An overview of object classes ============================= In `adegraphics`, a user-level function produces a plot that is stored (and returned) as an object. The class architecture of the objects created by `adegraphics` functions is described in [Figure 1](#classes).
This class management highlights a hierarchy with two parent classes: - `ADEg` for simple graphs. It contains the display of a single data set using only one kind of representation (e.g., arrows, points, lines, etc.) - `ADEgS` for multiple graphs. It contains a collection of at least two simple graphs (`ADEg`, `trellis` or `ADEgS`) The `ADEg` class has five child classes which are also subdivided in several child classes. Each of these five child classes is dedicated for a particular graphical data representation: - `ADEg.S1`: unidimensional graph of a numeric score - `ADEg.S2`: bidimensional graph of xy coordinates (`matrix` or `data.frame` object) - `ADEg.C1`: bidimensional graph of a numeric score (bar chart or curve) - `ADEg.T`: heat map-like representation of a data table (`matrix`, `data.frame`, `dist` or `table` object) - `ADEg.Tr`: ternary plot of xyz coordinates (`matrix` or `data.frame` object) The `ADEg` class and its five child classes are virtual: it is not allowed to create object belonging to these classes. Users can only create objects belonging to child classes by calls to user functions (see the [User functions](#user-functions) section). Simple graph (`ADEg` object) ============================ In the `adegraphics` package, a graph is created by a call to a user function and stored as an R object. These functions allow to display the raw data but also the outputs of a multivariate analysis. The following sections describe the different graphical functions available in the package. User functions -------------- Several user functions are available to create a simple graph (stored as an `ADEg` object in R). Each function creates an object of a given class (see [Figure 1](#classes)). [Table 1](#functionsADEg) lists the different functions, their corresponding classes and a short description. The `ade4` users would not be lost: many functions have kept their names in `adegraphics`. The main changes are listed in [Table 2](#functionsADEgchanged).
Table 1: Graphical functions available in `adegraphics` Function Class of the returned object Description --------- ------------------------------ -------------- `s1d.barchart` `C1.barchart` 1-D plot of a numeric score by bars `s1d.curve` `C1.curve` 1-D plot of a numeric score linked by curves `s1d.curves` `C1.curves` 1-D plot of multiple scores linked by curves `s1d.density` `C1.density` 1-D plot of a numeric score by density curves `s1d.dotplot` `C1.dotplot` 1-D plot of a numeric score by dots `s1d.gauss` `C1.gauss` 1-D plot of a numeric score by Gaussian curves `s1d.hist` `C1.hist` 1-D plot of a numeric score by bars `s1d.interval` `C1.interval` 1-D plot of the interval between two numeric scores `s1d.boxplot` `S1.boxplot` 1-D box plot of a numeric score partitioned in classes `s1d.class` `S1.class` 1-D plot of a numeric score partitioned in classes `s1d.distri` `S1.distri` 1-D plot of a numeric score by means/tandard deviations computed using an external table of weights `s1d.label` `S1.label` 1-D plot of a numeric score with labels `s1d.match` `S1.match` 1-D plot of the matching between two numeric scores `s.arrow` `S2.arrow` 2-D scatter plot with arrows `s.class` `S2.class` 2-D scatter plot with a partition in classes `s.corcircle` `S2.corcircle` Correlation circle `s.density` `S2.density` 2-D scatter plot with kernel density estimation `s.distri` `S2.distri` 2-D scatter plot with means/standard deviations computed using an external table of weights `s.image` `S2.image` 2-D scatter plot with loess estimation of an additional numeric score `s.label` `S2.label` 2-D scatter plot with labels `s.logo` `S2.logo` 2-D scatter plot with logos (pixmap objects) `s.match` `S2.match` 2-D scatter plot of the matching between two sets of coordinates `s.Spatial` `S2.label` Mapping of a `Spatial*` object `s.traject` `S2.traject` 2-D scatter plot with trajectories `s.value` `S2.value` 2-D scatter plot with proportional symbols `table.image` `T.image` Heat map-like representation with colored cells `table.value` `T.value` or `T.cont` Heat map-like representation with proportional symbols `triangle.class` `Tr.class` Ternary plot with a partition in classes `triangle.label` `Tr.label` Ternary plot with labels `triangle.match` `Tr.match` Ternary plot of the matching between two sets of coordinates `triangle.traject` `Tr.match` Ternary plot with trajectories
Table 2: Changes in functions names between `ade4` and `adegraphics` Function in `ade4` Equivalence in `adegraphics` ------------------------------------------- ------------------------------ `table.cont`, `table.dist`, `table.value` `table.value` [^1] `table.paint` `table.image` `sco.boxplot` `s1d.boxplot` `sco.class` `s1d.class` `sco.distri` `s1d.distri` `sco.gauss` `s1d.gauss` `sco.label` `s1d.label` `sco.match` `s1d.match` `sco.quant` no equivalence `s.chull` `s.class`[^2] `s.kde2d` `s.density` `s.match.class` superposition of `s.match` and `s.class` `triangle.biplot` `triangle.match` `triangle.plot` `triangle.label` `s.multinom` `triangle.multinom` [^1]: The `table.value` function is now generic and can handle `dist` or `table` objects as arguments. [^2]: Convex hulls are now drawn by the `s.class` function (argument `chullSize`.) Arguments --------- The list of arguments of a function are given by the `args` function. ```{r label=chunk1} library(ade4) library(adegraphics) args(s.label) ``` Some arguments are very general and present in all user functions: - `plot`: a logical value indicating if the graph should be displayed - `storeData`: a logical value indicating if the data should be stored in the returned object. If `FALSE`, only the names of the data are stored. This allows to reduce the size of the returned object but it implies that the data should not be modified in the environment to plot again the graph. - `add`: a logical value indicating if the graph should be superposed on the graph already displayed in the current device (it replaces the argument `add.plot` in `ade4`). - `pos`: an integer indicating the position of the environment where the data are stored, relative to the environment where the function is called. Useful only if `storeData` is `FALSE`. - `…`: additional graphical parameters (see below) Some other arguments influence the graphical outputs and they are thus specific to the type of produced graph. [Figure 2](#gargsVSclass) summarizes some of these graphical parameters available for the different functions. We only reported the parameters stored in the `g.args` slot of the returned object (see the [Parameters in `g.args`](#gargs) section).
Figure 2: Specific arguments in each object class
```{r label=fig-gargsVSclass, include=TRUE, echo=FALSE, fig.width=7, fig.height=8}
source("gargsVSclass.R")
```
The `ade4` users would note that the names of some arguments have been modified in `adegraphics`. The [Appendix](#appendix) gives a full list of these modifications. Slots and Methods ----------------- A call to a graphical function (see the [User functions](#user-functions) section) returns an `ADEg` object. Each object is defined by a number of slots and several methods are associated to this class. Let us consider the `olympic` data set available in the `ade4` package. A principal component analysis (PCA) is applied on the `olympic$tab` table that contains the results for 33 participating athletes at the 1988 summer olympic games: ```{r label=chunk2} data(olympic) pca1 <- dudi.pca(olympic$tab, scannf = FALSE) ``` The barplot of eigenvalues is then drawn and stored in `g1`: ```{r label=plot1, fig.height=4, fig.width=4} g1 <- s1d.barchart(pca1$eig, p1d.horizontal = F, ppolygons.col = "white") ```
The class of the `g1` object is `C1.barchart` which extends the `ADEg` class: ```{r label=chunk3} class(g1) showClass("C1.barchart") ```
This object contains different slots: ```{r label=chunk4} slotNames(g1) ```
These slots are defined for each `ADEg` object and contain different types of information. The package `adegraphics` uses the capabilities of the `lattice` package to display a graph (by generating a `trellis` object). Hence, several slots contain information that will be passed in the call to the `lattice` functions: - `data`: a list containing information about the data. - `trellis.par`: a list of graphical parameters that are directly passed to the `lattice` functions using the `par.settings` argument (see the [Parameters in `trellis.par`](#trellispar) section). - `adeg.par`: a list of graphical parameters defined in `adegraphics`. The list of parameters can be obtained using the `adegpar` function (see the [Parameters in `adeg.par`](#adegpar) section). - `lattice.call`: a list of two elements containing the information required to create the `trellis` object: `graphictype` (the name of the `lattice` functions that should be used) and `arguments` (the list of parameter values required to obtain the `trellis` object). - `g.args`: a list containing at least the different values of the graphical arguments described in [Figure 2](#gargsVSclass) (see the [Parameters in `g.args`](#gargs) section). - `stats`: a list of internal preliminary computations performed to display the graph. - `s.misc`: a list of other internal parameters. - `Call`: an object of class `call` containing the matched call.
These different slots can be extracted using the `@` operator: ```{r label=chunk5} g1@data ```
All these slots are automatically filled during the object creation. The `trellis.par`, `adeg.par` and `g.args` can also be modified *a posteriori* using the `update` method (see the [Customizing a graph](#update) section). This allows to customize graphs after their creation. We consider the correlation circle that depicts the correlation between PCA axes and the results for each event: ```{r label=plot2, fig.width=4, fig.height=4} g2 <- s.corcircle(pca1$co) ``` ```{r label=chunk6} class(g2) g2@g.args ```
The argument `fullcircle` can be updated *a posteriori* so that the original object is modified: ```{r label=plot3, fig.width=4, fig.height=4} update(g2, fullcircle = FALSE) g2@g.args ``` Several other methods have been defined for the `ADEg` class allowing to extract information, modify or combine objects: - `getcall`, `getlatticecall` and `getstats`: these accessor methods return respectively the `Call`, the `lattice.call` and the `stats` slots. - `getparameters`: this method returns the `trellis.par` and/or the `adeg.par` slots. - `show`, `print` and `plot`: these methods display the `ADEg` object in the current device or in a new one. - `gettrellis`: this method returns the `ADEg` object as a `trellis` object. It can then be exploited using the `lattice` and `latticeExtra` packages. - `superpose`, `+` and `add.ADEg`: these methods superpose two `ADEg` graphs. It returns a multiple graph object of class `ADEgS` (see the [The basic methods for superposition, juxtaposition and insertion](#superpose) section). - `insert`: this method inserts an `ADEg` graph in an existing one or in the current device. It returns an `ADEgS` object (see the [The basic methods for superposition, juxtaposition and insertion](#superpose) section). - `cbindADEg`, `rbindADEg`: these methods combine several `ADEg` graphs. It returns an `ADEgS` object (see the [The basic methods for superposition, juxtaposition and insertion](#superpose) section). - `update`: this method modifies the graphical parameters after the `ADEg` creation. It updates the current display and returns the modified `ADEg` (see the [Customizing a graph](#update) section). For instance: ```{r label=chunk7} getcall(g1) ## equivalent to g1@Call ``` A biplot-like graph can be obtained using the `superpose` method. The result is a multiple graph: ```{r label=plot4, fig.width=4, fig.height=4} g3 <- s.label(pca1$li) g4 <- s.arrow(5 * pca1$c1, add = TRUE) class(g4) ``` In addition, some object classes have specific methods. For instance, a `zoom` method is available for `ADEg.S1` and `ADEg.S2` classes. For the `ADEg.S2` class, the method `addhist` (see the [The basic methods for superposition, juxtaposition and insertion](#superpose) section) decorates a 2-D graph by adding marginal distributions as histograms and density lines (this method replaces and extends the `s.hist` function of `ade4`). ```{r label=plot5, fig.width=4, fig.height=4} zoom(g3, zoom = 2, center = c(2, -2)) ``` Multiple graph (`ADEgS` object) =============================== The `adegraphics` package provides class `ADEgS` to manage easily the combination of several graphs. This class allows to deal with the superposition, insertion or juxtaposition of several graphs in a single object. An object of this class is a list containing several graphical objects and information about their positioning. Different ways to generate `ADEgS` objects are described below. Slots and Methods ----------------- The class `ADEgS` is used to store multiple graphs. Different slots are associated to this class (use the symbol `@` to extract information): - `ADEglist`: a list of graphs stored as `trellis`, `ADEg` and/or `ADEgS` objects. - `positions`: a matrix containing the positions of the graphs. It has four columns and as many rows as the number of graphical objects in the `ADEglist` slot. For each graph (i.e. row), it contains the coordinates of the bottom-left and top-right corners in `npc` units (i.e. normalized parent coordinates varying between 0 and 1). - `add`: a square binary matrix with as many rows and columns as the number of graphical objects in the `ADEglist` slot. It allows to manage the superposition of graphs: the value at the i-th row and j-th column is equal to 1 if the j-th graphical object is superposed on the i-th. Otherwise, this value is equal to 0. - `Call`: an object of class `call` containing the matched call. Several methods have been implemented to obtain information, edit or modify `ADEgS` objects. Several methods are inspired from the management of `list` in R: - `[`, `[[` and `$`: these methods extract one or more elements from the `ADEgS` object. - `getpositions`, `getgraphics` and `getcall`: these methods return the `positions`, the `ADEglist` and the `Call` slots, respectively. - `names` and `length`: these methods return the names and number of graphs contained in the object. - `[[<-` and `names<-`: these methods replace a graph or its name in an `ADEgS` object (acts on the `ADEglist` slot). - `show`, `plot` and `print`: these methods display the `ADEgS` object in the current device or in a new