.1022447 ## Mult(Exp(.), orig:dest).educ3 -1.4836854 -0.2362378 ## Mult(Exp(.), orig:dest).educ4 -2.5792398 -0.2953420 ## Mult(Exp(.), orig:dest).educ5 -Inf -0.7006889 coefData <- se(unidiff) cbind(coefData[1] - 1.96 * coefData[2], coefData[1] + 1.96 * coefData[2]) ### A far from quadratic example, also with eliminated parameters backPainLong <- expandCategorical(backPain, "pain") oneDimensional <- gnm(count ~ pain + Mult(pain, x1 + x2 + x3), eliminate = id, family = "poisson", constrain = "[.](painworse|x1)", constrainTo = c(0, 1), data = backPainLong) prof <- profile(oneDimensional, trace = TRUE) plot(prof) ## not quadratic for any non-eliminated parameter confint(prof) coefData <- se(oneDimensional) cbind(coefData[1] - 1.96 * coefData[2], coefData[1] + 1.96 * coefData[2]) } } \keyword{ models } \keyword{ nonlinear } gnm/man/gnm.Rd0000755000176000001440000004403112515147523012720 0ustar ripleyusers\name{gnm} \alias{gnm} \title{ Fitting Generalized Nonlinear Models } \description{ \code{gnm} fits generalised nonlinear models using an over-parameterised representation. Nonlinear terms are specified by calls to functions of class \code{"nonlin"}. } \usage{ gnm(formula, eliminate = NULL, ofInterest = NULL, constrain = numeric(0), constrainTo = numeric(length(constrain)), family = gaussian, data = NULL, subset, weights, na.action, method = "gnmFit", checkLinear = TRUE, offset, start = NULL, etastart = NULL, mustart = NULL, tolerance = 1e-06, iterStart = 2, iterMax = 500, trace = FALSE, verbose = TRUE, model = TRUE, x = TRUE, termPredictors = FALSE, ridge = 1e-08, ...) } \arguments{ \item{formula}{ a symbolic description of the nonlinear predictor. } \item{eliminate}{ a factor to be included as the first term in the model. \code{gnm} will exploit the structure of this factor to improve computational efficiency. See details. } \item{ofInterest}{ optional coefficients of interest, specified by a regular expression, a numeric vector of indices, a character vector of names, or "[?]" to select from a Tk dialog. If missing, it is assumed that all non-\code{eliminate}d coefficients are of interest. } \item{constrain}{ (non-eliminated) coefficients to constrain, specified by a regular expression, a numeric vector of indices, a logical vector, a character vector of names, or "[?]" to select from a Tk dialog. } \item{constrainTo}{ a numeric vector of the same length as \code{constrain} specifying the values to constrain to. By default constrained parameters will be set to zero. } \item{family}{ a specification of the error distribution and link function to be used in the model. This can be a character string naming a family function; a family function, or the result of a call to a family function. See \code{\link{family}} and \code{\link{wedderburn}} for possibilities. } \item{data}{ an optional data frame containing the variables in the model. If not found in \code{data}, the variables are taken from \code{environment(formula)}, typically the environment from which \code{gnm} is called.} \item{subset}{ an optional vector specifying a subset of observations to be used in the fitting process.} \item{weights}{ an optional vector of weights to be used in the fitting process.} \item{na.action}{ a function which indicates what should happen when the data contain \code{NA}s. If \code{data} is a contingency table, the default is \code{"exclude"}. Otherwise the default is first, any \code{na.action} attribute of \code{data}; second, any \code{na.action} setting of \code{options}, and third, \code{na.fail}.} \item{method}{ the method to be used: either \code{"gnmFit"} to fit the model using the default maximum likelihood algorithm, \code{"coefNames"} to return a character vector of names for the coefficients in the model, \code{"model.matrix"} to return the model matrix, \code{"model.frame"} to return the model frame, or the name of a function providing an alternative fitting algorithm. } \item{checkLinear}{ logical: if \code{TRUE} \code{glm.fit} is used when the predictor is found to be linear } \item{offset}{ this can be used to specify an a priori known component to be added to the predictor during fitting. \code{offset} terms can be included in the formula instead or as well, and if both are specified their sum is used.} \item{start}{ a vector of starting values for the parameters in the model; if a starting value is \code{NA}, the default starting value will be used. Starting values need not be specified for eliminated parameters. } \item{etastart}{ starting values for the linear predictor. } \item{mustart}{ starting values for the vector of means. } \item{tolerance}{ a positive numeric value specifying the tolerance level for convergence. } \item{iterStart}{ a positive integer specifying the number of start-up iterations to perform. } \item{iterMax}{ a positive integer specifying the maximum number of main iterations to perform. } \item{trace}{ a logical value indicating whether the deviance should be printed after each iteration. } \item{verbose}{ logical: if \code{TRUE} and model includes nonlinear terms, progress indicators are printed as the model is fitted, including a diagnostic error message if the algorithm fails. } \item{model}{ logical: if \code{TRUE} the model frame is returned. } \item{x}{ logical: if \code{TRUE} the local design matrix from the last iteration is included as a component of returned model object. } \item{termPredictors}{ logical: if \code{TRUE}, a matrix is returned with a column for each term in the model, containing the additive contribution of that term to the predictor. } \item{ridge}{numeric, a positive value for the ridge constant to be used in the fitting algorithm} \item{\dots}{ further arguments passed to fitting function. } } \details{ Models for \code{gnm} are specified by giving a symbolic description of the nonlinear predictor, of the form \code{response ~ terms}. The \code{response} is typically a numeric vector, see later in this section for alternatives. The usual symbolic language may be used to specify any linear terms, see \code{\link{formula}} for details. Nonlinear terms may be specified by calls to functions of class "nonlin". There are several "nonlin" functions in the \code{gnm} package. Some of these specify simple mathematical functions of predictors: \code{Exp}, \code{Mult}, and \code{Inv}. Others specify more specialised nonlinear terms, in particular \code{MultHomog} specifies homogeneous multiplicative interactions and \code{Dref} specifies diagonal reference terms. Users may also define their own "nonlin" functions, see \code{\link{nonlin.function}} for details. The \code{eliminate} argument may be used to specify a factor that is to be included as the first term in the model (since an intercept is then redundant, none is fitted). The structure of the factor is exploited to improve computational efficiency --- substantially so if the \code{eliminate}d factor has a large number of levels. Use of \code{eliminate} is designed for factors that are required in the model but are not of direct interest (e.g., terms needed to fit multinomial-response models as conditional Poisson models). See \code{\link{backPain}} for an example. The \code{ofInterest} argument may be used to specify coefficients of interest, the indices of which are returned in the \code{ofInterest} component of the model object. \code{print()} displays of the model object or its components obtained using accessor functions such as \code{coef()} etc, will only show these coefficients. In addition methods for \code{"gnm"} objects which may be applied to a subset of the parameters are by default applied to the coefficients of interest. See \code{\link{ofInterest}} for accessor and replacement functions. For contingency tables, the data may be provided as an object of class \code{"table"} from which the frequencies will be extracted to use as the response. In this case, the response should be specified as \code{Freq} in the model formula. The \code{"predictors"}, \code{"fitted.values"}, \code{"residuals"}, \code{"prior.weights"}, \code{"weights"}, \code{"y"} and \code{"offset"} components of the returned \code{gnm} fit will be tables with the same format as the data, completed with \code{NA}s where necessary. For binomial models, the \code{response} may be specified as a factor in which the first level denotes failure and all other levels denote success, as a two-column matrix with the columns giving the numbers of successes and failures, or as a vector of the proportions of successes. The \code{gnm} fitting algorithm consists of two stages. In the start-up iterations, any nonlinear parameters that are not specified by either the \code{start} argument of \code{gnm} or a plug-in function are updated one parameter at a time, then the linear parameters are jointly updated before the next iteration. In the main iterations, all the parameters are jointly updated, until convergence is reached or the number or iterations reaches \code{iterMax}. To solve the (typically rank-deficient) least squares problem at the heart of the \code{gnm} fitting algorithm, the design matrix is standardized and regularized (in the Levenberg-Marquardt sense) prior to solving; the \code{ridge} argument provides a degree of control over the regularization performed (smaller values may sometimes give faster convergence but can lead to numerical instability). Convergence is judged by comparing the squared components of the score vector with corresponding elements of the diagonal of the Fisher information matrix. If, for all components of the score vector, the ratio is less than \code{tolerance^2}, or the corresponding diagonal element of the Fisher information matrix is less than 1e-20, iterations cease. If the algorithm has not converged by \code{iterMax} iterations, \code{\link{exitInfo}} can be used to print information on the parameters which failed the convergence criteria at the last iteration. By default, \code{gnm} uses an over-parameterized representation of the model that is being fitted. Only minimal identifiability constraints are imposed, so that in general a random parameterization is obtained. The parameter estimates are ordered so that those for any linear terms appear first. \code{\link{getContrasts}} may be used to obtain estimates of specified scaled contrasts, if these contrasts are identifiable. For example, \code{getContrasts} may be used to estimate the contrasts between the first level of a factor and the rest, and obtain standard errors. If appropriate constraints are known in advance, or have been determined from a \code{gnm} fit, the model may be (re-)fitted using the \code{constrain} argument to specify coefficients which should be set to values specified by \code{constrainTo}. Constraints should only be specified for non-eliminated parameters. \code{\link{update}} provides a convenient way of re-fitting a \code{gnm} model with new constraints. } \value{ If \code{method = "gnmFit"}, \code{gnm} returns \code{NULL} if the algorithm has failed and an object of class \code{"gnm"} otherwise. A \code{"gnm"} object inherits first from \code{"glm"} then \code{"lm"} and is a list containing the following components: \item{ call }{ the matched call. } \item{ formula }{ the formula supplied. } \item{ constrain }{ a numeric vector specifying any coefficients that were constrained in the fitting process. } \item{ constrainTo }{ a numeric vector of the same length as \code{constrain} specifying the values which constrained parameters were set to. } \item{ family }{ the \code{family} object used. } \item{ prior.weights }{ the case weights initially supplied. } \item{ terms }{ the \code{terms} object used. } \item{ data }{ the \code{data} argument. } \item{ na.action }{ the \code{na.action} attribute of the model frame } \item{ xlevels }{ a record of the levels of the factors used in fitting. } \item{ y }{ the response used. } \item{ offset }{ the offset vector used. } \item{ coefficients }{ a named vector of non-eliminated coefficients, with an attribute \code{"eliminated"} specifying the eliminated coefficients if \code{eliminate} is non-\code{NULL}. } \item{ eliminate }{ the \code{eliminate} argument. } \item{ ofInterest }{ a named numeric vector of indices corresponding to non-eliminated coefficients, or \code{NULL}. } \item{ predictors }{ the fitted values on the link scale. } \item{ fitted.values }{ the fitted mean values, obtained by transforming the predictors by the inverse of the link function. } \item{ deviance }{ up to a constant, minus twice the maximised log-likelihood. Where sensible, the constant is chosen so that a saturated model has deviance zero. } \item{ aic }{ Akaike's \emph{An Information Criterion}, minus twice the maximized log-likelihood plus twice the number of parameters (so assuming that the dispersion is known).