one. - `superpose` and `+`: these methods superpose two graphs. It returns a multiple graph object of class `ADEgS` (see the [The basic methods for superposition, juxtaposition and insertion](#superpose) section). - `insert`: this method inserts a graph in an existing one or in the current device. It returns a multiple graph object of class `ADEgS` (see the [The basic methods for superposition, juxtaposition and insertion](#superpose) section). - `cbindADEg`, `rbindADEg`: these methods combine several graphs. It returns an `ADEgS` object (see the [The basic methods for superposition, juxtaposition and insertion](#superpose) section). - `update`: this method modifies the names and/or the `positions` of the graphs contained in an `ADEgS` object. It updates the current display and returns the modified `ADEgS`. We will show in the next sections how these methods can be used to deal with `ADEgS` objects. Creating an `ADEgS` object by hand ---------------------------------- The `ADEgS` objects can be created by easy manipulation of several simple graphs. Some methods (e.g., `insert`, `superpose`) can be used to create a compilation of graphs by hand. ### The basic methods for superposition, juxtaposition and insertion The functions `superpose`, `+` and `add.ADEg` allow the superposition of an `ADEg`/`ADEgS` object on an `ADEg`/`ADEgS` object. The vector `olympic$score` contains the total number of points computed for each participant. This vector is used to generate a `factor` partitioning the participants in two groups according to their final result (more or less than 8000 points): ```{r label=chunk8} fac.score <- factor(olympic$score < 8000, labels = c("MT8000", "LT8000")) ``` These two groups can be represented on the PCA factorial map using the `s.class` function: ```{r label=plot6, fig.width=4, fig.height=4} g5 <- s.class(pca1$li, fac.score, col = c("red", "blue"), chullSize = 1, ellipseSize = 0, plabels.cex = 2, pbackground.col = "grey85", paxes.draw = TRUE) ``` The graph with the labels (object `g3`) can then be superposed on this one: ```{r label=plot7, fig.width=4, fig.height=4} g6 <- superpose(g5, g3, plot = TRUE) ## equivalent to g5 + g3 class(g6) ``` In the case of a superposition, the graphical parameters (e.g., background and limits) of the first graph (the one below) are used as a reference and applied to the second one (the one above). Note that it is also possible to use the `add = TRUE` argument in the call of a simple user function (functions described in [Table 1](#functionsADEg)) to perform a superposition. The graph `g6` can also be obtained by: ```{r label=chunk9, eval=FALSE} g5 s.label(pca1$li, add = TRUE) ``` The functions `cbindADEg` and `rbindADEg` allows to combine several graphical objects (`ADEg`, `ADEgS` or `trellis`) by rows or by columns. The new created `ADEgS` contains the list of the reduced graphs: ```{r label=plot8, fig.width=6, fig.height=6} rbindADEg(cbindADEg(g2, g3), cbindADEg(g5, g6), plot = TRUE) ``` The function `insert` allows the insertion of a graphical object on another one (`ADEg` or `ADEgS`). It takes the position of the inserted graph as an argument: ```{r label=plot9, fig.width=4, fig.height=4} g7 <- insert(g2, g6, posi = c(0.65, 0.65, 0.95, 0.95)) class(g7) ``` The different methods associated to the `ADEgS` class allow to obtain information and to modify the multiple graph: ```{r label=chunk10} length(g7) names(g7) names(g7) <- c("chulls", "labels", "cor") class(g7[1]) class(g7[[1]]) class(g7$chulls) ``` The multiple graph contains three simple graphs. It can be easily updated. For instance, the size of the inserted graph can be modified: ```{r label=plot10, fig.width=4, fig.height=4} pos.mat <- getpositions(g7) pos.mat pos.mat[3,] <- c(0.1, 0.7, 0.3, 0.9) update(g7, positions = pos.mat) ``` The graphs themselves can be modified, without affecting the global structure of the `ADEgS` object. Here, we replace the correlation circle by the barplot of eigenvalues: ```{r label=plot11, fig.width=4, fig.height=4} g7[[3]] <- g1 g7 ``` The `addhist` method adds univariate marginal distributions around an `ADEg.S2` and returns an `ADEgS` object: ```{r label=plot12, fig.width=4, fig.height=4} addhist(g3) ``` More examples are available in the help page by typing `example(superpose)`, `example(insert)`, `example(add.ADEg)` and `example(addhist)` in the R session. ### The `ADEgS` function The `ADEgS` function provides the most elementary and flexible way to create an `ADEgS` object. The different arguments of the function are: - `adeglist`: a list of several `trellis`, `ADEg` and/or `ADEgS` objects. - `positions`: a matrix with four columns and as many rows as the number of graphical objects in the `ADEglist` slot. For each subgraph, i.e. in each row, the coordinates of the top-right and the bottom-left hand corners are given in `npc` units (i.e., normalized parent coordinates varying from 0 to 1). - `layout`: an alternative way to specify the positions of graphs. It could be a vector of length 2 indicating the number of rows and columns used to split the device (similar to `mfrow` parameter in basic graphs). It could also be a matrix specifying the location of the graphs: each value in this matrix should be 0 or a positive integer (similar to `layout` function for basic graphs). - `add`: a square matrix with as many rows and columns as the number of graphical objects in the `ADEglist` slot. The value at the i-th row and j-th column is equal to 1 if the j-th graphical object is superposed to i-th one. Otherwise, this value is equal to 0. - `plot`: a logical value indicating if the graphs should be displayed. When users fill only one argument among `positions`, `layout` and `add`, the other values are automatically computed to define the `ADEgS` object. We illustrate the different possibilities to create objects with the `ADEgS` function. Simple juxtaposition using a vector as layout: ```{r label=plot13, fig.width=6, fig.height=3} ADEgS(adeglist = list(g2, g3), layout = c(1, 2)) ``` Layout specified as a matrix: ```{r label=plot14, fig.width=5, fig.height=5} mlay <- matrix(c(1, 1, 0, 1, 1, 0, 0, 0, 2), byrow = T, nrow = 3) mlay ADEgS(adeglist = list(g6, g2), layout = mlay) ``` Using the matrix of positions offers a very flexible way to arrange the different graphs: ```{r label=plot15, fig.width=5, fig.height=5} mpos <- rbind(c(0, 0.3, 0.7, 1), c(0.5, 0, 1, 0.5)) ADEgS(adeglist = list(g3, g5), positions = mpos) ``` Lastly, superposition can also be specified using the `add` argument: ```{r label=plot16, fig.width=4, fig.height=4} ADEgS(list(g5, g3), add = matrix(c(0, 1, 0, 0), byrow = TRUE, ncol = 2)) ``` More examples are available in the help page by typing `example(ADEgS)` in the R session. Automatic collections --------------------- The package `adegraphics` contains functionalities to create collections of graphs. These collections are based on a simple graph repeated for different groups of individuals, variables or axes. The building process of these collections is quite simple (definition of arguments in the call of a user function) and leads to the creation of an `ADEgS` object. ### Partitioning the data (`facets`) The `adegraphics` package allows to split up the data by one variable (`factor`) and to plot the subsets of data together. This possibility of conditional plot is available for all user functions (except the `table.*` functions) by setting the `facets` argument. This is directly inspired by the functionalities offered in the `lattice` and `ggplot2` packages. Let us consider the `jv73` data set. The table `jv73$morpho` contains the measures of 6 variables describing the geomorphology of 92 sites. A PCA can be performed on this data set: ```{r label=plot17, fig.width=4, fig.height=4} data(jv73) pca2 <- dudi.pca(jv73$morpho, scannf = FALSE) s.label(pca2$li) ``` The sites are located on 12 rivers (`jv73$fac.riv`) and it is possible to represent the PCA scores for each river using the `facets` argument: ```{r label=plot18, fig.width=7, fig.height=5.2} g8 <- s.label(pca2$li, facets = jv73$fac.riv) length(g8) names(g8) ``` The `ADEgS` returned object contains the 12 plots. It is then possible to focus on a given river (e.g., the Doubs river) by considering only a subplot (e.g., type `g8$Doubs`). The `facets` functionality is very general and available for the majority of `adegraphics` functions. For instance, with the `s.class` function: ```{r label=plot19, fig.width=7, fig.height=5.2} s.class(pca2$li, fac = jv73$fac.riv, col = rainbow(12), facets = jv73$fac.riv) ``` ### Multiple axes All 2-D functions (i.e. `s.*`) returning an object inheriting from the `ADEg.S2` class have the `xax` et `yax` arguments. These arguments allow to choose which column of the main argument (i.e. `df`) should be plotted as x and y axes. As in `ade4`, these two arguments can be simple integers. In `adegraphics`, the user can also specify vectors as `xax` and/or `yax` arguments to obtain multiple graphs. Here, we represent the different correlation circles associated to the first four PCA axes of the olympic data set: ```{r label=plot20, fig.width=6, fig.height=6} pca1 <- dudi.pca(olympic$tab, scannf = FALSE, nf = 4) g9 <- s.corcircle(pca1$co, xax = 1:2, yax = 3:4) length(g9) names(g9) g9@positions ``` ### Multiple score All 1-D functions (i.e. `s1d.*`) returning an object inheriting from the `ADEg.C1` or `ADEg.S1` classes have the `score` argument. Usually, this argument should be a numeric vector but it is also possible to consider an object with several columns (`data.frame` or `matrix`). In this case, an `ADEgS` object is returned in which one graph by column is created. For instance for the `olympic` data set, we can represent the link between the global performance (`fac.score`) and the PCA scores on the first four axes (`pca1$li`): ```{r label=plot21, fig.width=6, fig.height=6} dim(pca1$li) s1d.boxplot(pca1$li, fac.score, col = c("red", "blue"), psub = list(position = "topleft", cex = 2)) ``` ### Multiple variable Some user functions (`s1d.density`, `s1d.gauss`, `s1d.boxplot`, `s1d.class`, `s.class`, `s.image`, `s.traject`, `s.value`, `triangle.class`) have an argument named `fac` or `z`. This argument can have several columns (`data.frame` or `matrix`) so that each column is used to create a separate graph. For instance, we can represent the distribution of the 6 environmental variables on the PCA factorial map of the `jv73$tab` data set: ```{r label=plot22, fig.width=6, fig.height=4} s.value(pca2$li, pca2$tab, symbol = "circle") ``` ### Outputs of the `ade4` package Lastly, we reimplemented all the graphical functions of the `ade4` package designed to represent the outputs of a multivariate analysis. The functions `ade4::plot.*`, `ade4::kplot.*`, `ade4::scatter.*` and `ade4::score.*` return `ADEgS` objects. It is now very easy to represent or modify these graphical outputs: ```{r label=plot23, fig.width=6, fig.height=6} data(meaudret) pca3 <- dudi.pca(meaudret$env, scannf = FALSE) pca4 <- dudi.pca(meaudret$spe, scale = FALSE, scannf = FALSE) coi1 <- coinertia(pca3, pca4, scannf = FALSE, nf = 3) g10 <- plot(coi1) class(g10) names(g10) g10@Call ``` Customizing a graph =================== Compared to the `ade4` package, the main advantage of `adegraphics` concerns the numerous possibilities to customize a graph using several graphical parameters. These parameters are stored in slots `trellis.par`, `adeg.par` and `g.args` (see the [Slots and Methods](#slots) section) of an `ADEg` object. These parameters can be defined during the creation of a graph or updated *a posteriori* (using the `update` method). Parameters in `trellis.par` --------------------------- The `trellis.par` slot is a list of parameters that are directly included in the call of functions of the `lattice` package. The name of parameters and their default value are given by the `trellis.par.get` function of `lattice`. ```{r label=chunk11} library(lattice) sort(names(trellis.par.get())) ``` Hence, modifications of some of these parameters will modify the graphical display of an `ADEg` object. For instance, margins are defined using `layout.widths` and `layout.heights` parameters, `clip` parameter allows to overpass panel boundaries and `axis.line` and `axis.text` allow to customize lines and text of axes. ```{r label=plot24, fig.width=7, fig.height=3.5} d <- scale(olympic$tab) g11 <- table.image(d, plot = FALSE) g12 <- table.image(d, axis.line = list(col = "blue"), axis.text = list(col = "red"), plot = FALSE) ADEgS(c(g11, g12), layout = c(1, 2)) ``` Parameters in `adeg.par` ------------------------ The `adeg.par` slot is a list of graphical parameters specific to the `adegraphics` package. The name of parameters and their default value are available using the `adegpar` function which is inspired by the `par` function of the `graphics` package. ```{r label=chunk12} names(adegpar()) ``` A description of these parameters is available in the help page of the function (`?adegpar`). Note that each `adeg.par` parameter starts by the letter ’p’ and its name relates to the type of graphical element considered (`ptable` is for tables display, `ppoints` for points, `parrows` for arrows, etc). Each element of this list can contain one or more sublists. Details on a sublist are obtained using its name either as a parameter of the `adegpar` function or after the `$` symbol. For example, if we want to know the different parameters to manage the display of points: ```{r label=chunk13} adegpar("ppoints") adegpar()$ppoints ``` The full list of available parameters is summarized in [Figure 3](#paramVSparam).
Figure 3: Parameters that can be set with the `adegpar` function.
```{r label=fig-paramVSparam, echo=FALSE, fig.width=7, fig.height=7}
source("paramVSparam.R")
```
The ordinate represents the different sublists and the abscissa gives the name of the parameters available in each sublist. Note that some row names have two keys separated by a dot: the first key indicates the first level of the sublist, etc. For example `plabels.boxes` is the sublist `boxes` of the sublist `plabels`. The parameters `border`,`col`, `alpha`, `lwd`, `lty` and `draw` in `plabels.boxes` allow to control the aspect of the boxes around labels. According to the function called, only some of the full list of `adeg.par` parameters are useful to modify the graphical display. [Figure 4](#paramVSfunction) indicates which parameters can affect the display of an object created by a given user function. For example, the background (`pbackground` parameter) can be modified for all functions whereas the display of ellipses (`pellipses` parameter) affects only three functions.
Figure 4: Effect of `adeg.par` parameters in `adegraphics` functions.
```{r label=fig-paramVSfunction, echo=FALSE, fig.width=7, fig.height=10}
source("paramVSfunction.R")
```
### Global assignment The `adegpar` function allows to modify globally the values of graphical parameters so that changes will affect all subsequent displays. For example, we update the size/color of labels and add axes to a plot: ```{r label=plot25, fig.width=6, fig.height=3} oldadegpar <- adegpar() adegpar("plabels") g13 <- s.label(dfxy = pca1$li, plot = FALSE) adegpar(plabels = list(col = "blue", cex = 1.5), paxes.draw = TRUE) adegpar("plabels") g14 <- s.label(dfxy = pca1$li, plot = FALSE) ADEgS(c(g13, g14), layout = c(1, 2)) ``` As the `adegpar` function can accept numerous graphical parameters, it can be used to define some graphical themes. The next releases of `adegraphics` will offer functionalities to easily create, edit and store graphical themes. Here, we reassign the original default parameters: ```{r label=chunk14} adegpar(oldadegpar) ``` ### Local assignment A second option is to update the graphical parameters locally so that the changes will only modify the object created. This is done using the dots (`...`) argument in the call to a user function. In this case, the default values of parameters in the global environment are not modified: ```{r label=plot26, fig.width=4, fig.height=4} adegpar("ppoints") s.label(dfxy = pca1$li, plabels.cex = 0, ppoints = list(col = c(2, 4, 5), cex = 1.5, pch = 15)) adegpar("ppoints") ``` In the previous example, we can see that parameters can be either specified using a ’`.`’ separator or a list. For instance, using `plabels.cex = 0` or `plabels = list(cex = 0)` is strictly equivalent. Moreover, partial names can be used if there is no ambiguity (such as `plab.ce = 0` in our example). Parameters in `g.args` ---------------------- The `g.args` slot is a list of parameters specific to the function used (and thus to the class of the returned object). Several parameters are very general and used in all `adegraphics` functions: - `xlim`, `ylim`: limits of the graph on the x and y axes - `main`, `sub`: main title and subtitle - `xlab`, `ylab`: labels of the x and y axes - `scales`: a list determining how the x and y axes (tick marks dans labels) are drawn; this is the `scales` parameter of the `xyplot` function of `lattice` The `ADEg.S2` objects can also contain spatial information (map stored as a `Spatial` object or neighborhood stored as a `nb` object): - `Sp`, `sp.layout`: objects from the `sp` package to display spatial objects, `Sp` for maps and `sp.layout` for spatial widgets as a North arrow, scale, etc. - `nbobject`: object of class `nb` or `listw` to display neighbor graphs. When the `facets` (see the [Partitioning the data (`facets`)](#facets) section) argument is used, users can modify the parameter `samelimits`: if it is `TRUE`, all graphs have the same limits whereas limits are computed for each subgraph independently when it is `FALSE`. For example, considering the `jv73` data set, each subgraph is computed with its own limits and labels are then more scattered: ```{r label=plot27, fig.width=7, fig.height=5.2} s.label(pca2$li, facets = jv73$fac.riv, samelimits = FALSE) ``` Several other `g.args` parameters can be updated according to the class of the created object (see [Figure 2](#gargsVSclass)). Parameters applied on a `ADEgS` ------------------------------- Users can either apply the changes to all graphs or to update only one graph. Of an `ADEgS`, to apply changes on all the graphs contained in an `ADEgS`, the syntax is similar to the one described for an `ADEg` object. For example, background color can be changed for all graphs in `g10` using the `pbackground.col` parameter. ```{r label=plot28, fig.width=6, fig.height=6} g15 <- plot(coi1, pbackground.col = "steelblue") ``` To change the parameters of a given graph, the name of the parameter must be preceded by the name of the subgraph. This supposes that the names of subgraphs are known. For example, to modify only two graphs: ```{r label=plot29, fig.width=6, fig.height=6} names(g15) plot(coi1, XYmatch.pbackground.col = "steelblue", XYmatch.pgrid.col = "red", eig.ppolygons.col = "orange") ``` Using `adegraphics` functions in your package ============================================= In this section, we illustrate how `adegraphics` functionalities can be used to implement graphical functions in your own package. We created an objet of class `track` that contains a vector of distance and time. ```{r label=chunk15} tra1 <- list() tra1$time <- runif(300) tra1$distance <- tra1$time * 5 + rnorm(300) class(tra1) <- "track" ``` For an object of the class `track`, we wish to represent different components of the data: - an histogram of distances - an histogram of speeds (i.e., distance / time) - a 2D plot representing the distance, the time and the line corresponding to the linear model that predict distance by time The corresponding multiple plot can be done using `adegraphics` functions: ```{r label=plot30, fig.width=7, fig.height=2.3} g1 <- s1d.hist(tra1$distance, psub.text = "distance", ppolygons.col = "blue", pgrid.draw = FALSE, plot = FALSE) g2 <- s1d.hist(tra1$distance / tra1$time, psub.text = "speed", ppolygons.col = "red", plot = FALSE) g31 <- s.label(cbind(tra1$time, tra1$distance), paxes = list(aspectratio = "fill", draw = TRUE), plot = FALSE) g32 <- xyplot(tra1$distance ~ tra1$time, aspect = g31@adeg.par$paxes$aspectratio, panel = function(x, y) {panel.lmline(x, y)}) g3 <- superpose(g31, g32) G <- ADEgS(list(g1, g2, g3)) ``` To facilitate the graphical representation of an object of class `track`, the simplest solution is to design a function `plot` for this class. We illustrate how to define such function with a particular emphasis on the management of graphical parameters. The function is provided below and we detail the different steps. ```{r label=chunk16} plot.track <- function(x, pos = -1, storeData = TRUE, plot = TRUE, ...) { ## step 1 : sort parameters for each graph graphsnames <- c("histDist", "histSpeed", "regression") sortparameters <- sortparamADEgS(..., graphsnames = graphsnames, nbsubgraphs = c(1, 1, 2)) ## step 2 : define default values for graphical parameters params <- list() params[[1]] <- list(psub = list(text = "distance"), ppolygons = list(col = "blue"), pgrid = list(draw = FALSE)) params[[2]] <- list(psub = list(text = "speed"), ppolygons = list(col = "red"), pgrid = list(draw = FALSE)) params[[3]] <- list() params[[3]]$l1 <- list(paxes = list(aspectratio = "fill", draw = TRUE)) params[[3]]$l2 <- list() names(params) <- graphsnames sortparameters <- modifyList(params, sortparameters, keep.null = TRUE) ## step 3 : create each individual plot (ADEg) g1 <- do.call("s1d.hist", c(list(score = substitute(x$distance), plot = FALSE, storeData = storeData, pos = pos - 2), sortparameters[[1]])) g2 <- do.call("s1d.hist", c(list(score = substitute(x$distance / x$time), plot = FALSE, storeData = storeData, pos = pos - 2), sortparameters[[2]])) g31 <- do.call("s.label", c(list(dfxy = substitute(cbind(x$time, x$distance)), plot = FALSE, storeData = storeData, pos = pos - 2), sortparameters[[3]][[1]])) g32 <- xyplot(x$distance ~ x$time, aspect = g31@adeg.par$paxes$aspectratio, panel = function(x, y) {panel.lmline(x, y)}) g3 <- do.call("superpose", list(g31, g32)) g3@Call <- call("superpose", g31@Call, g32$call) ## step 4 : create the multiple plot (ADEgS) lay <- matrix(1:3, 1, 3) object <- new(Class = "ADEgS", ADEglist = list(g1, g2, g3), positions = layout2position(lay), add = matrix(0, ncol = 3, nrow = 3), Call = match.call()) names(object) <- graphsnames if(plot) print(object) invisible(object) } ``` In the first step, the arguments given by the user through the dots (…) argument are managed. A name is given to each subgraph and stored in the vector `graphnames`. Then, the function `sortparamADEgS` associates the graphical parameters of the dots (…) argument to each subgraph. If a prefix is specified and matches the name of a graph (e.g., `histDist.pbackground.col = grey`), the parameter is applied only to the graphic specified (e.g., called `histDist`). If no prefix is specified (e.g., `pbackground.col = grey`), the parameter is applied to all subgraphs. The function `sortparamADEgS` returns a list (length equal to the number of subgraph) of lists of graphical parameters.\ In the second step, default values for some graphical parameters are modified. The default parameters are stored in a list which has the same structure that the one produced by `sortparamADEgS` (i.e., names corresponding to those contained in `graphsnames`). Then, the `modifyList` function is applied to merge user and defaults values of paramaters (if a parameter is specified by the user and in the default, the value given by the user is used).\ In the third step, each subgraph is created. Here, we create two `C1.hist` objects and superpose a `S2.label` object and a `trellis` one. The functions `do.call` and `substitute` are used to provide a pretty call for each subgraph (stored in the `Call` slot).\ In a final step, the multiple graph is build through the creation of a new `ADEgS` object and possibly plotted.\ The `plot.track` function can then be used by: ```{r label=plot31, fig.width=7, fig.height=2.3} plot(tra1) ``` Graphical parameters can be modified by: ```{r label=plot32, fig.width=7, fig.height=2.3} plot(tra1, histDist.ppoly.col = "green", pbackground.col = "grey") ``` Examples ======== Labels customization -------------------- ```{r label=plot33, fig.width=6, fig.height=3} data(meaudret) g16 <- s.label(pca3$li, plot = FALSE) g17 <- s.label(pca3$li, ppoints.col= "red", plabels = list(box = list(draw = FALSE), optim = TRUE), plot = FALSE) ADEgS(c(g16, g17), layout = c(1, 2)) ``` Ellipses, stars and convex hulls -------------------------------- ```{r label=plot34, fig.width=6, fig.height=6} g18 <- s.class(pca3$li, fac = meaudret$design$season, plot = FALSE) g19 <- s.class(pca3$li, fac = meaudret$design$season, ellipseSize = 0, chullSize = 1, starSize = 0.5, col = TRUE, plot = FALSE) g20 <- s.class(pca3$li, fac = meaudret$design$season, pellipses.lwd = 2, pellipses.border = 2:5, pellipses.col = 2:5, plot = FALSE) g21 <- s.class(pca3$li, fac = meaudret$design$season, ellipseSize = 0, chullSize = 0, ppolygons.lwd = 2, plines.col = 2:5, starSize = 1.2, plot = FALSE) ADEgS(c(g18, g19, g20, g21), layout = c(2, 2)) ``` Values and legend ----------------- ```{r label=plot35, fig.width=6, fig.height=6} data(rpjdl) coa2 <- dudi.coa(rpjdl$fau, scannf = FALSE, nf = 3) g22 <- s.value(coa2$li, coa2$li[,3], plot = FALSE) g23 <- s.value(coa2$li, coa2$li[,3], method = "color", ppoints.cex = 0.8, plegend.size= 0.8, plot = FALSE) g24 <- s.value(coa2$li, coa2$li[,3], plegend.size = 0.8, ppoints.cex = 0.8, symbol = "square", method = "color", key = list(columns = 1), col = colorRampPalette(c("yellow", "blue"))(6), plot = FALSE) g25 <- s.value(coa2$li, coa2$li[, 3], center = 0, method = "size", ppoints.cex = 0.6, symbol = "circle", col = c("yellow", "red"), plot = FALSE) ADEgS(c(g22, g23, g24, g25), layout = c(2, 2)) ``` 1-D plot -------- ```{r label=plot36, fig.width=6, fig.height=6} score1 <- c(rnorm(1000, mean = -0.5, sd = 0.5), rnorm(1000, mean = 1)) fac1 <- rep(c("A", "B"), each = 1000) g26 <- s1d.density(score1, fac1, pback.col = "grey75", plot = FALSE) g27 <- s1d.density(score1, fac1, col = c(2, 4), plot = FALSE) g28 <- s1d.density(score1, fac1, col = c(2, 4), p1d.reverse = TRUE, p1d.horizontal = FALSE, p1d.rug.draw = FALSE, plot = FALSE) g29 <- s1d.density(score1, fac1, col = c(2, 4), ppolygons.alpha = 0.2, p1d = list(rug = list(tck = 1, line = FALSE)), plot = FALSE) ADEgS(c(g26, g27, g28, g29), layout = c(2, 2)) ``` Maps and neighbor graphs ------------------------ ```{r label=plot37, fig.width=6, fig.height=3} library(Guerry) library(sp) data(gfrance85) region.names <- data.frame(gfrance85)[, 5] col.region <- colors()[c(149, 254, 468, 552, 26)] g30 <- s.class(coordinates(gfrance85), region.names, porigin.include = FALSE, plot = FALSE) g31 <- s.class(coordinates(gfrance85), region.names, ellipseSize = 0, starSize = 0, Sp = gfrance85, pgrid.draw = FALSE, pSp.col = col.region[region.names], pSp.alpha = 0.4, plot = FALSE) ADEgS(c(g30, g31), layout = c(1, 2)) ``` ```{r label=plot38, fig.width=6, fig.height=4} s.Spatial(gfrance85[,7:12]) ``` ```{r label=plot39, fig.width=6, fig.height=3} data(mafragh, package = "ade4") g32 <- s.label(mafragh$xy, nb = mafragh$nb, plot = FALSE) g33 <- s.label(mafragh$xy, nb = mafragh$nb, pnb.ed.col = "red", plab.cex = 0, pnb.node = list(cex = 3, col = "blue"), ppoints.col = "green", plot = FALSE) ADEgS(c(g32, g33), layout = c(1, 2)) ``` Ternary plots ------------- ```{r label=plot40, fig.width=6, fig.height=3} data(euro123, package = "ade4") df <- rbind.data.frame(euro123$in78, euro123$in86, euro123$in97) row.names(df) <- paste(row.names(euro123$in78), rep(c(1, 2, 3), rep(12, 3)), sep = "") g34 <- triangle.label(df, label = row.names(df), showposition = TRUE, plot = FALSE) g35 <- triangle.label(euro123$in78, plabels.cex = 0, ppoints.cex = 2, addmean = TRUE, show = FALSE, plot = FALSE) ADEgS(c(g34, g35), layout = c(1, 2)) ``` Appendix ============================= This appendix summarizes the main changes between `ade4` and `adegraphics`. Each line corresponds to a graphical argument defined in `ade4` and its equivalent in `adegraphics` is given. | Arguments in `ade4` | Functions in `ade4` | `g.args` in `adegraphics` | `adeg.par` in `adegraphics` | | | ------------------- | ----------------------| ----------------------------| ------------------------------|---| | `abline.x` | `table.cont` | `ablineX` | | | | `abline.y` | `table.cont` | `ablineY` | | | | `abmean.x` | `table.cont` | `meanX` | | | | `abmean.y` | `table.cont` | `meanY` | | | | `addaxes` | `s.arrow`, `s.chull`, `s.class`, `s.distri`, `s.image`, `s.kde2d`, `s.label`, `s.logo`, `s.match`, `s.traject`, `s.value`, `triangle.class`, `triangle.plot` | |`paxes.draw` | | | `area` | `s.arrow`, `s.chull`, `s.class`, `s.distri`, `s.image`, `s.kde2d`, `s.label`, `s.logo`, `s.match`, `s.traject`, `s.value` | `Sp` | | a `Sp` object | | `axesell` | `s.class`, `s.distri`, `triangle.class` | | `pellipses.axes.draw` | | | `box` | `s.corcircle`, `triangle.plot` | | `pbackground.box` | | | `boxes` | `s.arrow`, `s.label`, `sco.class`, `sco.label`, `sco.match` | | `plabels.boxes.draw` | | | `cellipse` | `s.class`, `s.distri`, `triangle.class` | `ellipseSize` | | | | `cgrid` | `s.arrow`, `s.class`, `s.chull`, `s.corcircle`, `s.distri`, `s.image`, `s.kde2d`, `s.label`, `s.logo`, `s.match`, `s.traject`, `s.value`, `sco.boxplot`, `sco.class`, `sco.distri`, `sco.gauss`, `sco.label`, `sco.match` | | `pgrid.nint` | both play on the grid mesh, but they are not strictly equivalent | | `clabel` | `s.arrow`, `s.class`, `s.chull`, `s.corcircle`, `s.distri`, `s.kde2d`, `s.label`, `s.match`, `s.traject`, `sco.boxplot`, `sco.class`, `sco.distri`, `sco.gauss`, `sco.label`, `sco.match`, `triangle.plot` | | `plabels.cex` | | | `clabel` | `table.dist` | | | `axis.text = list()` `lattice` parameter | | `clabel.col` | `table.cont`, `table.paint`, `table.value` | | | `axis.text = list()` `lattice` parameter | | `clabel.row` | `table.cont`, `table.paint`, `table.value` | | | `axis.text = list()` `lattice` parameter | | `clegend` | `s.value`, `table.cont`, `table.value` | | `plegend.size` `ppoints.cex` | parameters setting the legend size | | `clegend` | `table.paint` | | `plegend.size` | | | `clogo` | `s.logo` | | `ppoints.cex` | | | `cneig` | `s.image`, `s.kde2d`, `s.label`, `s.logo`, `s.value` | | `pnb.edge.lwd` | | | `col.labels` | `table.cont`, `table.paint`, `table.value` | `labelsy` | | | | `contour` | `s.arrow`, `s.class`, `s.chull`, `s.distri`, `s.image`, `s.kde2d`, `s.label`, `s.logo`, `s.match`, `s.traject`, `s.value` | `Sp` | | a `Sp` object | | `contour.plot` | `s.image` | `region` | | | | `cpoints`, `cpoint` | `s.arrow`, `s.class`, `s.chull`, `s.distri`, `s.kde2d`, `s.label`, `s.match`, `s.traject`, `s.value`, `sco.class`, `sco.label`, `sco.match`, `triangle.class`, `triangle.plot` | | `ppoints.cex` | | | `csize` | `s.value`, `table.cont`, `table.dist`, `table.paint`, `table.value` | `ppoints.cex` | | | | `csize` | `sco.distri` | `sdSize` | | | | `cstar` | `s.class`, `s.distri`, `triangle.class` | `starSize` | | | | `csub` | `s.arrow`, `s.chull`, `s.class`, `s.corcircle`, `s.distri`, `s.image`, `s.kde2d`, `s.label`, `s.logo`, `s.match`, `s.traject`, `s.value`, `sco.boxplot`, `sco.class`, `sco.distri`, `sco.gauss`, `sco.label`, `sco.match`, `triangle.class`, `triangle.plot` | | `psub.cex` | | | `draw.line` | `triangle.biplot`, `triangle.class`, `triangle.plot` | | `pgrid.draw` | | | `edge` | `s.arrow`, `s.match`, `s.traject` | | `parrows.length` | setting the length of the arrows to 0 is equivalent to `edge = FALSE` | | `grid` | `s.arrow`, `s.chull`, `s.class`, `s.corcircle`, `s.distri`, `s.image`, `s.kde2d`, `s.label`, `s.logo`, `s.match`, `s.traject`, `s.value`, `sco.boxplot`, `sco.class`, `sco.distri`, `sco.gauss`, `sco.label`, `sco.match`, `table.cont`, `table.dist`, `table.value` | | `pgrid.draw` | | | `horizontal` | `sco.class`, `sco.gauss`, `sco.label`, `sco.match` | | `p1d.horizontal` | | | `image.plot` | `s.image` | `contour` | | | | `includeorigin`, `include.origin` | `s.arrow`, `s.chull`, `s.class`, `s.distri`, `s.image`, `s.kde2d`, `s.label`, `s.logo`, `s.match`, `s.traject`, `s.value`, `sco.boxplot`, `sco.class`, `sco.distri`, `sco.gauss`, `sco.label`, `sco.match` | | `porigin.include` | | | `kgrid` | `s.image` | `gridsize` | | | | `klogo` | `s.logo` | | | no correspondence | | `labeltriangle` | `triangle.class` , `triangle.plot` | | | no correspondence | | `legen` | `sco.gauss` | `labelplot` | | | | `neig` | `s.image`, `s.kde2d`, `s.label`, `s.logo`, `s.value` | `nbobject` | | a `nb` object | | `optchull` | `s.chull` | `chullSize` | | | | `origin` | `s.arrow`, `s.chull`, `s.class`, `s.corcircle`, `s.distri`, `s.image`, `s.kde2d`, `s.label`, `s.logo`, `s.match`, `s.traject`, `s.value`, `sco.boxplot`, `sco.class`, `sco.distri`, `sco.gauss`, `sco.label`, `sco.match` | | `porigin.origin` | | | `pch` | `s.arrow`, `s.chull`, `s.class`, `s.distri`, `s.kde2d`, `s.label`, `s.match`, `s.traject`, `s.value`, `sco.boxplot`, `sco.class`, `sco.label`, `sco.match`, `triangle.class`, `triangle.plot`, `table.cont` | | `ppoints.pch` | | | `pixmap` | `s.arrow`, `s.chull`, `s.class`, `s.distri`, `s.image`, `s.kde2d`, `s.label`, `s.logo`, `s.match`, `s.traject`, `s.value` | | | no correspondence | | `pos.lab` | `sco.class`, `sco.label`, `sco.match` | | `p1d.labpos` | | | `possub` | `s.arrow`, `s.chull`, `s.class`, `s.corcircle`, `s.distri`, `s.image`, `s.kde2d`, `s.label`, `s.logo`, `s.match`, `s.traject`, `s.value`, `sco.class`, `sco.gauss`, `sco.label`, `sco.match`, `triangle.class`, `triangle.plot` | | `psub.pos` | | | `rectlogo` | `s.logo` | `rect` | | | | `reverse` | `sco.class`, `sco.gauss`, `sco.label`, `sco.match` | | `p1d.reverse` | | | `row.labels` | `table.cont`, `table.paint`, `table.value` | `labelsx` | | | | `scale` | `triangle.class`, `triangle.plot` | `adjust` | | | | `show.position` | `triangle.class`, `triangle.plot` | `showposition` | | | | `sub` | `s.arrow`, `s.chull`, `s.class`, `s.corcircle`, `s.distri`, `s.image`, `s.kde2d`, `s.label`, `s.logo`, `s.match`, `s.traject`, `s.value`, `sco.boxplot`, `sco.class`, `sco.distri`, `sco.gauss`, `sco.label`, `sco.match`, `triangle.class`, `triangle.plot` | | `psub.text` | | | `y.rank` | `sco.distri` | `yrank` | | | | `zmax` | `s.value` | | | set to default max(abs(z)) | | | | | | | References ============================= adegraphics/inst/doc/adegraphics.html 0000644 0001751 0000144 00013047760 13024020511 017432 0 ustar hornik users
The adegraphics package
Alice Julien-Laferrière, Aurélie Siberchicot and Stéphane Dray
2016-12-13
The adegraphics package is a complete reimplementation of the graphical functionalities of the ade4 package (Dray and Dufour 2007). The package has been initially designed to improve the representation of the outputs of multivariate analyses performed with ade4 but as its graphical functionalities are very general, they can be used for other purposes.
The adegraphics package provides a flexible environment to produce, edit and manipulate graphs. We adopted an object oriented approach (a graph is an object) using S4 classes and methods and used the visualization system provided by the lattice (Sarkar 2008) and grid (Murrell 2005) packages. In adegraphics, graphs are R objects that can be edited, stored, combined, saved, removed, etc.
Note that we tried to facilitate the handling of adegraphics by ade4 users. Hence, the name of functions and parameters has been preserved in many cases. The main changes are listed in the appendix of this vignette so that it should be quite easy to use adegraphics. However, several new functionalities (graphical parameters, creation and manipulation of graphical objects, etc.) are now available and detailed in this vignette.
The adelist mailing list can be used to send questions and/or comments on adegraphics (see http://listes.univ-lyon1.fr/wws/info/adelist)
1 An overview of object classes
In adegraphics, a user-level function produces a plot that is stored (and returned) as an object. The class architecture of the objects created by adegraphics functions is described in Figure 1.
This class management highlights a hierarchy with two parent classes:
ADEgfor simple graphs. It contains the display of a single data set using only one kind of representation (e.g., arrows, points, lines, etc.)ADEgSfor multiple graphs. It contains a collection of at least two simple graphs (ADEg,trellisorADEgS)
The ADEg class has five child classes which are also subdivided in several child classes. Each of these five child classes is dedicated for a particular graphical data representation:
ADEg.S1: unidimensional graph of a numeric scoreADEg.S2: bidimensional graph of xy coordinates (matrixordata.frameobject)ADEg.C1: bidimensional graph of a numeric score (bar chart or curve)ADEg.T: heat map-like representation of a data table (matrix,data.frame,distortableobject)ADEg.Tr: ternary plot of xyz coordinates (matrixordata.frameobject)
The ADEg class and its five child classes are virtual: it is not allowed to create object belonging to these classes. Users can only create objects belonging to child classes by calls to user functions (see the User functions section).
2 Simple graph (ADEg object)
In the adegraphics package, a graph is created by a call to a user function and stored as an R object. These functions allow to display the raw data but also the outputs of a multivariate analysis. The following sections describe the different graphical functions available in the package.
2.1 User functions
Several user functions are available to create a simple graph (stored as an ADEg object in R). Each function creates an object of a given class (see Figure 1). Table 1 lists the different functions, their corresponding classes and a short description. The ade4 users would not be lost: many functions have kept their names in adegraphics. The main changes are listed in Table 2.
Table 1: Graphical functions available in adegraphics
| Function | Class of the returned object | Description |
|---|---|---|
s1d.barchart |
C1.barchart |
1-D plot of a numeric score by bars |
s1d.curve |
C1.curve |
1-D plot of a numeric score linked by curves |
s1d.curves |
C1.curves |
1-D plot of multiple scores linked by curves |
s1d.density |
C1.density |
1-D plot of a numeric score by density curves |
s1d.dotplot |
C1.dotplot |
1-D plot of a numeric score by dots |
s1d.gauss |
C1.gauss |
1-D plot of a numeric score by Gaussian curves |
s1d.hist |
C1.hist |
1-D plot of a numeric score by bars |
s1d.interval |
C1.interval |
1-D plot of the interval between two numeric scores |
s1d.boxplot |
S1.boxplot |
1-D box plot of a numeric score partitioned in classes |
s1d.class |
S1.class |
1-D plot of a numeric score partitioned in classes |
s1d.distri |
S1.distri |
1-D plot of a numeric score by means/tandard deviations computed using an external table of weights |
s1d.label |
S1.label |
1-D plot of a numeric score with labels |
s1d.match |
S1.match |
1-D plot of the matching between two numeric scores |
s.arrow |
S2.arrow |
2-D scatter plot with arrows |
s.class |
S2.class |
2-D scatter plot with a partition in classes |
s.corcircle |
S2.corcircle |
Correlation circle |
s.density |
S2.density |
2-D scatter plot with kernel density estimation |
s.distri |
S2.distri |
2-D scatter plot with means/standard deviations computed using an external table of weights |
s.image |
S2.image |
2-D scatter plot with loess estimation of an additional numeric score |
s.label |
S2.label |
2-D scatter plot with labels |
s.logo |
S2.logo |
2-D scatter plot with logos (pixmap objects) |
s.match |
S2.match |
2-D scatter plot of the matching between two sets of coordinates |
s.Spatial |
S2.label |
Mapping of a Spatial* object |
s.traject |
S2.traject |
2-D scatter plot with trajectories |
s.value |
S2.value |
2-D scatter plot with proportional symbols |
table.image |
T.image |
Heat map-like representation with colored cells |
table.value |
T.value or T.cont |
Heat map-like representation with proportional symbols |
triangle.class |
Tr.class |
Ternary plot with a partition in classes |
triangle.label |
Tr.label |
Ternary plot with labels |
triangle.match |
Tr.match |
Ternary plot of the matching between two sets of coordinates |
triangle.traject |
Tr.match |
Ternary plot with trajectories |
Table 2: Changes in functions names between ade4 and adegraphics
Function in ade4 |
Equivalence in adegraphics |
|---|---|
table.cont, table.dist, table.value |
table.value1 |
table.paint |
table.image |
sco.boxplot |
s1d.boxplot |
sco.class |
s1d.class |
sco.distri |
s1d.distri |
sco.gauss |
s1d.gauss |
sco.label |
s1d.label |
sco.match |
s1d.match |
sco.quant |
no equivalence |
s.chull |
s.class2 |
s.kde2d |
s.density |
s.match.class |
superposition of s.match and s.class |
triangle.biplot |
triangle.match |
triangle.plot |
triangle.label |
s.multinom |
triangle.multinom |
2.2 Arguments
The list of arguments of a function are given by the args function.
library(ade4)
library(adegraphics)##
## Attaching package: 'adegraphics'## The following objects are masked from 'package:ade4':
##
## kplotsepan.coa, s.arrow, s.class, s.corcircle, s.distri,
## s.image, s.label, s.logo, s.match, s.traject, s.value,
## table.value, triangle.classargs(s.label)## function (dfxy, labels = rownames(dfxy), xax = 1, yax = 2, facets = NULL,
## plot = TRUE, storeData = TRUE, add = FALSE, pos = -1, ...)
## NULLSome arguments are very general and present in all user functions:
plot: a logical value indicating if the graph should be displayedstoreData: a logical value indicating if the data should be stored in the returned object. IfFALSE, only the names of the data are stored. This allows to reduce the size of the returned object but it implies that the data should not be modified in the environment to plot again the graph.add: a logical value indicating if the graph should be superposed on the graph already displayed in the current device (it replaces the argumentadd.plotinade4).pos: an integer indicating the position of the environment where the data are stored, relative to the environment where the function is called. Useful only ifstoreDataisFALSE.…: additional graphical parameters (see below)
Some other arguments influence the graphical outputs and they are thus specific to the type of produced graph. Figure 2 summarizes some of these graphical parameters available for the different functions. We only reported the parameters stored in the g.args slot of the returned object (see the Parameters in g.args section).