} \item{ iter }{ the number of main iterations.} \item{ conv }{ logical indicating whether the main iterations converged, with an attribute for use by \code{\link{exitInfo}} if \code{FALSE}. } \item{ weights }{ the \emph{working} weights, that is, the weights used in the last iteration.} \item{ residuals }{ the \emph{working} residuals, that is, the residuals from the last iteration. } \item{ df.residual }{ the residual degrees of freedom. } \item{ rank }{ the numeric rank of the fitted model. } The list may also contain the components \code{model}, \code{x}, or \code{termPredictors} if requested in the arguments to \code{gnm}. If a table was passed to \code{data} and the default for \code{na.action} was not overridden, the list will also contain a \code{table.attr} component, for use by the extractor functions. If a binomial \code{gnm} model is specified by giving a two-column response, the weights returned by \code{prior.weights} are the total numbers of cases (factored by the supplied case weights) and the component \code{y} of the result is the proportion of successes. The function \code{\link{summary.gnm}} may be used to obtain and print a summary of the results, whilst \code{\link{plot.gnm}} may be used for model diagnostics. The generic functions \code{\link{formula}}, \code{\link{family}}, \code{\link{terms}}, \code{\link{coefficients}}, \code{\link{fitted.values}}, \code{\link{deviance}}, \code{\link{extractAIC}}, \code{\link{weights}}, \code{\link{residuals}}, \code{\link{df.residual}}, \code{\link{model.frame}}, \code{\link{model.matrix}}, \code{\link{vcov}} and \code{\link{termPredictors}} maybe used to extract components from the object returned by \code{\link{gnm}} or to construct the relevant objects where necessary. Note that the generic functions \code{\link{weights}} and \code{\link{residuals}} do not act as straight-forward accessor functions for \code{gnm} objects, but return the prior weights and deviance residuals respectively, as for \code{glm} objects. } \references{ Cautres, B, Heath, A F and Firth, D (1998). Class, religion and vote in Britain and France. \emph{La Lettre de la Maison Francaise} \bold{8}. } \author{ Heather Turner and David Firth } \note{ Regular expression matching is performed using \code{grep} with default settings. } \seealso{ \code{\link{formula}} for the symbolic language used to specify formulae. \code{\link{Diag}} and \code{\link{Symm}} for specifying special types of interaction. \code{Exp}, \code{Mult}, \code{Inv}, \code{\link{MultHomog}}, \code{\link{Dref}} and \code{\link{nonlin.function}} for incorporating nonlinear terms in the \code{formula} argument to \code{gnm}. \code{\link{residuals.glm}} and the generic functions \code{\link{coef}}, \code{\link{fitted}}, etc. for extracting components from \code{gnm} objects. \code{\link{exitInfo}} to print more information on last iteration when \code{gnm} has not converged. \code{\link{getContrasts}} to estimate (identifiable) scaled contrasts from a \code{gnm} model. } \examples{ ### Analysis of a 4-way contingency table set.seed(1) print(cautres) ## Fit a "double UNIDIFF" model with the religion-vote and class-vote ## interactions both modulated by nonnegative election-specific ## multipliers. doubleUnidiff <- gnm(Freq ~ election:vote + election:class:religion + Mult(Exp(election), religion:vote) + Mult(Exp(election), class:vote), family = poisson, data = cautres) ## Examine the multipliers of the class-vote log odds ratios ofInterest(doubleUnidiff) <- pickCoef(doubleUnidiff, "class:vote[).]") coef(doubleUnidiff) ## Coefficients of interest: ## Mult(Exp(.), class:vote).election1 ## -0.38357138 ## Mult(Exp(.), class:vote).election2 ## 0.29816599 ## Mult(Exp(.), class:vote).election3 ## 0.06580307 ## Mult(Exp(.), class:vote).election4 ## -0.02174104 ## Re-parameterize by setting first multiplier to zero getContrasts(doubleUnidiff, ofInterest(doubleUnidiff)) ## estimate SE ## Mult(Exp(.), class:vote).election1 0.0000000 0.0000000 ## Mult(Exp(.), class:vote).election2 0.6817374 0.2401644 ## Mult(Exp(.), class:vote).election3 0.4493745 0.2473521 ## Mult(Exp(.), class:vote).election4 0.3618301 0.2534754 ## quasiSE quasiVar ## Mult(Exp(.), class:vote).election1 0.22854401 0.052232363 ## Mult(Exp(.), class:vote).election2 0.07395886 0.005469913 ## Mult(Exp(.), class:vote).election3 0.09475938 0.008979340 ## Mult(Exp(.), class:vote).election4 0.10934798 0.011956981 ## Same thing but with last multiplier as reference category: getContrasts(doubleUnidiff, rev(ofInterest(doubleUnidiff))) ## estimate SE ## Mult(Exp(.), class:vote).election4 0.00000000 0.0000000 ## Mult(Exp(.), class:vote).election3 0.08754436 0.1446833 ## Mult(Exp(.), class:vote).election2 0.31990727 0.1320022 ## Mult(Exp(.), class:vote).election1 -0.36183013 0.2534754 ## quasiSE quasiVar ## Mult(Exp(.), class:vote).election4 0.10934798 0.011956981 ## Mult(Exp(.), class:vote).election3 0.09475938 0.008979340 ## Mult(Exp(.), class:vote).election2 0.07395886 0.005469913 ## Mult(Exp(.), class:vote).election1 0.22854401 0.052232363 ## Re-fit model with first multiplier set to zero doubleUnidiffConstrained <- update(doubleUnidiff, constrain = ofInterest(doubleUnidiff)[1]) ## Examine the multipliers of the class-vote log odds ratios coef(doubleUnidiffConstrained)[ofInterest(doubleUnidiff)] ## ...as using 'getContrasts' (to 4 d.p.). } \keyword{ models } \keyword{ regression } \keyword{ nonlinear } gnm/man/gnm-package.Rd0000644000176000001440000000316712515147523014313 0ustar ripleyusers\name{gnm-package} \alias{gnm-package} \docType{package} \title{ Generalized Nonlinear Models } \description{ Functions to specify, fit and evaluate generalized nonlinear models. } \details{ \code{gnm} provides functions to fit generalized nonlinear models by maximum likelihood. Such models extend the class of generalized linear models by allowing nonlinear terms in the predictor. Some special cases are models with multiplicative interaction terms, such as the UNIDIFF and row-column association models from sociology and the AMMI and GAMMI models from crop science; stereotype models for ordered categorical response, and diagonal reference models for dependence on a square two-way classification. \code{gnm} is a major re-working of an earlier Xlisp-Stat package, "Llama". Over-parameterized representations of models are used throughout; functions are provided for inference on estimable parameter combinations, as well as standard methods for diagnostics etc. The following documentation provides further information on the \code{gnm} package: \describe{ \item{gnmOverview}{\code{vignette("gnmOverview", package = "gnm")}} \item{NEWS}{\code{file.show(system.file("NEWS", package = "gnm"))}} } } \author{ Heather Turner and David Firth Maintainer: Heather Turner } \references{ http://www.warwick.ac.uk/go/gnm } \keyword{ package } \keyword{ models } \keyword{ regression } \keyword{ nonlinear } \seealso{ \code{\link{gnm}} for the model fitting function, with links to associated functions. } \examples{ demo(gnm) } gnm/man/asGnm.Rd0000644000176000001440000000311112515147523013173 0ustar ripleyusers\name{asGnm} \alias{asGnm} \title{ Coerce Linear Model to gnm Object } \description{ \code{asGnm} is a generic function which coerces objects of class "glm" or "lm" to an object of class "gnm". } \usage{ asGnm(object, ...) } \arguments{ \item{object}{ an object of class "glm" or "lm". } \item{\dots}{ additional arguments for method functions. } } \details{ Components are added to or removed from \code{object} to produce an object of class "gnm". This can be useful in model building, see examples. } \value{ An object of class "gnm" - see \code{\link{gnm}} for full description. } \references{ Vargas, M, Crossa, J, van Eeuwijk, F, Sayre, K D and Reynolds, M P (2001). Interpreting treatment by environment interaction in agronomy trials. \emph{Agronomy Journal} \bold{93}, 949--960. } \author{ Heather Turner } \seealso{ \code{\link{gnm}}, \code{\link{glm}}, \code{\link{lm}} } \examples{ set.seed(1) ## Scale yields to reproduce analyses reported in Vargas et al (2001) yield.scaled <- wheat$yield * sqrt(3/1000) treatment <- interaction(wheat$tillage, wheat$summerCrop, wheat$manure, wheat$N, sep = "") ## Fit linear model mainEffects <- lm(yield.scaled ~ year + treatment, data = wheat) ## Convert to gnm object to allow addition of Mult() term svdStart <- residSVD(mainEffects, year, treatment, 3) bilinear1 <- update(asGnm(mainEffects), . ~ . + Mult(year, treatment), start = c(coef(mainEffects), svdStart[,1])) } \keyword{ models } \keyword{ regression } gnm/man/model.matrix.gnm.Rd0000644000176000001440000000201312515147523015311 0ustar ripleyusers\name{model.matrix.gnm} \alias{model.matrix.gnm} \title{ Local Design Matrix for a Generalized Nonlinear Model } \description{ This method extracts or evaluates a local design matrix for a generalized nonlinear model } \usage{ \method{model.matrix}{gnm}(object, coef = NULL, ...) } \arguments{ \item{object}{ an object of class \code{gnm}. } \item{coef}{ if specified, the vector of (non-eliminated) coefficients at which the local design matrix is evaluated. } \item{...