The ade4 users would note that the names of some arguments have been modified in adegraphics. The Appendix gives a full list of these modifications.
2.3 Slots and Methods
A call to a graphical function (see the User functions section) returns an ADEg object. Each object is defined by a number of slots and several methods are associated to this class. Let us consider the olympic data set available in the ade4 package. A principal component analysis (PCA) is applied on the olympic$tab table that contains the results for 33 participating athletes at the 1988 summer olympic games:
data(olympic)
pca1 <- dudi.pca(olympic$tab, scannf = FALSE)The barplot of eigenvalues is then drawn and stored in g1:
g1 <- s1d.barchart(pca1$eig, p1d.horizontal = F, ppolygons.col = "white")
The class of the g1 object is C1.barchart which extends the ADEg class:
class(g1)## [1] "C1.barchart"
## attr(,"package")
## [1] "adegraphics"showClass("C1.barchart")## Class "C1.barchart" [package "adegraphics"]
##
## Slots:
##
## Name: data trellis.par adeg.par lattice.call g.args
## Class: list list list list list
##
## Name: stats s.misc Call
## Class: list list call
##
## Extends:
## Class "ADEg.C1", directly
## Class "ADEg", by class "ADEg.C1", distance 2
## Class "ADEgORtrellis", by class "ADEg.C1", distance 3
## Class "ADEgORADEgSORtrellis", by class "ADEg.C1", distance 3
This object contains different slots:
slotNames(g1)## [1] "data" "trellis.par" "adeg.par" "lattice.call"
## [5] "g.args" "stats" "s.misc" "Call"
These slots are defined for each ADEg object and contain different types of information. The package adegraphics uses the capabilities of the lattice package to display a graph (by generating a trellis object). Hence, several slots contain information that will be passed in the call to the lattice functions:
data: a list containing information about the data.trellis.par: a list of graphical parameters that are directly passed to thelatticefunctions using thepar.settingsargument (see the Parameters intrellis.parsection).adeg.par: a list of graphical parameters defined inadegraphics. The list of parameters can be obtained using theadegparfunction (see the Parameters inadeg.parsection).lattice.call: a list of two elements containing the information required to create thetrellisobject:graphictype(the name of thelatticefunctions that should be used) andarguments(the list of parameter values required to obtain thetrellisobject).g.args: a list containing at least the different values of the graphical arguments described in Figure 2 (see the Parameters ing.argssection).stats: a list of internal preliminary computations performed to display the graph.s.misc: a list of other internal parameters.Call: an object of classcallcontaining the matched call.
These different slots can be extracted using the @ operator:
g1@data## $score
## [1] 3.4182381 2.6063931 0.9432964 0.8780212 0.5566267 0.4912275 0.4305952
## [8] 0.3067981 0.2669494 0.1018542
##
## $frame
## [1] 24
##
## $storeData
## [1] TRUE
All these slots are automatically filled during the object creation. The trellis.par, adeg.par and g.args can also be modified a posteriori using the update method (see the Customizing a graph section). This allows to customize graphs after their creation.
We consider the correlation circle that depicts the correlation between PCA axes and the results for each event:
g2 <- s.corcircle(pca1$co)
class(g2)## [1] "S2.corcircle"
## attr(,"package")
## [1] "adegraphics"g2@g.args## $fullcircle
## [1] TRUE
##
## $xlim
## [1] -1.2 1.2
##
## $ylim
## [1] -1.2 1.2
##
## $scales
## $scales$draw
## [1] FALSE
The argument fullcircle can be updated a posteriori so that the original object is modified:
update(g2, fullcircle = FALSE)
g2@g.args## $fullcircle
## [1] FALSE
##
## $xlim
## [1] -0.8815395 0.9544397
##
## $ylim
## [1] -0.6344523 1.2015270
##
## $scales
## $scales$draw
## [1] FALSESeveral other methods have been defined for the ADEg class allowing to extract information, modify or combine objects:
getcall,getlatticecallandgetstats: these accessor methods return respectively theCall, thelattice.calland thestatsslots.getparameters: this method returns thetrellis.parand/or theadeg.parslots.show,printandplot: these methods display theADEgobject in the current device or in a new one.gettrellis: this method returns theADEgobject as atrellisobject. It can then be exploited using thelatticeandlatticeExtrapackages.superpose,+andadd.ADEg: these methods superpose twoADEggraphs. It returns a multiple graph object of classADEgS(see the The basic methods for superposition, juxtaposition and insertion section).insert: this method inserts anADEggraph in an existing one or in the current device. It returns anADEgSobject (see the The basic methods for superposition, juxtaposition and insertion section).cbindADEg,rbindADEg: these methods combine severalADEggraphs. It returns anADEgSobject (see the The basic methods for superposition, juxtaposition and insertion section).update: this method modifies the graphical parameters after theADEgcreation. It updates the current display and returns the modifiedADEg(see the Customizing a graph section).
For instance:
getcall(g1) ## equivalent to g1@Call## s1d.barchart(score = pca1$eig, p1d.horizontal = F, ppolygons.col = "white")A biplot-like graph can be obtained using the superpose method. The result is a multiple graph:
g3 <- s.label(pca1$li)
g4 <- s.arrow(5 * pca1$c1, add = TRUE)
class(g4)## [1] "ADEgS"
## attr(,"package")
## [1] "adegraphics"In addition, some object classes have specific methods. For instance, a zoom method is available for ADEg.S1 and ADEg.S2 classes. For the ADEg.S2 class, the method addhist (see the The basic methods for superposition, juxtaposition and insertion section) decorates a 2-D graph by adding marginal distributions as histograms and density lines (this method replaces and extends the s.hist function of ade4).
zoom(g3, zoom = 2, center = c(2, -2))
3 Multiple graph (ADEgS object)
The adegraphics package provides class ADEgS to manage easily the combination of several graphs. This class allows to deal with the superposition, insertion or juxtaposition of several graphs in a single object. An object of this class is a list containing several graphical objects and information about their positioning. Different ways to generate ADEgS objects are described below.
3.1 Slots and Methods
The class ADEgS is used to store multiple graphs. Different slots are associated to this class (use the symbol @ to extract information):
ADEglist: a list of graphs stored astrellis,ADEgand/orADEgSobjects.positions: a matrix containing the positions of the graphs. It has four columns and as many rows as the number of graphical objects in theADEglistslot. For each graph (i.e. row), it contains the coordinates of the bottom-left and top-right corners innpcunits (i.e. normalized parent coordinates varying between 0 and 1).add: a square binary matrix with as many rows and columns as the number of graphical objects in theADEglistslot. It allows to manage the superposition of graphs: the value at the i-th row and j-th column is equal to 1 if the j-th graphical object is superposed on the i-th. Otherwise, this value is equal to 0.Call: an object of classcallcontaining the matched call.
Several methods have been implemented to obtain information, edit or modify ADEgS objects. Several methods are inspired from the management of list in R:
[,[[and$: these methods extract one or more elements from theADEgSobject.getpositions,getgraphicsandgetcall: these methods return thepositions, theADEglistand theCallslots, respectively.namesandlength: these methods return the names and number of graphs contained in the object.[[<-andnames<-: these methods replace a graph or its name in anADEgSobject (acts on theADEglistslot).show,plotandprint: these methods display theADEgSobject in the current device or in a new one.superposeand+: these methods superpose two graphs. It returns a multiple graph object of classADEgS(see the The basic methods for superposition, juxtaposition and insertion section).insert: this method inserts a graph in an existing one or in the current device. It returns a multiple graph object of classADEgS(see the The basic methods for superposition, juxtaposition and insertion section).cbindADEg,rbindADEg: these methods combine several graphs. It returns anADEgSobject (see the The basic methods for superposition, juxtaposition and insertion section).update: this method modifies the names and/or thepositionsof the graphs contained in anADEgSobject. It updates the current display and returns the modifiedADEgS.
We will show in the next sections how these methods can be used to deal with ADEgS objects.
3.2 Creating an ADEgS object by hand
The ADEgS objects can be created by easy manipulation of several simple graphs. Some methods (e.g., insert, superpose) can be used to create a compilation of graphs by hand.
3.2.1 The basic methods for superposition, juxtaposition and insertion
The functions superpose, + and add.ADEg allow the superposition of an ADEg/ADEgS object on an ADEg/ADEgS object.
The vector olympic$score contains the total number of points computed for each participant. This vector is used to generate a factor partitioning the participants in two groups according to their final result (more or less than 8000 points):
fac.score <- factor(olympic$score < 8000, labels = c("MT8000", "LT8000"))These two groups can be represented on the PCA factorial map using the s.class function:
g5 <- s.class(pca1$li, fac.score, col = c("red", "blue"), chullSize = 1, ellipseSize = 0,
plabels.cex = 2, pbackground.col = "grey85", paxes.draw = TRUE)
The graph with the labels (object g3) can then be superposed on this one:
g6 <- superpose(g5, g3, plot = TRUE) ## equivalent to g5 + g3
class(g6)## [1] "ADEgS"
## attr(,"package")
## [1] "adegraphics"In the case of a superposition, the graphical parameters (e.g., background and limits) of the first graph (the one below) are used as a reference and applied to the second one (the one above). Note that it is also possible to use the add = TRUE argument in the call of a simple user function (functions described in Table 1) to perform a superposition. The graph g6 can also be obtained by:
g5
s.label(pca1$li, add = TRUE)The functions cbindADEg and rbindADEg allows to combine several graphical objects (ADEg, ADEgS or trellis) by rows or by columns. The new created ADEgS contains the list of the reduced graphs:
rbindADEg(cbindADEg(g2, g3), cbindADEg(g5, g6), plot = TRUE)
The function insert allows the insertion of a graphical object on another one (ADEg or ADEgS). It takes the position of the inserted graph as an argument:
g7 <- insert(g2, g6, posi = c(0.65, 0.65, 0.95, 0.95))
class(g7)## [1] "ADEgS"
## attr(,"package")
## [1] "adegraphics"The different methods associated to the ADEgS class allow to obtain information and to modify the multiple graph:
length(g7)## [1] 3names(g7)## [1] "g1" "g2" "X"names(g7) <- c("chulls", "labels", "cor")
class(g7[1])## [1] "ADEgS"
## attr(,"package")
## [1] "adegraphics"class(g7[[1]])## [1] "S2.class"
## attr(,"package")
## [1] "adegraphics"class(g7$chulls)## [1] "S2.class"
## attr(,"package")
## [1] "adegraphics"The multiple graph contains three simple graphs. It can be easily updated. For instance, the size of the inserted graph can be modified:
pos.mat <- getpositions(g7)
pos.mat## [,1] [,2] [,3] [,4]
## 0.00 0.00 1.00 1.00
## 0.00 0.00 1.00 1.00
## positions 0.65 0.65 0.95 0.95pos.mat[3,] <- c(0.1, 0.7, 0.3, 0.9)
update(g7, positions = pos.mat)
The graphs themselves can be modified, without affecting the global structure of the ADEgS object. Here, we replace the correlation circle by the barplot of eigenvalues:
g7[[3]] <- g1
g7
The addhist method adds univariate marginal distributions around an ADEg.S2 and returns an ADEgS object:
addhist(g3)
More examples are available in the help page by typing example(superpose), example(insert), example(add.ADEg) and example(addhist) in the R session.
3.2.2 The ADEgS function
The ADEgS function provides the most elementary and flexible way to create an ADEgS object. The different arguments of the function are:
adeglist: a list of severaltrellis,ADEgand/orADEgSobjects.positions: a matrix with four columns and as many rows as the number of graphical objects in theADEglistslot. For each subgraph, i.e. in each row, the coordinates of the top-right and the bottom-left hand corners are given innpcunits (i.e., normalized parent coordinates varying from 0 to 1).layout: an alternative way to specify the positions of graphs. It could be a vector of length 2 indicating the number of rows and columns used to split the device (similar tomfrowparameter in basic graphs). It could also be a matrix specifying the location of the graphs: each value in this matrix should be 0 or a positive integer (similar tolayoutfunction for basic graphs).add: a square matrix with as many rows and columns as the number of graphical objects in theADEglistslot. The value at the i-th row and j-th column is equal to 1 if the j-th graphical object is superposed to i-th one. Otherwise, this value is equal to 0.plot: a logical value indicating if the graphs should be displayed.
When users fill only one argument among positions, layout and add, the other values are automatically computed to define the ADEgS object.
We illustrate the different possibilities to create objects with the ADEgS function. Simple juxtaposition using a vector as layout:
ADEgS(adeglist = list(g2, g3), layout = c(1, 2))
Layout specified as a matrix:
mlay <- matrix(c(1, 1, 0, 1, 1, 0, 0, 0, 2), byrow = T, nrow = 3)
mlay## [,1] [,2] [,3]
## [1,] 1 1 0
## [2,] 1 1 0
## [3,] 0 0 2ADEgS(adeglist = list(g6, g2), layout = mlay)
Using the matrix of positions offers a very flexible way to arrange the different graphs:
mpos <- rbind(c(0, 0.3, 0.7, 1), c(0.5, 0, 1, 0.5))
ADEgS(adeglist = list(g3, g5), positions = mpos)
Lastly, superposition can also be specified using the add argument:
ADEgS(list(g5, g3), add = matrix(c(0, 1, 0, 0), byrow = TRUE, ncol = 2))
More examples are available in the help page by typing example(ADEgS) in the R session.
3.3 Automatic collections
The package adegraphics contains functionalities to create collections of graphs. These collections are based on a simple graph repeated for different groups of individuals, variables or axes. The building process of these collections is quite simple (definition of arguments in the call of a user function) and leads to the creation of an ADEgS object.
3.3.1 Partitioning the data (facets)
The adegraphics package allows to split up the data by one variable (factor) and to plot the subsets of data together. This possibility of conditional plot is available for all user functions (except the table.* functions) by setting the facets argument. This is directly inspired by the functionalities offered in the lattice and ggplot2 packages.
Let us consider the jv73 data set. The table jv73$morpho contains the measures of 6 variables describing the geomorphology of 92 sites. A PCA can be performed on this data set:
data(jv73)
pca2 <- dudi.pca(jv73$morpho, scannf = FALSE)
s.label(pca2$li)
The sites are located on 12 rivers (jv73$fac.riv) and it is possible to represent the PCA scores for each river using the facets argument:
g8 <- s.label(pca2$li, facets = jv73$fac.riv)
length(g8)## [1] 12names(g8)## [1] "Allaine" "Audeux" "Clauge" "Cuisance" "Cusancin"
## [6] "Dessoubre" "Doubs" "Doulonnes" "Drugeon" "Furieuse"
## [11] "Lison" "Loue"The ADEgS returned object contains the 12 plots. It is then possible to focus on a given river (e.g., the Doubs river) by considering only a subplot (e.g., type g8$Doubs). The facets functionality is very general and available for the majority of adegraphics functions. For instance, with the s.class function:
s.class(pca2$li, fac = jv73$fac.riv, col = rainbow(12), facets = jv73$fac.riv)
3.3.2 Multiple axes
All 2-D functions (i.e. s.*) returning an object inheriting from the ADEg.S2 class have the xax et yax arguments. These arguments allow to choose which column of the main argument (i.e. df) should be plotted as x and y axes. As in ade4, these two arguments can be simple integers. In adegraphics, the user can also specify vectors as xax and/or yax arguments to obtain multiple graphs. Here, we represent the different correlation circles associated to the first four PCA axes of the olympic data set:
pca1 <- dudi.pca(olympic$tab, scannf = FALSE, nf = 4)
g9 <- s.corcircle(pca1$co, xax = 1:2, yax = 3:4)
length(g9)## [1] 4names(g9)## [1] "x1y3" "x2y3" "x1y4" "x2y4"g9@positions## [,1] [,2] [,3] [,4]
## [1,] 0.0 0.5 0.5 1.0
## [2,] 0.5 0.5 1.0 1.0
## [3,] 0.0 0.0 0.5 0.5
## [4,] 0.5 0.0 1.0 0.53.3.3 Multiple score
All 1-D functions (i.e. s1d.*) returning an object inheriting from the ADEg.C1 or ADEg.S1 classes have the score argument. Usually, this argument should be a numeric vector but it is also possible to consider an object with several columns (data.frame or matrix). In this case, an ADEgS object is returned in which one graph by column is created. For instance for the olympic data set, we can represent the link between the global performance (fac.score) and the PCA scores on the first four axes (pca1$li):
dim(pca1$li)## [1] 33 4s1d.boxplot(pca1$li, fac.score, col = c("red", "blue"),
psub = list(position = "topleft", cex = 2))
3.3.4 Multiple variable
Some user functions (s1d.density, s1d.gauss, s1d.boxplot, s1d.class, s.class, s.image, s.traject, s.value, triangle.class) have an argument named fac or z. This argument can have several columns (data.frame or matrix) so that each column is used to create a separate graph. For instance, we can represent the distribution of the 6 environmental variables on the PCA factorial map of the jv73$tab data set:
s.value(pca2$li, pca2$tab, symbol = "circle")
3.3.5 Outputs of the ade4 package
Lastly, we reimplemented all the graphical functions of the ade4 package designed to represent the outputs of a multivariate analysis. The functions ade4::plot.*, ade4::kplot.*, ade4::scatter.* and ade4::score.* return ADEgS objects. It is now very easy to represent or modify these graphical outputs:
data(meaudret)
pca3 <- dudi.pca(meaudret$env, scannf = FALSE)
pca4 <- dudi.pca(meaudret$spe, scale = FALSE, scannf = FALSE)
coi1 <- coinertia(pca3, pca4, scannf = FALSE, nf = 3)
g10 <- plot(coi1)
class(g10)## [1] "ADEgS"
## attr(,"package")
## [1] "adegraphics"names(g10)## [1] "Xax" "Yax" "eig" "XYmatch" "Yloadings" "Xloadings"g10@Call## plot.coinertia(x = coi1)4 Customizing a graph
Compared to the ade4 package, the main advantage of adegraphics concerns the numerous possibilities to customize a graph using several graphical parameters. These parameters are stored in slots trellis.par, adeg.par and g.args (see the Slots and Methods section) of an ADEg object. These parameters can be defined during the creation of a graph or updated a posteriori (using the update method).
4.1 Parameters in trellis.par
The trellis.par slot is a list of parameters that are directly included in the call of functions of the lattice package. The name of parameters and their default value are given by the trellis.par.get function of lattice.
library(lattice)
sort(names(trellis.par.get()))## [1] "add.line" "add.text" "as.table"
## [4] "axis.components" "axis.line" "axis.text"
## [7] "background" "box.3d" "box.dot"
## [10] "box.rectangle" "box.umbrella" "clip"
## [13] "dot.line" "dot.symbol" "fontsize"
## [16] "grid.pars" "layout.heights" "layout.widths"
## [19] "panel.background" "par.main.text" "par.sub.text"
## [22] "par.xlab.text" "par.ylab.text" "par.zlab.text"
## [25] "plot.line" "plot.polygon" "plot.symbol"
## [28] "reference.line" "regions" "shade.colors"
## [31] "strip.background" "strip.border" "strip.shingle"
## [34] "superpose.line" "superpose.polygon" "superpose.symbol"Hence, modifications of some of these parameters will modify the graphical display of an ADEg object. For instance, margins are defined using layout.widths and layout.heights parameters, clip parameter allows to overpass panel boundaries and axis.line and axis.text allow to customize lines and text of axes.
d <- scale(olympic$tab)
g11 <- table.image(d, plot = FALSE)
g12 <- table.image(d, axis.line = list(col = "blue"), axis.text = list(col = "red"),
plot = FALSE)
ADEgS(c(g11, g12), layout = c(1, 2))
4.2 Parameters in adeg.par
The adeg.par slot is a list of graphical parameters specific to the adegraphics package. The name of parameters and their default value are available using the adegpar function which is inspired by the par function of the graphics package.
names(adegpar())## [1] "p1d" "parrows" "paxes" "pbackground" "pellipses"
## [6] "pgrid" "plabels" "plegend" "plines" "pnb"
## [11] "porigin" "ppalette" "ppoints" "ppolygons" "pSp"
## [16] "psub" "ptable"A description of these parameters is available in the help page of the function (?adegpar). Note that each adeg.par parameter starts by the letter ’p’ and its name relates to the type of graphical element considered (ptable is for tables display, ppoints for points, parrows for arrows, etc). Each element of this list can contain one or more sublists. Details on a sublist are obtained using its name either as a parameter of the adegpar function or after the $ symbol. For example, if we want to know the different parameters to manage the display of points:
adegpar("ppoints")## $ppoints
## $ppoints$alpha
## [1] 1
##
## $ppoints$cex
## [1] 1
##
## $ppoints$col
## [1] "black"
##
## $ppoints$pch
## [1] 20
##
## $ppoints$fill
## [1] "black"adegpar()$ppoints## $alpha
## [1] 1
##
## $cex
## [1] 1
##
## $col
## [1] "black"
##
## $pch
## [1] 20
##
## $fill
## [1] "black"The full list of available parameters is summarized in Figure 3.
The ordinate represents the different sublists and the abscissa gives the name of the parameters available in each sublist. Note that some row names have two keys separated by a dot: the first key indicates the first level of the sublist, etc. For example plabels.boxes is the sublist boxes of the sublist plabels. The parameters border,col, alpha, lwd, lty and draw in plabels.boxes allow to control the aspect of the boxes around labels.
According to the function called, only some of the full list of adeg.par parameters are useful to modify the graphical display. Figure 4 indicates which parameters can affect the display of an object created by a given user function. For example, the background (pbackground parameter) can be modified for all functions whereas the display of ellipses (pellipses parameter) affects only three functions.
4.2.1 Global assignment
The adegpar function allows to modify globally the values of graphical parameters so that changes will affect all subsequent displays. For example, we update the size/color of labels and add axes to a plot:
oldadegpar <- adegpar()
adegpar("plabels")## $plabels
## $plabels$alpha
## [1] 1
##
## $plabels$cex
## [1] 1
##
## $plabels$col
## [1] "black"
##
## $plabels$srt
## [1] "horizontal"
##
## $plabels$optim
## [1] FALSE
##
## $plabels$boxes
## $plabels$boxes$alpha
## [1] 1
##
## $plabels$boxes$border
## [1] "black"
##
## $plabels$boxes$col
## [1] "white"
##
## $plabels$boxes$draw
## [1] TRUE
##
## $plabels$boxes$lwd
## [1] 1
##
## $plabels$boxes$lty
## [1] 1g13 <- s.label(dfxy = pca1$li, plot = FALSE)
adegpar(plabels = list(col = "blue", cex = 1.5), paxes.draw = TRUE)
adegpar("plabels")## $plabels
## $plabels$alpha
## [1] 1
##
## $plabels$cex
## [1] 1.5
##
## $plabels$col
## [1] "blue"
##
## $plabels$srt
## [1] "horizontal"
##
## $plabels$optim
## [1] FALSE
##
## $plabels$boxes
## $plabels$boxes$alpha
## [1] 1
##
## $plabels$boxes$border
## [1] "black"
##
## $plabels$boxes$col
## [1] "white"
##
## $plabels$boxes$draw
## [1] TRUE
##
## $plabels$boxes$lwd
## [1] 1
##
## $plabels$boxes$lty
## [1] 1g14 <- s.label(dfxy = pca1$li, plot = FALSE)
ADEgS(c(g13, g14), layout = c(1, 2))
As the adegpar function can accept numerous graphical parameters, it can be used to define some graphical themes. The next releases of adegraphics will offer functionalities to easily create, edit and store graphical themes. Here, we reassign the original default parameters:
adegpar(oldadegpar)4.2.2 Local assignment
A second option is to update the graphical parameters locally so that the changes will only modify the object created. This is done using the dots (...) argument in the call to a user function. In this case, the default values of parameters in the global environment are not modified:
adegpar("ppoints")## $ppoints
## $ppoints$alpha
## [1] 1
##
## $ppoints$cex
## [1] 1
##
## $ppoints$col
## [1] "black"
##
## $ppoints$pch
## [1] 20
##
## $ppoints$fill
## [1] "black"s.label(dfxy = pca1$li, plabels.cex = 0, ppoints = list(col = c(2, 4, 5), cex = 1.5, pch = 15))
adegpar("ppoints")## $ppoints
## $ppoints$alpha
## [1] 1
##
## $ppoints$cex
## [1] 1
##
## $ppoints$col
## [1] "black"
##
## $ppoints$pch
## [1] 20
##
## $ppoints$fill
## [1] "black"In the previous example, we can see that parameters can be either specified using a ’.’ separator or a list. For instance, using plabels.cex = 0 or plabels = list(cex = 0) is strictly equivalent. Moreover, partial names can be used if there is no ambiguity (such as plab.ce = 0 in our example).
4.3 Parameters in g.args
The g.args slot is a list of parameters specific to the function used (and thus to the class of the returned object). Several parameters are very general and used in all adegraphics functions:
xlim,ylim: limits of the graph on the x and y axesmain,sub: main title and subtitlexlab,ylab: labels of the x and y axesscales: a list determining how the x and y axes (tick marks dans labels) are drawn; this is thescalesparameter of thexyplotfunction oflattice
The ADEg.S2 objects can also contain spatial information (map stored as a Spatial object or neighborhood stored as a nb object):
Sp,sp.layout: objects from thesppackage to display spatial objects,Spfor maps andsp.layoutfor spatial widgets as a North arrow, scale, etc.nbobject: object of classnborlistwto display neighbor graphs.
When the facets (see the Partitioning the data (facets) section) argument is used, users can modify the parameter samelimits: if it is TRUE, all graphs have the same limits whereas limits are computed for each subgraph independently when it is FALSE. For example, considering the jv73 data set, each subgraph is computed with its own limits and labels are then more scattered:
s.label(pca2$li, facets = jv73$fac.riv, samelimits = FALSE)
Several other g.args parameters can be updated according to the class of the created object (see Figure 2).
4.4 Parameters applied on a ADEgS
Users can either apply the changes to all graphs or to update only one graph. Of an ADEgS, to apply changes on all the graphs contained in an ADEgS, the syntax is similar to the one described for an ADEg object. For example, background color can be changed for all graphs in g10 using the pbackground.col parameter.
g15 <- plot(coi1, pbackground.col = "steelblue")
To change the parameters of a given graph, the name of the parameter must be preceded by the name of the subgraph. This supposes that the names of subgraphs are known. For example, to modify only two graphs:
names(g15)## [1] "Xax" "Yax" "eig" "XYmatch" "Yloadings" "Xloadings"plot(coi1, XYmatch.pbackground.col = "steelblue", XYmatch.pgrid.col = "red",
eig.ppolygons.col = "orange")
5 Using adegraphics functions in your package
In this section, we illustrate how adegraphics functionalities can be used to implement graphical functions in your own package. We created an objet of class track that contains a vector of distance and time.
tra1 <- list()
tra1$time <- runif(300)
tra1$distance <- tra1$time * 5 + rnorm(300)
class(tra1) <- "track"For an object of the class track, we wish to represent different components of the data:
an histogram of distances
an histogram of speeds (i.e., distance / time)
a 2D plot representing the distance, the time and the line corresponding to the linear model that predict distance by time
The corresponding multiple plot can be done using adegraphics functions:
g1 <- s1d.hist(tra1$distance, psub.text = "distance", ppolygons.col = "blue",
pgrid.draw = FALSE, plot = FALSE)
g2 <- s1d.hist(tra1$distance / tra1$time, psub.text = "speed", ppolygons.col = "red",
plot = FALSE)
g31 <- s.label(cbind(tra1$time, tra1$distance), paxes = list(aspectratio = "fill",
draw = TRUE), plot = FALSE)
g32 <- xyplot(tra1$distance ~ tra1$time, aspect = g31@adeg.par$paxes$aspectratio,
panel = function(x, y) {panel.lmline(x, y)})
g3 <- superpose(g31, g32)
G <- ADEgS(list(g1, g2, g3))
To facilitate the graphical representation of an object of class track, the simplest solution is to design a function plot for this class. We illustrate how to define such function with a particular emphasis on the management of graphical parameters. The function is provided below and we detail the different steps.