}{ further arguments. } } \value{ If \code{coef = NULL}, the local design matrix with columns corresponding to the non-eliminated parameters evaluated at \code{coef(object)} (extracted from \code{object} if possible). Otherwise, the local design matrix evaluated at \code{coef}. } \author{ Heather Turner } \seealso{ \code{\link{gnm}}, \code{\link{model.matrix}} } \examples{ example(mentalHealth) model.matrix(RC1model) model.matrix(RC1model, coef = seq(coef(RC1model))) } \keyword{ models } \keyword{ regression } \keyword{ nonlinear } gnm/man/erikson.Rd0000644000176000001440000000616412515147523013613 0ustar ripleyusers\name{erikson} \alias{erikson} \docType{data} \title{Intergenerational Class Mobility in England/Wales, France and Sweden} \description{ Intergenerational class mobility among the male populations of England and Wales; France, and Sweden. } \usage{erikson} \format{ A table of counts, with classifying factors \code{origin} (father's class; levels \code{I}, \code{II}, \code{III}, \code{IVa}, \code{IVb}, \code{IVc}, \code{V/VI}, \code{VIIa}, \code{VIIb}) \code{destination} (son's class; levels as before), and \code{country} (son's country of residence; levels \code{EW}, \code{F}, \code{S}). } \source{ Hauser, R. M. (1984) Vertical Class Mobility in England, France and Sweden. \emph{Acta Sociol.}, \bold{27(2)}, 87-110. } \references{ Erikson, R., GoldThorpe, J. H. and Portocarero, L. (1982) Social Fluidity in Industrial Nations: England, France and Sweden. \emph{Brit. J. Sociol.} \bold{33(1)}, 1-34. Xie, Y. (1992) The Log-multiplicative Layer Effect Model for Comparing Mobility Tables. \emph{Am. Sociol. Rev.} \bold{57(3)}, 380-395. } \examples{ set.seed(1) ### Collapse to 7 by 7 table as in Erikson (1982) erikson <- as.data.frame(erikson) lvl <- levels(erikson$origin) levels(erikson$origin) <- levels(erikson$destination) <- c(rep(paste(lvl[1:2], collapse = " + "), 2), lvl[3], rep(paste(lvl[4:5], collapse = " + "), 2), lvl[6:9]) erikson <- xtabs(Freq ~ origin + destination + country, data = erikson) ### Fit the models given in first half of Table 3 of Xie (1992) ## Null association between origin and destination nullModel <- gnm(Freq ~ country*origin + country*destination, family = poisson, data = erikson) ## Full interaction, common to all countries commonInteraction <- update(nullModel, ~ . + origin:destination) ## Full Interaction, different multiplier for each country multInteraction <- update(nullModel, ~ . + Mult(Exp(country), origin:destination)) ### Create array of interaction levels as in Table 2 of Xie (1992) levelMatrix <- matrix(c(2, 3, 4, 6, 5, 6, 6, 3, 3, 4, 6, 4, 5, 6, 4, 4, 2, 5, 5, 5, 5, 6, 6, 5, 1, 6, 5, 2, 4, 4, 5, 6, 3, 4, 5, 5, 4, 5, 5, 3, 3, 5, 6, 6, 5, 3, 5, 4, 1), 7, 7, byrow = TRUE) ### Fit models in second half of Table 3 in Xie (1992) ## Interaction specified by levelMatrix, common to all countries commonTopo <- update(nullModel, ~ . + Topo(origin, destination, spec = levelMatrix)) ## Interaction specified by levelMatrix, different multiplier for ## each country multTopo <- update(nullModel, ~ . + Mult(Exp(country), Topo(origin, destination, spec = levelMatrix))) ## Interaction specified by levelMatrix, different effects for ## each country separateTopo <- update(nullModel, ~ . + country:Topo(origin, destination, spec = levelMatrix)) } \keyword{datasets} gnm/man/predict.gnm.Rd0000644000176000001440000000712112515147523014345 0ustar ripleyusers\name{predict.gnm} \alias{predict.gnm} \title{ Predict Method for Generalized Nonlinear Models } \description{ Obtains predictions and optionally estimates standard errors of those predictions from a fitted generalized nonlinear model object. } \usage{ \method{predict}{gnm}(object, newdata = NULL, type = c("link", "response", "terms"), se.fit = FALSE, dispersion = NULL, terms = NULL, na.action = na.exclude, ...) } %- maybe also 'usage' for other objects documented here. \arguments{ \item{object}{ a fitted object of class inheriting from \code{"gnm"}. } \item{newdata}{ optionally, a data frame in which to look for variables with which to predict. If omitted, the fitted predictors are used. } \item{type}{ the type of prediction required. The default is on the scale of the predictors; the alternative \code{"response"} is on the scale of the response variable. Thus for a default binomial model the default predictions are of log-odds (probabilities on logit scale) and \code{type = "response"} gives the predicted probabilities. The \code{"terms"} option returns a matrix giving the fitted values of each term in the model formula on the predictor scale. The value of this argument can be abbreviated. } \item{se.fit}{ logical switch indicating if standard errors are required. } \item{dispersion}{ the dispersion of the fit to be assumed in computing the standard errors. If omitted, that returned by \code{summary} applied to the object is used. } \item{terms}{ with \code{type="terms"} by default all terms are returned. A character vector specifies which terms are to be returned } \item{na.action}{ function determining what should be done with missing values in \code{newdata}. The default is to predict \code{NA}. } \item{\dots}{ further arguments passed to or from other methods. } } \details{ If \code{newdata} is omitted the predictions are based on the data used for the fit. In that case how cases with missing values in the original fit is determined by the \code{na.action} argument of that fit. If \code{na.action = na.omit} omitted cases will not appear in the residuals, whereas if \code{na.action = na.exclude} they will appear (in predictions and standard errors), with residual value \code{NA}. See also \code{\link{napredict}}. } \value{ If \code{se = FALSE}, a vector or matrix of predictions. If \code{se = TRUE}, a list with components \item{ fit }{ predictions.} \item{ se.fit }{ estimated standard errors.} \item{ residual.scale }{ a scalar giving the square root of the dispersion used in computing the standard errors.} } \references{ Chambers, J. M. and Hastie, T. J. (1992) \emph{Statistical Models in S }} \author{ Heather Turner } \note{Variables are first looked for in 'newdata' and then searched for in the usual way (which will include the environment of the formula used in the fit). A warning will be given if the variables found are not of the same length as those in 'newdata' if it was supplied.} \seealso{ \code{\link{gnm}} } \examples{ set.seed(1) ## Fit an association model with homogeneous row-column effects RChomog <- gnm(Freq ~ origin + destination + Diag(origin, destination) + MultHomog(origin, destination), family = poisson, data = occupationalStatus) ## Fitted values (expected counts) predict(RChomog, type = "response", se.fit = TRUE) ## Fitted values on log scale predict(RChomog, type = "link", se.fit = TRUE) } \keyword{ models } \keyword{ nonlinear } gnm/man/Topo.Rd0000644000176000001440000000527212515147523013061 0ustar ripleyusers\name{Topo} \alias{Topo} \title{ Topological Interaction of Factors } \description{ Given two or more factors \code{Topo} creates an interaction factor as specified by an array of levels, which may be arbitrarily structured. } \usage{ Topo(..., spec = NULL) } \arguments{ \item{\dots}{ two or more factors } \item{spec}{ an array of levels, with dimensions corresponding to the number of levels of each factor in the interaction } } \value{ A factor of levels extracted from the levels array given in \code{spec}, using the given factors as index variables. } \references{ Erikson, R., GoldThorpe, J. H. and Portocarero, L. (1982) Social Fluidity in Industrial Nations: England, France and Sweden. \emph{Brit. J. Sociol.} \bold{33(1)}, 1-34. Xie, Y. (1992) The Log-multiplicative Layer Effect Model for Comparing Mobility Tables. \emph{Am. Sociol. Rev.} \bold{57(3)}, 380-395. } \author{ David Firth } \seealso{ \code{\link{Symm}} and \code{\link{Diag}} for special cases } \examples{ set.seed(1) ### Collapse to 7 by 7 table as in Erikson (1982) erikson <- as.data.frame(erikson) lvl <- levels(erikson$origin) levels(erikson$origin) <- levels(erikson$destination) <- c(rep(paste(lvl[1:2], collapse = " + "), 2), lvl[3], rep(paste(lvl[4:5], collapse = " + "), 2), lvl[6:9]) erikson <- xtabs(Freq ~ origin + destination + country, data = erikson) ### Create array of interaction levels as in Table 2 of Xie (1992) levelMatrix <- matrix(c(2, 3, 4, 6, 5, 6, 6, 3, 3, 4, 6, 4, 5, 6, 4, 4, 2, 5, 5, 5, 5, 6, 6, 5, 1, 6, 5, 2, 4, 4, 5, 6, 3, 4, 5, 5, 4, 5, 5, 3, 3, 5, 6, 6, 5, 3, 5, 4, 1), 7, 7, byrow = TRUE) ### Fit the levels models given in Table 3 of Xie (1992) ## Null association between origin and destination nullModel <- gnm(Freq ~ country:origin + country:destination, family = poisson, data = erikson) ## Interaction specified by levelMatrix, common to all countries commonTopo <- update(nullModel, ~ . + Topo(origin, destination, spec = levelMatrix)) ## Interaction specified by levelMatrix, different multiplier for ## each country multTopo <- update(nullModel, ~ . + Mult(Exp(country), Topo(origin, destination, spec = levelMatrix))) ## Interaction specified by levelMatrix, different effects for ## each country separateTopo <- update(nullModel, ~ . + country:Topo(origin, destination, spec = levelMatrix)) } \keyword{ models } gnm/man/se.Rd0000644000176000001440000000542212515147523012544 0ustar ripleyusers\name{se} \alias{se} \title{ Standard Errors of Linear Parameter Combinations in gnm Models } \description{ Computes approximate standard errors for (a selection of) individual parameters or one or more linear combinations of the parameters in a \code{\link{gnm}} (generalized nonlinear model) object. By default, a check is made first on the estimability of each specified combination. } \usage{ se(model, estimate = NULL, checkEstimability = TRUE, Vcov = NULL, dispersion = NULL, ...) } \arguments{ \item{model}{ a model object of class \code{"gnm"}.} \item{estimate}{ (optional) specifies parameters or linear combinations of parameters for which to find standard errors. In the first case either a character vector of names, a numeric vector of indices or \code{"[?]"} to select from a Tk dialog. In the second case coefficients given as a vector or the rows of a matrix, such that \code{NROW(estimate)} is equal to \code{length(coef(model))}. If missing, standard errors are returned for all (non-eliminated) parameters in the model.} \item{checkEstimability}{ logical: should the estimability of all specified combinations be checked?} \item{Vcov}{ either NULL, or a matrix } \item{dispersion}{ either NULL, or a positive number } \item{\dots}{ possible further arguments for \code{\link{checkEstimable}}. } } \note{ In the case where \code{estimate} is a numeric vector, \code{se} will assume that indices have been specified if all the values of \code{estimate} are in \code{seq(length(coef(model))}. Where both \code{Vcov} and \code{dispersion} are supplied, the variance-covariance matrix of estimated model coefficients is taken to be \code{Vcov * dispersion}. } \value{ A data frame with two columns: \item{Estimate }{The estimated parameter combinations} \item{Std. Error }{Their estimated standard errors} If available, the column names of \code{coefMatrix} will be used to name the rows. } \author{ David Firth and Heather Turner } \seealso{ \code{\link{gnm}}, \code{\link{getContrasts}}, \code{\link{checkEstimable}}, \code{\link{ofInterest}}} \examples{ set.seed(1) ## Fit the "UNIDIFF" mobility model across education levels unidiff <- gnm(Freq ~ educ*orig + educ*dest + Mult(Exp(educ), orig:dest), ofInterest = "[.]educ", family = poisson, data = yaish, subset = (dest != 7)) ## Deviance is 200.3 ## Get estimate and se for the contrast between educ4 and educ5 in the ## UNIDIFF multiplier mycontrast <- numeric(length(coef(unidiff))) mycontrast[ofInterest(unidiff)[4:5]] <- c(1, -1) se(unidiff, mycontrast) ## Get all of the contrasts with educ5 in the UNIDIFF multipliers getContrasts(unidiff, rev(ofInterest(unidiff))) } \keyword{ models } \keyword{ regression } \keyword{ nonlinear } gnm/man/wedderburn.Rd0000644000176000001440000000503112515147523014272 0ustar ripleyusers\name{wedderburn} \alias{wedderburn} \title{ Wedderburn Quasi-likelihood Family } \description{ Creates a \code{\link{family}} object for use with \code{\link{glm}}, \code{\link{gnm}}, etc., for the variance function \eqn{[\mu(1-\mu)]^2} introduced by Wedderburn (1974) for response values in [0,1]. } \usage{ wedderburn(link = "logit") } \arguments{ \item{link}{ The name of a link function. Allowed are "logit", "probit" and "cloglog". } } \value{ An object of class \code{\link{family}}. } \references{ Gabriel, K R (1998). Generalised bilinear regression. \emph{Biometrika} \bold{85}, 689--700. McCullagh, P and Nelder, J A (1989). \emph{Generalized Linear Models} (2nd ed). Chapman and Hall. Wedderburn, R W M (1974). Quasilikelihood functions, generalized linear models and the Gauss-Newton method. \emph{Biometrika} \bold{61}, 439--47. } \author{ Modification of \code{\link{binomial}} by the R Core Team. Adapted for the Wedderburn quasi-likelihood family by David Firth. } \note{ The reported deviance involves an arbitrary constant (see McCullagh and Nelder, 1989, p330); for estimating dispersion, use the Pearson chi-squared statistic instead. } \seealso{ \code{\link{glm}}, \code{\link{gnm}}, \code{\link{family}} } \examples{ set.seed(1) ### Use data from Wedderburn (1974), see ?barley ### Fit Wedderburn's logit model with variance proportional to the ### square of mu(1-mu) logitModel <- glm(y ~ site + variety, family = wedderburn, data = barley) fit <- fitted(logitModel) print(sum((barley$y - fit)^2 / (fit * (1-fit))^2)) ## Agrees with the chi-squared value reported in McCullagh and Nelder ## (1989, p331), which differs slightly from Wedderburn's reported value. ### Fit the biplot model as in Gabriel (1998, p694) biplotModel <- gnm(y ~ -1 + instances(Mult(site, variety), 2), family = wedderburn, data = barley) barleySVD <- svd(matrix(biplotModel$predictors, 10, 9)) A <- sweep(barleySVD$v, 2, sqrt(barleySVD$d), "*")[, 1:2] B <- sweep(barleySVD$u, 2, sqrt(barleySVD$d), "*")[, 1:2] ## These are essentially A and B as in Gabriel (1998, p694), from which ## the biplot is made by plot(rbind(A, B), pch = c(LETTERS[1:9], as.character(1:9), "X")) ### Fit the double-additive model as in Gabriel (1998, p697) variety.binary <- factor(match(barley$variety, c(2,3,6), nomatch = 0) > 0, labels = c("Rest", "2,3,6")) doubleAdditive <- gnm(y ~ variety + Mult(site, variety.binary), family = wedderburn, data = barley) } \keyword{ models } gnm/man/meanResiduals.Rd0000644000176000001440000000604512515147523014733 0ustar ripleyusers\name{meanResiduals} \alias{meanResiduals} \title{Average Residuals within Factor Levels} \description{ Computes the mean working residuals from a model fitted using Iterative Weighted Least Squares for each level of a factor or interaction of factors. } \usage{meanResiduals(object, by, standardized=TRUE, as.table=TRUE, ...)} \arguments{ \item{object}{model object for which \code{object$residuals} gives the working residuals and \code{object$weights} gives the working weights.} \item{by}{either a formula specifying a factor or interaction of factors (recommended), or a list of factors (the elements of which must correspond exactly to observations in the model frame). When a list of factors is specified, their interaction is used to specify the grouping factor.} \item{standardized}{logical: if \code{TRUE}, the mean residuals are standardized to be approximately standard normal.} \item{as.table}{logical: logical: if \code{TRUE} and \code{by} specifies an interaction of factors, the result is returned as a table cross-classified by these factors.} \item{...}{currently ignored} } \details{ For level \eqn{i} of the grouping factor \eqn{A} the mean working residual is defined as \deqn{\frac{r_{ij} * w_{ij}}{\sum_{j = 1}^{n_i} w_{ij}}}{ (r_ij * w_ij)/(sum_(j = 1)^(n_i) w_ij)} where \eqn{r_{ij}}{r_ij} is the \eqn{j}'th residual for level \eqn{i}, \eqn{w_{ij}}{w_ij} is the corresponding working weight and \eqn{n_i} is the number of observations for level \eqn{i}. The denominator gives the weight corresponding to mean residual. For non-aggregated residuals, i.e. when the factor has one level per observation, the residuals are the same as Pearson residuals. } \author{Heather Turner} \value{An object of class \code{"meanResiduals"}, for which \code{print} and \code{summary} methods are provided. A \code{"meanResiduals"} object is a list containing the following elements: \item{ call }{ the call used to create the model object from which the mean residuals are derived. } \item{ by }{ a label for the grouping factor. } \item{ residuals }{ the mean residuals. } \item{ df }{ the degrees of freedom associated with the mean residuals. } \item{ standardized }{ the \code{standardized} argument. } \item{ weights }{ the weights corresponding to the mean residuals. } } \examples{ ## Fit a conditional independence model, leaving out ## the uninformative subtable for dest == 7: CImodel <- gnm(Freq ~ educ*orig + educ*dest, family = poisson, data = yaish, subset = (dest != 7)) ## compute mean residuals over origin and destination meanRes <- meanResiduals(CImodel, ~ orig:dest) meanRes summary(meanRes) \dontrun{ ## requires vcdExtra package ## display mean residuals for origin and destination library(vcdExtra) mosaic(CImodel, ~orig+dest) } ## non-aggregated residuals res1 <- meanResiduals(CImodel, ~ educ:orig:dest) res2 <- residuals(CImodel, type = "pearson") all.equal(as.numeric(res1), as.numeric(res2)) } \keyword{ models } \keyword{ regression } \keyword{ nonlinear } gnm/man/MultHomog.Rd0000644000176000001440000000570012515147523014047 0ustar ripleyusers\name{MultHomog} \alias{MultHomog} \title{Specify a Multiplicative Interaction with Homogeneous Effects in a gnm Model Formula} \description{ A function of class \code{"nonlin"} to specify a multiplicative interaction with homogeneous effects in the formula argument to \code{\link{gnm}}. } \usage{ MultHomog(..., inst = NULL) } \arguments{ \item{\dots}{ a comma-separated list of two or more factors. } \item{inst}{ (optional) an integer specifying the instance number of the term. } } \details{ \code{MultHomog} specifies instances of a multiplicative interaction in which the constituent multipliers are the effects of two or more factors and the effects of these factors are constrained to be equal when the factor levels are equal. Thus the interaction effect would be \deqn{\gamma_i\gamma_j...}{gamma_i gamma_j ...} for an observation with level \eqn{i} of the first factor, level \eqn{j} of the second factor and so on, where \eqn{\gamma_l}{gamma_l} is the effect for level \eqn{l} of the homogeneous multiplicative factor. If the factors passed to \code{MultHomog} do not have exactly the same levels, the set of levels is taken to be the union of the factor levels, sorted into increasing order. } \value{ A list with the anticipated components of a \code{"nonlin"} function: \item{ predictors }{ the factors passed to \code{MultHomog}} \item{ common }{ an index to specify that common effects are to be estimated across the factors } \item{ term }{ a function to create a deparsed mathematical expression of the term, given labels for the predictors.} \item{ call }{ the call to use as a prefix for parameter labels. } } \references{ Goodman, L. A. (1979) Simple Models for the Analysis of Association in Cross-Classifications having Ordered Categories. \emph{J. Am. Stat. Assoc.}, \bold{74(367)}, 537-552. } \note{Currently, \code{MultHomog} can only be used to specify a one-dimensional interaction. See examples for a workaround to specify interactions with more than one dimension. } \author{ Heather Turner } \seealso{\code{\link{gnm}}, \code{\link{formula}}, \code{\link{instances}}, \code{\link{nonlin.function}}, \code{\link{Mult}}} \examples{ set.seed(1) ### Fit an association model with homogeneous row-column effects rc1 <- gnm(Freq ~ r + c + Diag(r,c) + MultHomog(r, c), family = poisson, data = friend) rc1 \dontrun{ ### Extend to two-component interaction rc2 <- update(rc1, . ~ . + MultHomog(r, c, inst = 2), etastart = rc1$predictors) rc2 } ### For factors with a large number of levels, save time by ### setting diagonal elements to NA rather than fitting exactly; ### skipping start-up iterations may also save time dat <- as.data.frame(friend) id <- with(dat, r == c) dat[id,] <- NA rc2 <- gnm(Freq ~ r + c + instances(MultHomog(r, c), 2), family = poisson, data = dat, iterStart = 0) } \keyword{ models } \keyword{ regression } \keyword{ nonlinear } gnm/man/voting.Rd0000644000176000001440000000412212515147523013437 0ustar ripleyusers\name{voting} \alias{voting} \docType{data} \title{Data on Social Mobility and the Labour Vote} \description{ Voting data from the 1987 British general election, cross-classified by the class of the head of household and the class of their father. } \usage{voting} \format{ A data frame with 25 observations on the following 4 variables. \describe{ \item{\code{percentage}}{the percentage of the cell voting Labour.} \item{\code{total}}{the cell count.} \item{\code{origin}}{a factor describing the father's class with levels \code{1:5}.} \item{\code{destination}}{a factor describing the head of household's class with levels \code{1:5}.} } } \source{ Clifford, P. and Heath, A. F. (1993) The Political Consequences of Social Mobility. \emph{J. Roy. Stat. Soc. A}, \bold{156(1)}, 51-61. } \examples{ ### Examples from Clifford and Heath paper ### (Results differ slightly - possible transcription error in ### published data?) set.seed(1) ## reconstruct counts voting Labour/non-Labour count <- with(voting, percentage/100 * total) yvar <- cbind(count, voting$total - count) ## fit diagonal reference model with constant weights classMobility <- gnm(yvar ~ -1 + Dref(origin, destination), family = binomial, data = voting) DrefWeights(classMobility) ## create factors indicating movement in and out of salariat (class 1) upward <- with(voting, origin != 1 & destination == 1) downward <- with(voting, origin == 1 & destination != 1) ## fit separate weights for the "socially mobile" groups socialMobility <- gnm(yvar ~ -1 + Dref(origin, destination, delta = ~ 1 + downward + upward), family = binomial, data = voting) DrefWeights(socialMobility) ## fit separate weights for downwardly mobile groups only downwardMobility <- gnm(yvar ~ -1 + Dref(origin, destination, delta = ~ 1 + downward), family = binomial, data = voting) DrefWeights(downwardMobility) } \keyword{datasets} gnm/man/parameters.Rd0000644000176000001440000000172012515147523014275 0ustar ripleyusers\name{parameters} \alias{parameters} \title{ Extract Constrained and Estimated Parameters from a gnm Object} \description{ A function to extract non-eliminated parameters from a \code{"gnm"} object, including parameters that were constrained. } \usage{ parameters(object) } \arguments{ \item{object}{ an object of class \code{"gnm"}. } } \details{ \code{parameters} acts like \code{coefficients} except that for constrained parameters, the value at which the parameter was constrained is returned instead of \code{NA}. } \value{ A vector of parameters. } \author{ Heather Turner } \seealso{ \code{\link{coefficients}}, \code{\link{gnm}} } \examples{ RChomog <- gnm(Freq ~ origin + destination + Diag(origin, destination) + MultHomog(origin, destination), family = poisson, data = occupationalStatus, ofInterest = "MultHomog", constrain = "MultHomog.