plot.track <- function(x, pos = -1, storeData = TRUE, plot = TRUE, ...) {
## step 1 : sort parameters for each graph
graphsnames <- c("histDist", "histSpeed", "regression")
sortparameters <- sortparamADEgS(..., graphsnames = graphsnames,
nbsubgraphs = c(1, 1, 2))
## step 2 : define default values for graphical parameters
params <- list()
params[[1]] <- list(psub = list(text = "distance"), ppolygons = list(col = "blue"),
pgrid = list(draw = FALSE))
params[[2]] <- list(psub = list(text = "speed"), ppolygons = list(col = "red"),
pgrid = list(draw = FALSE))
params[[3]] <- list()
params[[3]]$l1 <- list(paxes = list(aspectratio = "fill", draw = TRUE))
params[[3]]$l2 <- list()
names(params) <- graphsnames
sortparameters <- modifyList(params, sortparameters, keep.null = TRUE)
## step 3 : create each individual plot (ADEg)
g1 <- do.call("s1d.hist", c(list(score = substitute(x$distance), plot = FALSE,
storeData = storeData, pos = pos - 2), sortparameters[[1]]))
g2 <- do.call("s1d.hist", c(list(score = substitute(x$distance / x$time),
plot = FALSE, storeData = storeData, pos = pos - 2), sortparameters[[2]]))
g31 <- do.call("s.label", c(list(dfxy = substitute(cbind(x$time, x$distance)), plot =
FALSE, storeData = storeData, pos = pos - 2), sortparameters[[3]][[1]]))
g32 <- xyplot(x$distance ~ x$time, aspect = g31@adeg.par$paxes$aspectratio,
panel = function(x, y) {panel.lmline(x, y)})
g3 <- do.call("superpose", list(g31, g32))
g3@Call <- call("superpose", g31@Call, g32$call)
## step 4 : create the multiple plot (ADEgS)
lay <- matrix(1:3, 1, 3)
object <- new(Class = "ADEgS", ADEglist = list(g1, g2, g3), positions =
layout2position(lay), add = matrix(0, ncol = 3, nrow = 3),
Call = match.call())
names(object) <- graphsnames
if(plot)
print(object)
invisible(object)
}In the first step, the arguments given by the user through the dots (…) argument are managed. A name is given to each subgraph and stored in the vector graphnames. Then, the function sortparamADEgS associates the graphical parameters of the dots (…) argument to each subgraph. If a prefix is specified and matches the name of a graph (e.g., histDist.pbackground.col = grey), the parameter is applied only to the graphic specified (e.g., called histDist). If no prefix is specified (e.g., pbackground.col = grey), the parameter is applied to all subgraphs. The function sortparamADEgS returns a list (length equal to the number of subgraph) of lists of graphical parameters.
In the second step, default values for some graphical parameters are modified. The default parameters are stored in a list which has the same structure that the one produced by sortparamADEgS (i.e., names corresponding to those contained in graphsnames). Then, the modifyList function is applied to merge user and defaults values of paramaters (if a parameter is specified by the user and in the default, the value given by the user is used).
In the third step, each subgraph is created. Here, we create two C1.hist objects and superpose a S2.label object and a trellis one. The functions do.call and substitute are used to provide a pretty call for each subgraph (stored in the Call slot).
In a final step, the multiple graph is build through the creation of a new ADEgS object and possibly plotted.
The plot.track function can then be used by:
plot(tra1)
Graphical parameters can be modified by:
plot(tra1, histDist.ppoly.col = "green", pbackground.col = "grey")
6 Examples
6.1 Labels customization
data(meaudret)
g16 <- s.label(pca3$li, plot = FALSE)
g17 <- s.label(pca3$li, ppoints.col= "red", plabels = list(box = list(draw = FALSE),
optim = TRUE), plot = FALSE)
ADEgS(c(g16, g17), layout = c(1, 2))
6.2 Ellipses, stars and convex hulls
g18 <- s.class(pca3$li, fac = meaudret$design$season, plot = FALSE)
g19 <- s.class(pca3$li, fac = meaudret$design$season, ellipseSize = 0,
chullSize = 1, starSize = 0.5, col = TRUE, plot = FALSE)
g20 <- s.class(pca3$li, fac = meaudret$design$season, pellipses.lwd = 2,
pellipses.border = 2:5, pellipses.col = 2:5, plot = FALSE)
g21 <- s.class(pca3$li, fac = meaudret$design$season, ellipseSize = 0,
chullSize = 0, ppolygons.lwd = 2, plines.col = 2:5, starSize = 1.2, plot = FALSE)
ADEgS(c(g18, g19, g20, g21), layout = c(2, 2))
6.3 Values and legend
data(rpjdl)
coa2 <- dudi.coa(rpjdl$fau, scannf = FALSE, nf = 3)
g22 <- s.value(coa2$li, coa2$li[,3], plot = FALSE)
g23 <- s.value(coa2$li, coa2$li[,3], method = "color", ppoints.cex = 0.8,
plegend.size= 0.8, plot = FALSE)
g24 <- s.value(coa2$li, coa2$li[,3], plegend.size = 0.8, ppoints.cex = 0.8,
symbol = "square", method = "color", key = list(columns = 1),
col = colorRampPalette(c("yellow", "blue"))(6), plot = FALSE)
g25 <- s.value(coa2$li, coa2$li[, 3], center = 0, method = "size", ppoints.cex = 0.6,
symbol = "circle", col = c("yellow", "red"), plot = FALSE)
ADEgS(c(g22, g23, g24, g25), layout = c(2, 2))
6.4 1-D plot
score1 <- c(rnorm(1000, mean = -0.5, sd = 0.5), rnorm(1000, mean = 1))
fac1 <- rep(c("A", "B"), each = 1000)
g26 <- s1d.density(score1, fac1, pback.col = "grey75", plot = FALSE)
g27 <- s1d.density(score1, fac1, col = c(2, 4), plot = FALSE)
g28 <- s1d.density(score1, fac1, col = c(2, 4), p1d.reverse = TRUE, p1d.horizontal = FALSE,
p1d.rug.draw = FALSE, plot = FALSE)
g29 <- s1d.density(score1, fac1, col = c(2, 4), ppolygons.alpha = 0.2,
p1d = list(rug = list(tck = 1, line = FALSE)), plot = FALSE)
ADEgS(c(g26, g27, g28, g29), layout = c(2, 2))
6.5 Maps and neighbor graphs
library(Guerry)
library(sp)
data(gfrance85)
region.names <- data.frame(gfrance85)[, 5]
col.region <- colors()[c(149, 254, 468, 552, 26)]
g30 <- s.class(coordinates(gfrance85), region.names, porigin.include = FALSE, plot = FALSE)
g31 <- s.class(coordinates(gfrance85), region.names, ellipseSize = 0, starSize = 0,
Sp = gfrance85, pgrid.draw = FALSE, pSp.col = col.region[region.names], pSp.alpha = 0.4,
plot = FALSE)
ADEgS(c(g30, g31), layout = c(1, 2))
s.Spatial(gfrance85[,7:12])
data(mafragh, package = "ade4")
g32 <- s.label(mafragh$xy, nb = mafragh$nb, plot = FALSE)
g33 <- s.label(mafragh$xy, nb = mafragh$nb, pnb.ed.col = "red", plab.cex = 0,
pnb.node = list(cex = 3, col = "blue"), ppoints.col = "green", plot = FALSE)
ADEgS(c(g32, g33), layout = c(1, 2))
6.6 Ternary plots
data(euro123, package = "ade4")
df <- rbind.data.frame(euro123$in78, euro123$in86, euro123$in97)
row.names(df) <- paste(row.names(euro123$in78), rep(c(1, 2, 3), rep(12, 3)), sep = "")
g34 <- triangle.label(df, label = row.names(df), showposition = TRUE, plot = FALSE)
g35 <- triangle.label(euro123$in78, plabels.cex = 0, ppoints.cex = 2, addmean = TRUE,
show = FALSE, plot = FALSE)
ADEgS(c(g34, g35), layout = c(1, 2))
7 Appendix
This appendix summarizes the main changes between ade4 and adegraphics. Each line corresponds to a graphical argument defined in ade4 and its equivalent in adegraphics is given.
Arguments in ade4 |
Functions in ade4 |
g.args in adegraphics |
adeg.par in adegraphics |
|
|---|---|---|---|---|
abline.x |
table.cont |
ablineX |
||
abline.y |
table.cont |
ablineY |
||
abmean.x |
table.cont |
meanX |
||
abmean.y |
table.cont |
meanY |
||
addaxes |
s.arrow, s.chull, s.class, s.distri, s.image, s.kde2d, s.label, s.logo, s.match, s.traject, s.value, triangle.class, triangle.plot |
paxes.draw |
||
area |
s.arrow, s.chull, s.class, s.distri, s.image, s.kde2d, s.label, s.logo, s.match, s.traject, s.value |
Sp |
a Sp object |
|
axesell |
s.class, s.distri, triangle.class |
pellipses.axes.draw |
||
box |
s.corcircle, triangle.plot |
pbackground.box |
||
boxes |
s.arrow, s.label, sco.class, sco.label, sco.match |
plabels.boxes.draw |
||
cellipse |
s.class, s.distri, triangle.class |
ellipseSize |
||
cgrid |
s.arrow, s.class, s.chull, s.corcircle, s.distri, s.image, s.kde2d, s.label, s.logo, s.match, s.traject, s.value, sco.boxplot, sco.class, sco.distri, sco.gauss, sco.label, sco.match |
pgrid.nint |
both play on the grid mesh, but they are not strictly equivalent | |
clabel |
s.arrow, s.class, s.chull, s.corcircle, s.distri, s.kde2d, s.label, s.match, s.traject, sco.boxplot, sco.class, sco.distri, sco.gauss, sco.label, sco.match, triangle.plot |
plabels.cex |
||
clabel |
table.dist |
axis.text = list() lattice parameter |
||
clabel.col |
table.cont, table.paint, table.value |
axis.text = list() lattice parameter |
||
clabel.row |
table.cont, table.paint, table.value |
axis.text = list() lattice parameter |
||
clegend |
s.value, table.cont, table.value |
plegend.size ppoints.cex |
parameters setting the legend size | |
clegend |
table.paint |
plegend.size |
||
clogo |
s.logo |
ppoints.cex |
||
cneig |
s.image, s.kde2d, s.label, s.logo, s.value |
pnb.edge.lwd |
||
col.labels |
table.cont, table.paint, table.value |
labelsy |
||
contour |
s.arrow, s.class, s.chull, s.distri, s.image, s.kde2d, s.label, s.logo, s.match, s.traject, s.value |
Sp |
a Sp object |
|
contour.plot |
s.image |
region |
||
cpoints, cpoint |
s.arrow, s.class, s.chull, s.distri, s.kde2d, s.label, s.match, s.traject, s.value, sco.class, sco.label, sco.match, triangle.class, triangle.plot |
ppoints.cex |
||
csize |
s.value, table.cont, table.dist, table.paint, table.value |
ppoints.cex |
||
csize |
sco.distri |
sdSize |
||
cstar |
s.class, s.distri, triangle.class |
starSize |
||
csub |
s.arrow, s.chull, s.class, s.corcircle, s.distri, s.image, s.kde2d, s.label, s.logo, s.match, s.traject, s.value, sco.boxplot, sco.class, sco.distri, sco.gauss, sco.label, sco.match, triangle.class, triangle.plot |
psub.cex |
||
draw.line |
triangle.biplot, triangle.class, triangle.plot |
pgrid.draw |
||
edge |
s.arrow, s.match, s.traject |
parrows.length |
setting the length of the arrows to 0 is equivalent to edge = FALSE |
|
grid |
s.arrow, s.chull, s.class, s.corcircle, s.distri, s.image, s.kde2d, s.label, s.logo, s.match, s.traject, s.value, sco.boxplot, sco.class, sco.distri, sco.gauss, sco.label, sco.match, table.cont, table.dist, table.value |
pgrid.draw |
||
horizontal |
sco.class, sco.gauss, sco.label, sco.match |
p1d.horizontal |
||
image.plot |
s.image |
contour |
||
includeorigin, include.origin |
s.arrow, s.chull, s.class, s.distri, s.image, s.kde2d, s.label, s.logo, s.match, s.traject, s.value, sco.boxplot, sco.class, sco.distri, sco.gauss, sco.label, sco.match |
porigin.include |
||
kgrid |
s.image |
gridsize |
||
klogo |
s.logo |
no correspondence | ||
labeltriangle |
triangle.class , triangle.plot |
no correspondence | ||
legen |
sco.gauss |
labelplot |
||
neig |
s.image, s.kde2d, s.label, s.logo, s.value |
nbobject |
a nb object |
|
optchull |
s.chull |
chullSize |
||
origin |
s.arrow, s.chull, s.class, s.corcircle, s.distri, s.image, s.kde2d, s.label, s.logo, s.match, s.traject, s.value, sco.boxplot, sco.class, sco.distri, sco.gauss, sco.label, sco.match |
porigin.origin |
||
pch |
s.arrow, s.chull, s.class, s.distri, s.kde2d, s.label, s.match, s.traject, s.value, sco.boxplot, sco.class, sco.label, sco.match, triangle.class, triangle.plot, table.cont |
ppoints.pch |
||
pixmap |
s.arrow, s.chull, s.class, s.distri, s.image, s.kde2d, s.label, s.logo, s.match, s.traject, s.value |
no correspondence | ||
pos.lab |
sco.class, sco.label, sco.match |
p1d.labpos |
||
possub |
s.arrow, s.chull, s.class, s.corcircle, s.distri, s.image, s.kde2d, s.label, s.logo, s.match, s.traject, s.value, sco.class, sco.gauss, sco.label, sco.match, triangle.class, triangle.plot |
psub.pos |
||
rectlogo |
s.logo |
rect |
||
reverse |
sco.class, sco.gauss, sco.label, sco.match |
p1d.reverse |
||
row.labels |
table.cont, table.paint, table.value |
labelsx |
||
scale |
triangle.class, triangle.plot |
adjust |
||
show.position |
triangle.class, triangle.plot |
showposition |
||
sub |
s.arrow, s.chull, s.class, s.corcircle, s.distri, s.image, s.kde2d, s.label, s.logo, s.match, s.traject, s.value, sco.boxplot, sco.class, sco.distri, sco.gauss, sco.label, sco.match, triangle.class, triangle.plot |
psub.text |
||
y.rank |
sco.distri |
yrank |
||
zmax |
s.value |
set to default max(abs(z)) | ||
References
Dray, S., and A. B. Dufour. 2007. “The Ade4 Package: Implementing the Duality Diagram for Ecologists.” Journal of Statistical Software 22 (4): 1–20.