*1") coefficients(RChomog) parameters(RChomog) } \keyword{ models } gnm/man/backPain.Rd0000644000176000001440000000617612515147523013654 0ustar ripleyusers\name{backPain} \alias{backPain} \docType{data} \title{ Data on Back Pain Prognosis, from Anderson (1984) } \description{ Data from a study of patients suffering from back pain. Prognostic variables were recorded at presentation and progress was categorised three weeks after treatment. } \usage{backPain} \format{ A data frame with 101 observations on the following 4 variables. \describe{ \item{x1}{length of previous attack.} \item{x2}{pain change.} \item{x3}{lordosis.} \item{pain}{an ordered factor describing the progress of each patient with levels \code{worse} < \code{same} < \code{slight.improvement} < \code{moderate.improvement} < \code{marked.improvement} < \code{complete.relief}. } } } \source{ \url{http://ideas.repec.org/c/boc/bocode/s419001.html} } \references{ Anderson, J. A. (1984) Regression and Ordered Categorical Variables. \emph{J. R. Statist. Soc. B}, \bold{46(1)}, 1-30. } \examples{ set.seed(1) summary(backPain) ### Re-express as count data backPainLong <- expandCategorical(backPain, "pain") ### Fit models described in Table 5 of Anderson (1984) ### Logistic family models noRelationship <- gnm(count ~ pain, eliminate = id, family = "poisson", data = backPainLong) ## stereotype model oneDimensional <- update(noRelationship, ~ . + Mult(pain, x1 + x2 + x3)) ## multinomial logistic threeDimensional <- update(noRelationship, ~ . + pain:(x1 + x2 + x3)) ### Models to determine distinguishability in stereotype model ## constrain scale of category-specific multipliers oneDimensional <- update(noRelationship, ~ . + Mult(pain, offset(x1) + x2 + x3)) ## obtain identifiable contrasts & id possibly indistinguishable slopes getContrasts(oneDimensional, pickCoef(oneDimensional, "[.]pain")) \dontrun{ ## (this part not needed for package testing) ## fit simpler models and compare .pain <- backPainLong$pain levels(.pain)[2:3] <- paste(levels(.pain)[2:3], collapse = " | ") fiveGroups <- update(noRelationship, ~ . + Mult(.pain, x1 + x2 + x3)) levels(.pain)[4:5] <- paste(levels(.pain)[4:5], collapse = " | ") fourGroups <- update(fiveGroups) levels(.pain)[2:3] <- paste(levels(.pain)[2:3], collapse = " | ") threeGroups <- update(fourGroups) ### Grouped continuous model, aka proportional odds model library(MASS) sixCategories <- polr(pain ~ x1 + x2 + x3, data = backPain) ### Obtain number of parameters and log-likelihoods for equivalent ### multinomial models as presented in Anderson (1984) logLikMultinom <- function(model, size){ object <- get(model) if (inherits(object, "gnm")) { l <- sum(object$y * log(object$fitted/size)) c(nParameters = object$rank - nlevels(object$eliminate), logLikelihood = l) } else c(nParameters = object$edf, logLikelihood = -deviance(object)/2) } size <- tapply(backPainLong$count, backPainLong$id, sum)[backPainLong$id] models <- c("threeDimensional", "oneDimensional", "noRelationship", "fiveGroups", "fourGroups", "threeGroups", "sixCategories") t(sapply(models, logLikMultinom, size)) } } \keyword{datasets} gnm/man/cautres.Rd0000644000176000001440000000567512515147523013615 0ustar ripleyusers\name{cautres} \alias{cautres} \docType{data} \title{ Data on Class, Religion and Vote in France} \description{ A 4-way contingency table of vote by class by religion in four French elections } \usage{cautres} \format{ A table of counts, with classifying factors \code{vote} (levels \code{1:2}), \code{class} (levels \code{1:6}) and \code{religion} (levels \code{1:4}) and \code{election} (levels \code{1:4}). } \source{ Bruno Cautres } \references{ Cautres, B, Heath, A F and Firth, D (1998). Class, religion and vote in Britain and France. \emph{La Lettre de la Maison Francaise} \bold{8}. } \examples{ set.seed(1) ## Fit a "double UNIDIFF" model with the religion-vote and class-vote ## interactions both modulated by nonnegative election-specific multipliers doubleUnidiff <- gnm(Freq ~ election*vote + election*class*religion + Mult(Exp(election), religion:vote) + Mult(Exp(election), class:vote), family = poisson, data = cautres) ## Deviance should be 133.04 ## Examine the multipliers of the class-vote log odds ratios ofInterest(doubleUnidiff) <- pickCoef(doubleUnidiff, "class:vote[).]") coef(doubleUnidiff) ## Coefficients of interest: ## Mult(Exp(.), class:vote).election1 ## -0.38357138 ## Mult(Exp(.), class:vote).election2 ## 0.29816599 ## Mult(Exp(.), class:vote).election3 ## 0.06580307 ## Mult(Exp(.), class:vote).election4 ## -0.02174104 ## Re-parameterize by setting Mult2.Factor1.election1 to zero getContrasts(doubleUnidiff, ofInterest(doubleUnidiff)) ## estimate SE ## Mult(Exp(.), class:vote).election1 0.0000000 0.0000000 ## Mult(Exp(.), class:vote).election2 0.6817374 0.2401644 ## Mult(Exp(.), class:vote).election3 0.4493745 0.2473521 ## Mult(Exp(.), class:vote).election4 0.3618301 0.2534754 ## quasiSE quasiVar ## Mult(Exp(.), class:vote).election1 0.22854401 0.052232363 ## Mult(Exp(.), class:vote).election2 0.07395886 0.005469913 ## Mult(Exp(.), class:vote).election3 0.09475938 0.008979340 ## Mult(Exp(.), class:vote).election4 0.10934798 0.011956981 ## Same thing but with election 4 as reference category: getContrasts(doubleUnidiff, rev(ofInterest(doubleUnidiff))) ## estimate SE ## Mult(Exp(.), class:vote).election4 0.00000000 0.0000000 ## Mult(Exp(.), class:vote).election3 0.08754436 0.1446833 ## Mult(Exp(.), class:vote).election2 0.31990727 0.1320022 ## Mult(Exp(.), class:vote).election1 -0.36183013 0.2534754 ## quasiSE quasiVar ## Mult(Exp(.), class:vote).election4 0.10934798 0.011956981 ## Mult(Exp(.), class:vote).election3 0.09475938 0.008979340 ## Mult(Exp(.), class:vote).election2 0.07395886 0.005469913 ## Mult(Exp(.), class:vote).election1 0.22854401 0.052232363 } \keyword{datasets} gnm/man/Dref.Rd0000644000176000001440000002221612515147523013015 0ustar ripleyusers\name{Dref} \alias{Dref} \alias{DrefWeights} \title{Specify a Diagonal Reference Term in a gnm Model Formula} \description{ Dref is a function of class \code{"nonlin"} to specify a diagonal reference term in the formula argument to \code{\link{gnm}}. } \usage{ Dref(..., delta = ~ 1) } \arguments{ \item{\dots}{a comma-separated list of two or more factors.} \item{delta}{a formula with no left-hand-side specifying the model for each factor weight.} } \details{ \code{Dref} specifies diagonal reference terms as introduced by Sobel (1981, 1985). Such terms comprise an additive component for each factor of the form \deqn{w_f\gamma_l}{w_f gamma_l} where \eqn{w_f} is the weight for factor \eqn{f}, \eqn{\gamma_l}{gamma_l} is the diagonal effect for level \eqn{l} and \eqn{l} is the level of factor \eqn{f} for the given data point. The weights are constrained to be nonnegative and to sum to one as follows \deqn{w_f = \frac{e^{\delta_f}}{\sum_i e^{\delta_i}}}{ w_f = exp(delta_f)/sum_i(exp(delta_i))} and the \eqn{\delta_f}{delta_f} are modelled as specified by the \code{delta} argument (constant weights by default). The returned parameters are those in the model for \eqn{\delta_f}{delta_f}, rather than the implied weights \eqn{w_f}. The \code{DrefWeights} function will take a fitted gnm model and return the weights \eqn{w_f}, along with their standard errors. If the factors passed to \code{Dref} do not have exactly the same levels, the set of levels in the diagonal reference term is taken to be the union of the factor levels, sorted into increasing order. } \value{ A list with the anticipated components of a "nonlin" function: \item{ predictors }{ the factors passed to \code{Dref} and the formulae for the weights. } \item{ common }{ an index to specify that common effects are to be estimated across the factors. } \item{ term }{ a function to create a deparsed mathematical expression of the term, given labels for the predictors.} \item{ start }{ a function to generate starting values for the parameters.} \item{ call }{ the call to use as a prefix for parameter labels. } } \references{Sobel, M. E. (1981), Diagonal mobility models: A substantively motivated class of designs for the analysis of mobility effects. \emph{American Sociological Review} \bold{46}, 893--906. Sobel, M. E. (1985), Social mobility and fertility revisited: Some new models for the analysis of the mobility effects hypothesis. \emph{American Sociological Review} \bold{50}, 699--712. Clifford, P. and Heath, A. F. (1993) The Political Consequences of Social Mobility. \emph{J. Roy. Stat. Soc. A}, \bold{156(1)}, 51-61. Van der Slik, F. W. P., De Graaf, N. D and Gerris, J. R. M. (2002) Conformity to Parental Rules: Asymmetric Influences of Father's and Mother's Levels of Education. \emph{European Sociological Review} \bold{18(4)}, 489 -- 502. } \author{ Heather Turner } \seealso{\code{\link{gnm}}, \code{\link{formula}}, \code{\link{nonlin.function}}} \examples{ ### Examples from Clifford and Heath paper ### (Results differ slightly - possible transcription error in ### published data?) set.seed(1) ## reconstruct counts voting Labour/non-Labour count <- with(voting, percentage/100 * total) yvar <- cbind(count, voting$total - count) ## fit diagonal reference model with constant weights classMobility <- gnm(yvar ~ -1 + Dref(origin, destination), family = binomial, data = voting) DrefWeights(classMobility) ## create factors indicating movement in and out of salariat (class 1) upward <- with(voting, origin != 1 & destination == 1) downward <- with(voting, origin == 1 & destination != 1) ## fit separate weights for the "socially mobile" groups socialMobility <- gnm(yvar ~ -1 + Dref(origin, destination, delta = ~ 1 + downward + upward), family = binomial, data = voting) DrefWeights(socialMobility) ## fit separate weights for downwardly mobile groups only downwardMobility <- gnm(yvar ~ -1 + Dref(origin, destination, delta = ~ 1 + downward), family = binomial, data = voting) DrefWeights(downwardMobility) \dontrun{ ### Examples from Van der Slik paper ### For illustration only - data not publically available ### Using data in data.frame named 'conformity', with variables ### MCFM - mother's conformity score ### FCFF - father's conformity score ### MOPLM - a factor describing the mother's education with 7 levels ### FOPLF - a factor describing the father's education with 7 levels ### AGEM - mother's birth cohort ### MRMM - mother's traditional role model ### FRMF - father's traditional role model ### MWORK - mother's employment ### MFCM - mother's family conflict score ### FFCF - father's family conflict score set.seed(1) ## Models for mothers' conformity score as specified in Figure 1 A <- gnm(MCFM ~ -1 + AGEM + MRMM + FRMF + MWORK + MFCM + Dref(MOPLM, FOPLF), family = gaussian, data = conformity, verbose = FALSE) A ## Call: ## gnm(formula = MCFM ~ -1 + AGEM + MRMM + FRMF + MWORK + MFCM + ## Dref(MOPLM, FOPLF), family = gaussian, data = conformity, ## verbose = FALSE) ## ## Coefficients: ## AGEM MRMM ## 0.06363 -0.32425 ## FRMF MWORK ## -0.25324 -0.06430 ## MFCM Dref(MOPLM, FOPLF)delta1 ## -0.06043 -0.33731 ## Dref(MOPLM, FOPLF)delta2 Dref(., .).MOPLM|FOPLF1 ## -0.02505 4.95121 ## Dref(., .).MOPLM|FOPLF2 Dref(., .).MOPLM|FOPLF3 ## 4.86329 4.86458 ## Dref(., .).MOPLM|FOPLF4 Dref(., .).MOPLM|FOPLF5 ## 4.72343 4.43516 ## Dref(., .).MOPLM|FOPLF6 Dref(., .).MOPLM|FOPLF7 ## 4.18873 4.43378 ## ## Deviance: 425.3389 ## Pearson chi-squared: 425.3389 ## Residual df: 576 ## Weights as in Table 4 DrefWeights(A) ## Refitting with parameters of first Dref weight constrained to zero ## $MOPLM ## weight se ## 0.4225636 0.1439829 ## ## $FOPLF ## weight se ## 0.5774364 0.1439829 F <- gnm(MCFM ~ -1 + AGEM + MRMM + FRMF + MWORK + MFCM + Dref(MOPLM, FOPLF, delta = ~1 + MFCM), family = gaussian, data = conformity, verbose = FALSE) F ## Call: ## gnm(formula = MCFM ~ -1 + AGEM + MRMM + FRMF + MWORK + MFCM + ## Dref(MOPLM, FOPLF, delta = ~1 + MFCM), family = gaussian, ## data = conformity, verbose = FALSE) ## ## ## Coefficients: ## AGEM ## 0.05818 ## MRMM ## -0.32701 ## FRMF ## -0.25772 ## MWORK ## -0.07847 ## MFCM ## -0.01694 ## Dref(MOPLM, FOPLF, delta = ~ . + MFCM).delta1(Intercept) ## 1.03515 ## Dref(MOPLM, FOPLF, delta = ~ 1 + .).delta1MFCM ## -1.77756 ## Dref(MOPLM, FOPLF, delta = ~ . + MFCM).delta2(Intercept) ## -0.03515 ## Dref(MOPLM, FOPLF, delta = ~ 1 + .).delta2MFCM ## 2.77756 ## Dref(., ., delta = ~ 1 + MFCM).MOPLM|FOPLF1 ## 4.82476 ## Dref(., ., delta = ~ 1 + MFCM).MOPLM|FOPLF2 ## 4.88066 ## Dref(., ., delta = ~ 1 + MFCM).MOPLM|FOPLF3 ## 4.83969 ## Dref(., ., delta = ~ 1 + MFCM).MOPLM|FOPLF4 ## 4.74850 ## Dref(., ., delta = ~ 1 + MFCM).MOPLM|FOPLF5 ## 4.42020 ## Dref(., ., delta = ~ 1 + MFCM).MOPLM|FOPLF6 ## 4.17957 ## Dref(., ., delta = ~ 1 + MFCM).MOPLM|FOPLF7 ## 4.40819 ## ## Deviance: 420.9022 ## Pearson chi-squared: 420.9022 ## Residual df: 575 ## ## ## Standard error for MFCM == 1 lower than reported by Van der Slik et al DrefWeights(F) ## Refitting with parameters of first Dref weight constrained to zero ## $MOPLM ## MFCM weight se ## 1 1 0.02974675 0.2277711 ## 2 0 0.74465224 0.2006916 ## ## $FOPLF ## MFCM weight se ## 1 1 0.9702532 0.2277711 ## 2 0 0.2553478 0.2006916 } } \keyword{ models } \keyword{ regression } \keyword{ nonlinear } gnm/man/MPinv.Rd0000644000176000001440000000337212515147523013170 0ustar ripleyusers\name{MPinv} \alias{MPinv} \title{ Moore-Penrose Pseudoinverse of a Real-valued Matrix } \description{ Computes the Moore-Penrose generalized inverse. } \usage{ MPinv(mat, tolerance = 100*.Machine$double.eps, rank = NULL, method = "svd") } \arguments{ \item{mat}{ a real matrix.} \item{tolerance}{ A positive scalar which determines the tolerance for detecting zeroes among the singular values. } \item{rank}{Either \code{NULL}, in which case the rank of \code{mat} is determined numerically; or an integer specifying the rank of \code{mat} if it is known. No check is made on the validity of any non-\code{NULL} value.} \item{method}{Character, one of \code{"svd", "chol"}. The specification \code{method = "chol"} is valid only for symmetric matrices. } } \details{ Real-valuedness is not checked, neither is symmetry when \code{method = "chol"}. } \value{ A matrix, with an additional attribute named \code{"rank"} containing the numerically determined rank of the matrix. } \references{ Harville, D. A. (1997). \emph{Matrix Algebra from a Statistician's Perspective}. New York: Springer. Courrieu, P. (2005). Fast computation of Moore-Penrose inverse matrices. \emph{Neural Information Processing} \bold{8}, 25--29 } \author{ David Firth and Heather Turner } \seealso{\code{\link[MASS]{ginv}}} \examples{ A <- matrix(c(1, 1, 0, 1, 1, 0, 2, 3, 4), 3, 3) B <- MPinv(A) A \%*\% B \%*\% A - A # essentially zero B \%*\% A \%*\% B - B # essentially zero attr(B, "rank") # here 2 ## demonstration that "svd" and "chol" deliver essentially the same ## results for symmetric matrices: A <- crossprod(A) MPinv(A) - MPinv(A, method = "chol") ## (essentially zero) } \keyword{ array } gnm/man/residSVD.Rd0000644000176000001440000000416412515147523013622 0ustar ripleyusers\name{residSVD} \alias{residSVD} \title{ Multiplicative Approximation of Model Residuals } \description{ This function uses the first \code{d} components of the singular value decomposition in order to approximate a vector of model residuals by a sum of \code{d} multiplicative terms, with the multiplicative structure determined by two specified factors. It applies to models of class \code{lm}, \code{glm} or \code{gnm}. } \usage{ residSVD(model, fac1, fac2, d = 1) } \arguments{ \item{model}{ an object of class \code{gnm}, \code{glm} or \code{lm} } \item{fac1}{ a factor } \item{fac2}{ a factor } \item{d}{ integer, the number of multiplicative terms to use in the approximation } } \details{ This function operates on the matrix of mean residuals, with rows indexed by \code{fac1} and columns indexed by \code{fac2}. For \code{glm} and \code{glm} models, the matrix entries are weighted working residuals. The primary use of \code{residSVD} is to generate good starting values for the parameters in \code{\link{Mult}} terms in models to be fitted using \code{\link{gnm}}. } \value{ If \code{d = 1}, a numeric vector; otherwise a numeric matrix with \code{d} columns. } \author{ David Firth and Heather Turner } \seealso{ \code{\link{gnm}}, \code{\link{Mult}}} \examples{ set.seed(1) ## Goodman RC1 association model fits well (deviance 3.57, df 8) mentalHealth$MHS <- C(mentalHealth$MHS, treatment) mentalHealth$SES <- C(mentalHealth$SES, treatment) ## independence model indep <- gnm(count ~ SES + MHS, family = poisson, data = mentalHealth) mult1 <- residSVD(indep, SES, MHS) ## Now use mult1 as starting values for the RC1 association parameters RC1model <- update(indep, . ~ . + Mult(SES, MHS), start = c(coef(indep), mult1), trace = TRUE) ## Similarly for the RC2 model: mult2 <- residSVD(indep, SES, MHS, d = 2) RC2model <- update(indep, . ~ . + instances(Mult(SES, MHS), 2), start = c(coef(indep), mult2), trace = TRUE) ## ## See also example(House2001), where good starting values matter much more! ## } \keyword{ models } \keyword{ regression } \keyword{ nonlinear } gnm/man/friend.Rd0000644000176000001440000000211112515147523013374 0ustar ripleyusers\name{friend} \alias{friend} \docType{data} \title{ Occupation of Respondents and Their Closest Friend } \description{ Cross-classification of the occupation of respondent and that of their closest friend. Data taken from wave 10 (year 2000) of the British Household Panel Survey. } \usage{friend} \format{ A table of counts, with classifying factors \code{r} (respondent's occupational category; levels \code{1:31}) and \code{c} (friend's occupational category; levels \code{1:31}). } \source{ Chan, T.W. and Goldthorpe, J.H. (2004) Is there a status order in contemporary British society: Evidence from the occupational structure of friendship, \emph{European Sociological Review}, \bold{20}, 383--401. } \examples{ set.seed(1) ### Fit an association model with homogeneous row-column effects rc1 <- gnm(Freq ~ r + c + Diag(r,c) + MultHomog(r, c), family = poisson, data = friend) rc1 \dontrun{ ### Extend to two-component interaction rc2 <- update(rc1, . ~ . + MultHomog(r, c, inst = 2), etastart = rc1$predictors) rc2 } } \keyword{datasets} gnm/man/mentalHealth.Rd0000644000176000001440000000416212515147523014543 0ustar ripleyusers\name{mentalHealth} \alias{mentalHealth} \docType{data} \title{ Data on Mental Health and Socioeconomic Status} \description{ A 2-way contingency table from a sample of residents of Manhattan. Classifying variables are child's mental impairment (\code{MHS}) and parents' socioeconomic status (\code{SES}). } \usage{mentalHealth} \format{ A data frame with 24 observations on the following 3 variables. \describe{ \item{\code{count}}{a numeric vector} \item{\code{SES}}{an ordered factor with levels \code{A} < \code{B} < \code{C} < \code{D} < \code{E} < \code{F}} \item{\code{MHS}}{an ordered factor with levels \code{well} < \code{mild} < \code{moderate} < \code{impaired}} } } \source{ From Agresti (2002, p381); originally in Srole et al. (1978, p289). } \references{ Agresti, A. (2002). \emph{Categorical Data Analysis} (2nd edn). New York: Wiley. Srole, L, Langner, T. S., Michael, S. T., Opler, M. K. and Rennie, T. A. C. (1978), \emph{Mental Health in the Metropolis: The Midtown Manhattan Study}. New York: NYU Press. } \examples{ set.seed(1) ## Goodman Row-Column association model fits well (deviance 3.57, df 8) mentalHealth$MHS <- C(mentalHealth$MHS, treatment) mentalHealth$SES <- C(mentalHealth$SES, treatment) RC1model <- gnm(count ~ SES + MHS + Mult(SES, MHS), family = poisson, data = mentalHealth) ## Row scores and column scores are both unnormalized in this ## parameterization of the model ## The scores can be normalized as in Agresti's eqn (9.15): rowProbs <- with(mentalHealth, tapply(count, SES, sum) / sum(count)) colProbs <- with(mentalHealth, tapply(count, MHS, sum) / sum(count)) mu <- getContrasts(RC1model, pickCoef(RC1model, "[.]SES"), ref = rowProbs, scaleRef = rowProbs, scaleWeights = rowProbs) nu <- getContrasts(RC1model, pickCoef(RC1model, "[.]MHS"), ref = colProbs, scaleRef = colProbs, scaleWeights = colProbs) all.equal(sum(mu$qv[,1] * rowProbs), 0) all.equal(sum(nu$qv[,1] * colProbs), 0) all.equal(sum(mu$qv[,1]^2 * rowProbs), 1) all.equal(sum(nu$qv[,1]^2 * colProbs), 1) } \keyword{datasets} gnm/man/getContrasts.Rd0000644000176000001440000001367712515147523014630 0ustar ripleyusers\name{getContrasts} \alias{getContrasts} \title{ Estimated Contrasts and Standard Errors for Parameters in a gnm Model } \description{ Computes contrasts or scaled contrasts for a set of (non-eliminated) parameters from a \code{\link{gnm}} model, and computes standard errors for the estimated contrasts. Where possible, quasi standard errors are also computed. } \usage{ getContrasts(model, set = NULL, ref = "first", scaleRef = "mean", scaleWeights = NULL, dispersion = NULL, check = TRUE, ...) } \arguments{ \item{model}{ a model object of class \code{"gnm"}.} \item{set}{ a vector of indices (numeric) or coefficient names (character). If \code{NULL}, a dialog will open for parameter selection. } \item{ref}{either a single numeric index, or a vector of real numbers which sum to 1, or one of the character strings \code{"first"}, \code{"last"} or \code{"mean"}.} \item{scaleRef}{as for \code{ref}} \item{scaleWeights}{either \code{NULL}, a vector of real numbers, \code{"unit"} or \code{"setLength"}.} \item{dispersion}{either \code{NULL}, or a positive number by which the model's variance-covariance matrix should be scaled.} \item{check}{\code{TRUE} or \code{FALSE} or a numeric vector -- for which of the specified parameter combinations should estimability be checked? If \code{TRUE}, all are checked; if \code{FALSE}, none is checked.} \item{\dots}{ arguments to pass to other functions. } } \details{ The indices in \code{set} must all be in \code{1:length(coef(object))}. If \code{set = NULL}, a dialog is presented for the selection of indices (model coefficients). For the set of coefficients selected, contrasts and their standard errors are computed. A check is performed first on the estimability of all such contrasts (if \code{check = TRUE}) or on a specified subset (if \code{check} is a numeric index vector). The specific contrasts to be computed are controlled by the choice of \code{ref}: this may be \code{"first"} (the default), for contrasts with the first of the selected coefficients, or \code{"last"} for contrasts with the last, or \code{"mean"} for contrasts with the arithmetic mean of the coefficients in the selected set; or it may be an arbitrary vector of weights (summing to 1, not necessarily all non-negative) which specify a weighted mean against which contrasts are taken; or it may be a single index specifying one of the coefficients with which all contrasts should be taken. Thus, for example, \code{ref = 1} is equivalent to \code{ref = "first"}, and \code{ref = c(1/3, 1/3, 1/3)} is equivalent to \code{ref = "mean"} when there are three coefficients in the selected \code{set}. The contrasts may be scaled by \deqn{\frac{1}{\sqrt{\sum_r v_r * d_r^2}}}{1/sqrt(sum(v * d))} where \eqn{d_r} is a contrast of the r'th coefficient in \code{set} with the reference level specified by \code{scaleRef} and \eqn{v} is a vector of weights (of the same length as \code{set}) specified by \code{scaleWeights}. If \code{scaleWeights} is \code{NULL} (the default), \code{scaleRef} is ignored and no scaling is performed. Other options for \code{scaleWeights} are \code{"unit"} for weights equal to one and \code{"setLength"} for weights equal to the reciprocal of \code{length(set)}. If \code{scaleRef} is the same as \code{ref}, these options constrain the sum of squared contrasts to 1 and \code{length(set)} respectively. Quasi-variances (and corresponding quasi standard errors) are reported for \bold{unscaled} contrasts where possible. These statistics are invariant to the choice of \code{ref}, see Firth (2003) or Firth and Menezes (2004) for more details. } \value{ An object of class \code{qv} --- see \code{\link[qvcalc]{qvcalc}}. } \author{ David Firth and Heather Turner } \seealso{ \code{\link{gnm}}, \code{\link{se}}, \code{\link{checkEstimable}}, \code{\link[qvcalc]{qvcalc}}, \code{\link{ofInterest}}} \references{ Firth, D (2003). Overcoming the reference category problem in the presentation of statistical models. \emph{Sociological Methodology} \bold{33}, 1--18. Firth, D and Menezes, R X de (2004). Quasi-variances. \emph{Biometrika} \bold{91}, 65--80. } \examples{ ### Unscaled contrasts ### set.seed(1) ## Fit the "UNIDIFF" mobility model across education levels -- see ?yaish unidiff <- gnm(Freq ~ educ*orig + educ*dest + Mult(Exp(educ), orig:dest), ofInterest = "[.]educ", family = poisson, data = yaish, subset = (dest != 7)) ## Examine the education multipliers (differences on the log scale): unidiffContrasts <- getContrasts(unidiff, ofInterest(unidiff)) plot(unidiffContrasts, main = "Unidiff multipliers (log scale): intervals based on quasi standard errors", xlab = "Education level", levelNames = 1:5) ### Scaled contrasts (elliptical contrasts) ### set.seed(1) ## Goodman Row-Column association model fits well (deviance 3.57, df 8) mentalHealth$MHS <- C(mentalHealth$MHS, treatment) mentalHealth$SES <- C(mentalHealth$SES, treatment) RC1model <- gnm(count ~ SES + MHS + Mult(SES, MHS), family = poisson, data = mentalHealth) ## Row scores and column scores are both unnormalized in this ## parameterization of the model ## The scores can be normalized as in Agresti's eqn (9.15): rowProbs <- with(mentalHealth, tapply(count, SES, sum) / sum(count)) colProbs <- with(mentalHealth, tapply(count, MHS, sum) / sum(count)) mu <- getContrasts(RC1model, pickCoef(RC1model, "[.]SES"), ref = rowProbs, scaleRef = rowProbs, scaleWeights = rowProbs) nu <- getContrasts(RC1model, pickCoef(RC1model, "[.]MHS"), ref = colProbs, scaleRef = colProbs, scaleWeights = colProbs) all.equal(sum(mu$qv[,1] * rowProbs), 0) all.equal(sum(nu$qv[,1] * colProbs), 0) all.equal(sum(mu$qv[,1]^2 * rowProbs), 1) all.equal(sum(nu$qv[,1]^2 * colProbs), 1) } \keyword{ models } \keyword{ regression } \keyword{ nonlinear } gnm/man/vcov.gnm.Rd0000644000176000001440000000747112515147523013700 0ustar ripleyusers\name{vcov.gnm} \alias{vcov.gnm} \title{ Variance-covariance Matrix for Parameters in a Generalized Nonlinear Model } \description{ This method extracts or computes a variance-covariance matrix for use in approximate inference on estimable parameter combinations in a generalized nonlinear model. } \usage{ \method{vcov}{gnm}(object, dispersion = NULL, with.eliminate = FALSE, ...) } %- maybe also 'usage' for other objects documented here. \arguments{ \item{object}{ a model object of class \code{gnm}. } \item{dispersion}{the dispersion parameter for the fitting family. By default it is obtained from \code{object}. } \item{with.eliminate}{logical; should parts of the variance-covariance matrix corresponding to eliminated coefficients be computed?} \item{\dots}{ as for \code{\link{vcov}}. } } \details{ The resultant matrix does not itself necessarily contain variances and covariances, since \code{gnm} typically works with over-parameterized model representations in which parameters are not all identified. Rather, the resultant matrix is to be used as the kernel of quadratic forms which are the variances or covariances for estimable parameter combinations. The matrix values are scaled by \code{dispersion}. If the dispersion is not specified, it is taken as \code{1} for the \code{binomial} and \code{Poisson} families, and otherwise estimated by the residual Chi-squared statistic divided by the residual degrees of freedom. The dispersion used is returned as an attribute of the matrix. The dimensions of the matrix correspond to the non-eliminated coefficients of the \code{"gnm"} object. If \code{use.eliminate = TRUE} then setting can sometimes give appreciable speed gains; see \code{\link{gnm}} for details of the \code{eliminate} mechanism. The \code{use.eliminate} argument is currently ignored if the model has full rank. } \value{ A matrix with number of rows/columns equal to \code{length(coef(object))}. If there are eliminated coefficients and \code{use.eliminate = TRUE}, the matrix will have the following attributes: \item{covElim }{ a matrix of covariances between the eliminated and non-eliminated parameters. } \item{varElim }{ a vector of variances corresponding to the eliminated parameters.} } \references{ Turner, H and Firth, D (2005). Generalized nonlinear models in R: An overview of the gnm package. At \url{http://cran.r-project.org}} \author{ David Firth } \note{ The \code{gnm} class includes generalized linear models, and it should be noted that the behaviour of \code{vcov.gnm} differs from that of \code{\link{vcov.glm}} whenever \code{any(is.na(coef(object)))} is \code{TRUE}. Whereas \code{vcov.glm} drops all rows and columns which correspond to \code{NA} values in \code{coef(object)}, \code{vcov.gnm} keeps those columns (which are full of zeros, since the \code{NA} represents a parameter which is fixed either by use of the \code{constrain} argument to \code{gnm} or by a convention to handle linear aliasing). } \seealso{ \code{\link{getContrasts}}, \code{\link{se}} } \examples{ set.seed(1) ## Fit the "UNIDIFF" mobility model across education levels unidiff <- gnm(Freq ~ educ*orig + educ*dest + Mult(Exp(educ), orig:dest), family = poisson, data = yaish, subset = (dest != 7)) ## Examine the education multipliers (differences on the log scale): ind <- pickCoef(unidiff, "[.]educ") educMultipliers <- getContrasts(unidiff, rev(ind)) ## Now get the same standard errors using a suitable set of ## quadratic forms, by calling vcov() directly: cmat <- contr.sum(ind) sterrs <- sqrt(diag(t(cmat) \%*\% vcov(unidiff)[ind, ind] \%*\% cmat)) all(sterrs == (educMultipliers$SE)[-1]) ## TRUE } \keyword{ models } \keyword{ regression } \keyword{ nonlinear } gnm/man/exitInfo.Rd0000644000176000001440000000315612515147523013724 0ustar ripleyusers\name{exitInfo} \alias{exitInfo} \title{ Print Exit Information for gnm Fit } \description{ A utility function to print information on final iteration in \code{gnm} fit, intended for use when \code{gnm} has not converged. } \usage{ exitInfo(object) } \arguments{ \item{object}{ a \code{gnm} object. } } \details{ If \code{gnm} has not converged within the pre-specified maximum number of iterations, it may be because the algorithm has converged to a non-solution of the likelihood equations. In order to determine appropriate action, it is necessary to differentiate this case from one of near-convergence to the solution. \code{exitInfo} prints the absolute score and the corresponding convergence criterion for all parameters which failed to meet the convergence criterion at the last iteration. Clearly a small number of parameters with scores close to the criterion suggests near-convergence. } \references{ Vargas, M, Crossa, J, van Eeuwijk, F, Sayre, K D and Reynolds, M P (2001). Interpreting treatment by environment interaction in agronomy trials. \emph{Agronomy Journal} \bold{93}, 949--960. } \author{ Heather Turner } \seealso{ \code{\link{gnm}}} \examples{ ## Fit a "double UNIDIFF" model with low iterMax for illustration! set.seed(1) doubleUnidiff <- gnm(Freq ~ election*vote + election*class*religion + Mult(Exp(election), religion:vote) + Mult(Exp(election), class:vote), family = poisson, data = cautres, iterMax = 10) exitInfo(doubleUnidiff) } \keyword{ models } \keyword{ regression } \keyword{ nonlinear } gnm/man/Inv.Rd0000644000176000001440000000322212515147523012665 0ustar ripleyusers\name{Inv} \alias{Inv} \title{ Specify the Reciprocal of a Predictor in a gnm Model Formula} \description{ A function of class \code{"nonlin"} to specify the reciprocal of a predictor in the formula argument to \code{\link{gnm}}. } \usage{ Inv(expression, inst = NULL) } \arguments{ \item{expression}{ a symbolic expression representing the (possibly nonlinear) predictor. } \item{inst}{ (optional) an integer specifying the instance number of the term. } } \details{ The \code{expression} argument is interpreted as the right hand side of a formula in an object of class \code{"formula"}, except that an intercept term is not added by default. Any function of class \code{"nonlin"} may be used in addition to the usual operators and functions. } \value{ A list with the components required of a \code{"nonlin"} function: \item{ predictors }{the \code{expression} argument passed to \code{Inv}} \item{ term }{a function to create a deparsed mathematical expression of the term, given a label for the predictor.} \item{ call }{the call to use as a prefix for parameter labels. } } \author{ Heather Turner } \seealso{ \code{\link{gnm}}, \code{\link{formula}}, \code{\link{nonlin.function}}} \examples{ ## One way to fit the logistic function without conditional ## linearity as in ?nls library(gnm) set.seed(1) DNase1 <- subset(DNase, Run == 1) test <- gnm(density ~ -1 + Mult(1, Inv(Const(1) + Exp(Mult(1 + offset(-log(conc)), Inv(1))))), start = c(NA, 0, 1), data = DNase1, trace = TRUE) coef(test) } \keyword{ models } \keyword{ regression } \keyword{ nonlinear } gnm/man/yaish.Rd0000644000176000001440000000471312515147523013254 0ustar ripleyusers\name{yaish} \alias{yaish} \docType{data} \title{ Class Mobility by Level of Education in Israel} \description{ A 3-way contingency table of father/son pairs, classified by father's social class (\code{orig}), son's social class (\code{dest}) and son's education level (\code{educ}). } \usage{yaish} \format{ A table of counts, with classifying factors \code{educ} (levels \code{1:5}), \code{orig} (levels \code{1:7}) and \code{dest} (levels \code{1:7}). } \source{Originally in Yaish (1998), see also Yaish (2004, p316).} \references{ Yaish, M (1998). Opportunities, Little Change. Class Mobility in Israeli Society: 1974-1991. D.Phil. Thesis, Nuffield College, University of Oxford. Yaish, M (2004). \emph{Class Mobility Trends in Israeli Society, 1974-1991.} Lewiston: Edwin Mellen Press. } \examples{ set.seed(1) ## Fit the "UNIDIFF" mobility model across education levels, leaving out ## the uninformative subtable for dest == 7: ## unidiff <- gnm(Freq ~ educ*orig + educ*dest + Mult(Exp(educ), orig:dest), family = poisson, data = yaish, subset = (dest != 7)) ## Deviance should be 200.3, 116 d.f. ## ## Look at the multipliers of the orig:dest association: ofInterest(unidiff) <- pickCoef(unidiff, "[.]educ") coef(unidiff) ## ## Coefficients of interest: ## Mult(Exp(.), orig:dest).educ1 Mult(Exp(.), orig:dest).educ2 ## -0.5513258 -0.7766976 ## Mult(Exp(.), orig:dest).educ3 Mult(Exp(.), orig:dest).educ4 ## -1.2947494 -1.5902644 ## Mult(Exp(.), orig:dest).educ5 ## -2.8008285 ## ## Get standard errors for the contrasts with educ1: ## getContrasts(unidiff, ofInterest(unidiff)) ## estimate SE quasiSE ## Mult(Exp(.), orig:dest).educ1 0.0000000 0.0000000 0.09757438 ## Mult(Exp(.), orig:dest).educ2 -0.2253718 0.1611874 0.12885847 ## Mult(Exp(.), orig:dest).educ3 -0.7434236 0.2335083 0.21182123 ## Mult(Exp(.), orig:dest).educ4 -1.0389386 0.3434256 0.32609380 ## Mult(Exp(.), orig:dest).educ5 -2.2495026 0.9453764 0.93560643 ## quasiVar ## Mult(Exp(.), orig:dest).educ1 0.00952076 ## Mult(Exp(.), orig:dest).educ2 0.01660450 ## Mult(Exp(.), orig:dest).educ3 0.04486823 ## Mult(Exp(.), orig:dest).educ4 0.10633716 ## Mult(Exp(.), orig:dest).educ5 0.87535940 ## ## Table of model residuals: ## residuals(unidiff) } \author{David Firth} \keyword{ datasets } gnm/man/nonlin.function.Rd0000644000176000001440000001564712515147523015270 0ustar ripleyusers\name{nonlin.function} \alias{nonlin.function} \title{ Functions to Specify Nonlinear Terms in gnm Models } \description{ Nonlinear terms maybe be specified in the formula argument to gnm by a call to a function of class \code{"nonlin"}. A \code{"nonlin"} function takes a list of arguments and returns a list of arguments for the internal \code{nonlinTerms} function. } \arguments{ \item{...}{ arguments required to define the term, e.g. symbolic representations of predictors in the term. } \item{inst}{(optional) an integer specifying the instance number of the term - for compatibility with \code{\link{instances}}. } } \value{ The function should return a list with the following components: \item{predictors}{ a list of symbolic expressions or formulae with no left hand side which represent (possibly nonlinear) predictors that form part of the term. Intercepts will be added by default to predictors specified by formulae. If predictors are named, these names will be used as a prefix for parameter labels or the parameter label itself in the single parameter case (in either case, prefixed by the call if supplied.) Predictors that may include an intercept should always be named or matched to a call. } \item{variables}{ an optional list of expressions representing variables in the term. } \item{term}{ a function which takes the arguments \code{predLabels} and \code{varLabels}, which are vectors of labels defined by \code{gnm} that correspond to the specified predictors and variables, and returns a deparsed mathematical expression of the full term. Only functions recognised by \code{deriv} should be used in the expression, e.g. \code{+} rather than \code{sum}.} \item{common}{ an optional numeric index of \code{predictors} with duplicated indices identifying single factor predictors for which homologous effects are to be estimated. } \item{call}{ an optional call to be used as a prefix for parameter labels, specified as an R expression. } \item{match}{ (if \code{call} is non-\code{NULL}) a numeric index of \code{predictors} specifying which arguments of \code{call} the predictors match to - zero indicating no match. If \code{NULL}, predictors will not be matched. It is recommended that matches are specified wherever possible, to ensure parameter labels are well-defined. Parameters in matched predictors are labelled using "dot-style" labelling, see examples.} \item{start}{ an optional function which takes a named vector of parameters corresponding to the predictors and returns a vector of starting values for those parameters. This function is ignored if the term is nested within another nonlinear term.} } \author{ Heather Turner } \seealso{ \code{\link{Const}} to specify a constant, \code{\link{Dref}} to specify a diagonal reference term, \code{\link{Exp}} to specify the exponential of a predictor, \code{\link{Inv}} to specify the reciprocal of a predictor, % \code{\link{Log}} to specify the natural logarithm of a predictor, % \code{\link{Logit}} to specify the logit of a predictor, \code{\link{Mult}} to specify a multiplicative interaction, \code{\link{MultHomog}} to specify a homogeneous multiplicative interaction, % \code{\link{Raise}} to raise a predictor to a constant power. } \examples{ ### Equivalent of weighted.MM function in ?nls weighted.MM <- function(resp, conc){ list(predictors = list(Vm = substitute(conc), K = 1), variables = list(substitute(resp), substitute(conc)), term = function(predictors, variables) { pred <- paste("(", predictors[1], "/(", predictors[2], " + ", variables[2], "))", sep = "") pred <- paste("(", variables[1], " - ", pred, ")/sqrt(", pred, ")", sep = "") }) } class(weighted.MM) <- "nonlin" ## use to fitted weighted Michaelis-Menten model Treated <- Puromycin[Puromycin$state == "treated", ] Pur.wt.2 <- gnm( ~ -1 + weighted.MM(rate, conc), data = Treated, start = c(Vm = 200, K = 0.1), verbose = FALSE) Pur.wt.2 ## ## Call: ## gnm(formula = ~-1 + weighted.MM(rate, conc), data = Treated, ## start = c(Vm = 200, K = 0.1), verbose = FALSE) ## ## Coefficients: ## Vm K ## 206.83477 0.05461 ## ## Deviance: 14.59690 ## Pearson chi-squared: 14.59690 ## Residual df: 10 ### The definition of MultHomog MultHomog <- function(..., inst = NULL){ dots <- match.call(expand.dots = FALSE)[["..."]] list(predictors = dots, common = rep(1, length(dots)), term = function(predictors, ...) { paste("(", paste(predictors, collapse = ")*("), ")", sep = "") }, call = as.expression(match.call())) } class(MultHomog) <- "nonlin" ## use to fit homogeneous multiplicative interaction set.seed(1) RChomog <- gnm(Freq ~ origin + destination + Diag(origin, destination) + MultHomog(origin, destination), ofInterest = "MultHomog", family = poisson, data = occupationalStatus, verbose = FALSE) RChomog ## ## Call: ## ## gnm(formula = Freq ~ origin + destination + Diag(origin, destination) + ## MultHomog(origin, destination), ofInterest = "MultHomog", family = poisson, ## data = occupationalStatus, verbose = FALSE) ## ## Coefficients of interest: ## MultHomog(origin, destination)1 ## -1.50089 ## MultHomog(origin, destination)2 ## -1.28260 ## MultHomog(origin, destination)3 ## -0.68443 ## MultHomog(origin, destination)4 ## -0.10055 ## MultHomog(origin, destination)5 ## -0.08338 ## MultHomog(origin, destination)6 ## 0.42838 ## MultHomog(origin, destination)7 ## 0.84452 ## MultHomog(., .).`origin|destination`8 ## 1.08809 ## ## Deviance: 32.56098 ## Pearson chi-squared: 31.20716 ## Residual df: 34 ## ## the definition of Exp Exp <- function(expression, inst = NULL){ list(predictors = list(substitute(expression)), term = function(predictors, ...) { paste("exp(", predictors, ")", sep = "") }, call = as.expression(match.call()), match = 1) } class(Exp) <- "nonlin" ## use to fit exponentional model x <- 1:100 y <- exp(- x / 10) set.seed(4) exp1 <- gnm(y ~ Exp(1 + x), verbose = FALSE) exp1 ## ## Call: ## gnm(formula = y ~ Exp(1 + x), verbose = FALSE) ## ## Coefficients: ## (Intercept) Exp(. + x).(Intercept) ## 1.549e-11 -7.934e-11 ## Exp(1 + .).x ## -1.000e-01 ## ## Deviance: 9.342418e-20 ## Pearson chi-squared: 9.342418e-20 ## Residual df: 97 } \keyword{ models } \keyword{ regression } \keyword{ nonlinear } gnm/man/Diag.Rd0000644000176000001440000000174212515147523013002 0ustar ripleyusers\name{Diag} \alias{Diag} \title{Equality of Two or More Factors} \description{ Converts two or more factors into a new factor whose value is 0 where the original factors are not all equal, and nonzero otherwise. } \usage{ Diag(..., binary = FALSE) } \arguments{ \item{\dots}{ One or more factors} \item{binary}{ Logical } } \value{ Either a factor (if \code{binary = FALSE}) or a 0-1 numeric vector (if \code{binary = TRUE}). } \details{ Used mainly in regression models for data classified by two or more factors with the same levels. By default, operates on k-level factors to produce a new factor having k+1 levels; if \code{binary = TRUE} is specified, the result is a coarser binary variable equal to 1 where all of the input factors are equal and 0 otherwise. } \author{ David Firth and Heather Turner} \seealso{\code{\link{Symm}}} \examples{ row <- gl(4, 4, 16) col <- gl(4, 1, 16) diag4by4 <- Diag(row, col) matrix(Diag(row, col, binary = TRUE), 4, 4) } \keyword{ models } gnm/man/checkEstimable.Rd0000644000176000001440000000412212515147523015034 0ustar ripleyusers\name{checkEstimable} \alias{checkEstimable} \title{ Check Whether One or More Parameter Combinations in a gnm Model are Identified } \description{ For each of a specified set of linear combinations of parameters from a \code{\link{gnm}} model, checks numerically whether the combination's estimate is invariant to re-parameterization of the model. } \usage{ checkEstimable(model, combMatrix = diag(length(coef(model))), tolerance = NULL) } \arguments{ \item{model}{ a model object of class \code{"gnm"} } \item{combMatrix}{ numeric: either a vector of length the same as \code{length(coef(model))}, or a matrix with that number of rows. Coefficients of one or more linear combinations of the model's parameters.} \item{tolerance}{ numeric: a threshold value for detection of non-estimability. If \code{NULL}, the default value of the \code{tol} argument to \code{\link[Matrix]{rankMatrix}} is used. } } \value{A logical vector of length equal to the number of parameter combinations tested; \code{NA} where a parameter combination is identically zero.} \author{ David Firth and Heather Turner } \seealso{ \code{\link{gnm}}, \code{\link{se}}, \code{\link{getContrasts}} } \references{ Catchpole, E.A. and Morgan, B.J.T. (1997). Detecting parameter redundancy. \emph{Biometrika}, \bold{84}, 187--196. } \examples{ set.seed(1) ## Fit the "UNIDIFF" mobility model across education levels unidiff <- gnm(Freq ~ educ*orig + educ*dest + Mult(Exp(educ), orig:dest), family = poisson, data = yaish, subset = (dest != 7)) ## Check whether multiplier contrast educ4 - educ5 is estimable ofInterest(unidiff) <- pickCoef(unidiff, "[.]educ") mycontrast <- numeric(length(coef(unidiff))) mycontrast[ofInterest(unidiff)[4:5]] <- c(1, -1) checkEstimable(unidiff, mycontrast) ## should be TRUE ## Check whether multiplier educ4 itself is estimable mycontrast[ofInterest(unidiff)[5]] <- 0 checkEstimable(unidiff, mycontrast) ## should be FALSE -- only *differences* are identified here } \keyword{ models } \keyword{ regression } \keyword{nonlinear}