Murrell, P. 2005. R Graphics. Chapman & Hall/CRC.
Sarkar, D. 2008. Lattice: Multivariate Data Visualization with R. Springer.
The `adegraphics` package is a complete reimplementation of the graphical functionalities of the `ade4` package [@Dray2007]. The package has been initially designed to improve the representation of the outputs of multivariate analyses performed with `ade4` but as its graphical functionalities are very general, they can be used for other purposes. The `adegraphics` package provides a flexible environment to produce, edit and manipulate graphs. We adopted an *object oriented* approach (a graph is an object) using `S4` classes and methods and used the visualization system provided by the `lattice` [@Sarkar2008] and `grid` [@Murrell2005] packages. In `adegraphics`, graphs are R objects that can be edited, stored, combined, saved, removed, etc. Note that we tried to facilitate the handling of `adegraphics` by `ade4` users. Hence, the name of functions and parameters has been preserved in many cases. The main changes are listed in the appendix of this vignette so that it should be quite easy to use `adegraphics`. However, several new functionalities (graphical parameters, creation and manipulation of graphical objects, etc.) are now available and detailed in this vignette. The *adelist* mailing list can be used to send questions and/or comments on `adegraphics` (see
An overview of object classes ============================= In `adegraphics`, a user-level function produces a plot that is stored (and returned) as an object. The class architecture of the objects created by `adegraphics` functions is described in [Figure 1](#classes).
This class management highlights a hierarchy with two parent classes: - `ADEg` for simple graphs. It contains the display of a single data set using only one kind of representation (e.g., arrows, points, lines, etc.) - `ADEgS` for multiple graphs. It contains a collection of at least two simple graphs (`ADEg`, `trellis` or `ADEgS`) The `ADEg` class has five child classes which are also subdivided in several child classes. Each of these five child classes is dedicated for a particular graphical data representation: - `ADEg.S1`: unidimensional graph of a numeric score - `ADEg.S2`: bidimensional graph of xy coordinates (`matrix` or `data.frame` object) - `ADEg.C1`: bidimensional graph of a numeric score (bar chart or curve) - `ADEg.T`: heat map-like representation of a data table (`matrix`, `data.frame`, `dist` or `table` object) - `ADEg.Tr`: ternary plot of xyz coordinates (`matrix` or `data.frame` object) The `ADEg` class and its five child classes are virtual: it is not allowed to create object belonging to these classes. Users can only create objects belonging to child classes by calls to user functions (see the [User functions](#user-functions) section). Simple graph (`ADEg` object) ============================ In the `adegraphics` package, a graph is created by a call to a user function and stored as an R object. These functions allow to display the raw data but also the outputs of a multivariate analysis. The following sections describe the different graphical functions available in the package. User functions -------------- Several user functions are available to create a simple graph (stored as an `ADEg` object in R). Each function creates an object of a given class (see [Figure 1](#classes)). [Table 1](#functionsADEg) lists the different functions, their corresponding classes and a short description. The `ade4` users would not be lost: many functions have kept their names in `adegraphics`. The main changes are listed in [Table 2](#functionsADEgchanged).
Table 1: Graphical functions available in `adegraphics` Function Class of the returned object Description --------- ------------------------------ -------------- `s1d.barchart` `C1.barchart` 1-D plot of a numeric score by bars `s1d.curve` `C1.curve` 1-D plot of a numeric score linked by curves `s1d.curves` `C1.curves` 1-D plot of multiple scores linked by curves `s1d.density` `C1.density` 1-D plot of a numeric score by density curves `s1d.dotplot` `C1.dotplot` 1-D plot of a numeric score by dots `s1d.gauss` `C1.gauss` 1-D plot of a numeric score by Gaussian curves `s1d.hist` `C1.hist` 1-D plot of a numeric score by bars `s1d.interval` `C1.interval` 1-D plot of the interval between two numeric scores `s1d.boxplot` `S1.boxplot` 1-D box plot of a numeric score partitioned in classes `s1d.class` `S1.class` 1-D plot of a numeric score partitioned in classes `s1d.distri` `S1.distri` 1-D plot of a numeric score by means/tandard deviations computed using an external table of weights `s1d.label` `S1.label` 1-D plot of a numeric score with labels `s1d.match` `S1.match` 1-D plot of the matching between two numeric scores `s.arrow` `S2.arrow` 2-D scatter plot with arrows `s.class` `S2.class` 2-D scatter plot with a partition in classes `s.corcircle` `S2.corcircle` Correlation circle `s.density` `S2.density` 2-D scatter plot with kernel density estimation `s.distri` `S2.distri` 2-D scatter plot with means/standard deviations computed using an external table of weights `s.image` `S2.image` 2-D scatter plot with loess estimation of an additional numeric score `s.label` `S2.label` 2-D scatter plot with labels `s.logo` `S2.logo` 2-D scatter plot with logos (pixmap objects) `s.match` `S2.match` 2-D scatter plot of the matching between two sets of coordinates `s.Spatial` `S2.label` Mapping of a `Spatial*` object `s.traject` `S2.traject` 2-D scatter plot with trajectories `s.value` `S2.value` 2-D scatter plot with proportional symbols `table.image` `T.image` Heat map-like representation with colored cells `table.value` `T.value` or `T.cont` Heat map-like representation with proportional symbols `triangle.class` `Tr.class` Ternary plot with a partition in classes `triangle.label` `Tr.label` Ternary plot with labels `triangle.match` `Tr.match` Ternary plot of the matching between two sets of coordinates `triangle.traject` `Tr.match` Ternary plot with trajectories
Table 2: Changes in functions names between `ade4` and `adegraphics` Function in `ade4` Equivalence in `adegraphics` ------------------------------------------- ------------------------------ `table.cont`, `table.dist`, `table.value` `table.value` [^1] `table.paint` `table.image` `sco.boxplot` `s1d.boxplot` `sco.class` `s1d.class` `sco.distri` `s1d.distri` `sco.gauss` `s1d.gauss` `sco.label` `s1d.label` `sco.match` `s1d.match` `sco.quant` no equivalence `s.chull` `s.class`[^2] `s.kde2d` `s.density` `s.match.class` superposition of `s.match` and `s.class` `triangle.biplot` `triangle.match` `triangle.plot` `triangle.label` `s.multinom` `triangle.multinom` [^1]: The `table.value` function is now generic and can handle `dist` or `table` objects as arguments. [^2]: Convex hulls are now drawn by the `s.class` function (argument `chullSize`.) Arguments --------- The list of arguments of a function are given by the `args` function. ```{r label=chunk1} library(ade4) library(adegraphics) args(s.label) ``` Some arguments are very general and present in all user functions: - `plot`: a logical value indicating if the graph should be displayed - `storeData`: a logical value indicating if the data should be stored in the returned object. If `FALSE`, only the names of the data are stored. This allows to reduce the size of the returned object but it implies that the data should not be modified in the environment to plot again the graph. - `add`: a logical value indicating if the graph should be superposed on the graph already displayed in the current device (it replaces the argument `add.plot` in `ade4`). - `pos`: an integer indicating the position of the environment where the data are stored, relative to the environment where the function is called. Useful only if `storeData` is `FALSE`. - `…`: additional graphical parameters (see below) Some other arguments influence the graphical outputs and they are thus specific to the type of produced graph. [Figure 2](#gargsVSclass) summarizes some of these graphical parameters available for the different functions. We only reported the parameters stored in the `g.args` slot of the returned object (see the [Parameters in `g.args`](#gargs) section).
Figure 2: Specific arguments in each object class
```{r label=fig-gargsVSclass, include=TRUE, echo=FALSE, fig.width=7, fig.height=8}
source("gargsVSclass.R")
```
The `ade4` users would note that the names of some arguments have been modified in `adegraphics`. The [Appendix](#appendix) gives a full list of these modifications. Slots and Methods ----------------- A call to a graphical function (see the [User functions](#user-functions) section) returns an `ADEg` object. Each object is defined by a number of slots and several methods are associated to this class. Let us consider the `olympic` data set available in the `ade4` package. A principal component analysis (PCA) is applied on the `olympic$tab` table that contains the results for 33 participating athletes at the 1988 summer olympic games: ```{r label=chunk2} data(olympic) pca1 <- dudi.pca(olympic$tab, scannf = FALSE) ``` The barplot of eigenvalues is then drawn and stored in `g1`: ```{r label=plot1, fig.height=4, fig.width=4} g1 <- s1d.barchart(pca1$eig, p1d.horizontal = F, ppolygons.col = "white") ```
The class of the `g1` object is `C1.barchart` which extends the `ADEg` class: ```{r label=chunk3} class(g1) showClass("C1.barchart") ```
This object contains different slots: ```{r label=chunk4} slotNames(g1) ```
These slots are defined for each `ADEg` object and contain different types of information. The package `adegraphics` uses the capabilities of the `lattice` package to display a graph (by generating a `trellis` object). Hence, several slots contain information that will be passed in the call to the `lattice` functions: - `data`: a list containing information about the data. - `trellis.par`: a list of graphical parameters that are directly passed to the `lattice` functions using the `par.settings` argument (see the [Parameters in `trellis.par`](#trellispar) section). - `adeg.par`: a list of graphical parameters defined in `adegraphics`. The list of parameters can be obtained using the `adegpar` function (see the [Parameters in `adeg.par`](#adegpar) section). - `lattice.call`: a list of two elements containing the information required to create the `trellis` object: `graphictype` (the name of the `lattice` functions that should be used) and `arguments` (the list of parameter values required to obtain the `trellis` object). - `g.args`: a list containing at least the different values of the graphical arguments described in [Figure 2](#gargsVSclass) (see the [Parameters in `g.args`](#gargs) section). - `stats`: a list of internal preliminary computations performed to display the graph. - `s.misc`: a list of other internal parameters. - `Call`: an object of class `call` containing the matched call.
These different slots can be extracted using the `@` operator: ```{r label=chunk5} g1@data ```
All these slots are automatically filled during the object creation. The `trellis.par`, `adeg.par` and `g.args` can also be modified *a posteriori* using the `update` method (see the [Customizing a graph](#update) section). This allows to customize graphs after their creation. We consider the correlation circle that depicts the correlation between PCA axes and the results for each event: ```{r label=plot2, fig.width=4, fig.height=4} g2 <- s.corcircle(pca1$co) ``` ```{r label=chunk6} class(g2) g2@g.args ```
The argument `fullcircle` can be updated *a posteriori* so that the original object is modified: ```{r label=plot3, fig.width=4, fig.height=4} update(g2, fullcircle = FALSE) g2@g.args ``` Several other methods have been defined for the `ADEg` class allowing to extract information, modify or combine objects: - `getcall`, `getlatticecall` and `getstats`: these accessor methods return respectively the `Call`, the `lattice.call` and the `stats` slots. - `getparameters`: this method returns the `trellis.par` and/or the `adeg.par` slots. - `show`, `print` and `plot`: these methods display the `ADEg` object in the current device or in a new one. - `gettrellis`: this method returns the `ADEg` object as a `trellis` object. It can then be exploited using the `lattice` and `latticeExtra` packages. - `superpose`, `+` and `add.ADEg`: these methods superpose two `ADEg` graphs. It returns a multiple graph object of class `ADEgS` (see the [The basic methods for superposition, juxtaposition and insertion](#superpose) section). - `insert`: this method inserts an `ADEg` graph in an existing one or in the current device. It returns an `ADEgS` object (see the [The basic methods for superposition, juxtaposition and insertion](#superpose) section). - `cbindADEg`, `rbindADEg`: these methods combine several `ADEg` graphs. It returns an `ADEgS` object (see the [The basic methods for superposition, juxtaposition and insertion](#superpose) section). - `update`: this method modifies the graphical parameters after the `ADEg` creation. It updates the current display and returns the modified `ADEg` (see the [Customizing a graph](#update) section). For instance: ```{r label=chunk7} getcall(g1) ## equivalent to g1@Call ``` A biplot-like graph can be obtained using the `superpose` method. The result is a multiple graph: ```{r label=plot4, fig.width=4, fig.height=4} g3 <- s.label(pca1$li) g4 <- s.arrow(5 * pca1$c1, add = TRUE) class(g4) ``` In addition, some object classes have specific methods. For instance, a `zoom` method is available for `ADEg.S1` and `ADEg.S2` classes. For the `ADEg.S2` class, the method `addhist` (see the [The basic methods for superposition, juxtaposition and insertion](#superpose) section) decorates a 2-D graph by adding marginal distributions as histograms and density lines (this method replaces and extends the `s.hist` function of `ade4`). ```{r label=plot5, fig.width=4, fig.height=4} zoom(g3, zoom = 2, center = c(2, -2)) ``` Multiple graph (`ADEgS` object) =============================== The `adegraphics` package provides class `ADEgS` to manage easily the combination of several graphs. This class allows to deal with the superposition, insertion or juxtaposition of several graphs in a single object. An object of this class is a list containing several graphical objects and information about their positioning. Different ways to generate `ADEgS` objects are described below. Slots and Methods ----------------- The class `ADEgS` is used to store multiple graphs. Different slots are associated to this class (use the symbol `@` to extract information): - `ADEglist`: a list of graphs stored as `trellis`, `ADEg` and/or `ADEgS` objects. - `positions`: a matrix containing the positions of the graphs. It has four columns and as many rows as the number of graphical objects in the `ADEglist` slot. For each graph (i.e. row), it contains the coordinates of the bottom-left and top-right corners in `npc` units (i.e. normalized parent coordinates varying between 0 and 1). - `add`: a square binary matrix with as many rows and columns as the number of graphical objects in the `ADEglist` slot. It allows to manage the superposition of graphs: the value at the i-th row and j-th column is equal to 1 if the j-th graphical object is superposed on the i-th. Otherwise, this value is equal to 0. - `Call`: an object of class `call` containing the matched call. Several methods have been implemented to obtain information, edit or modify `ADEgS` objects. Several methods are inspired from the management of `list` in R: - `[`, `[[` and `$`: these methods extract one or more elements from the `ADEgS` object. - `getpositions`, `getgraphics` and `getcall`: these methods return the `positions`, the `ADEglist` and the `Call` slots, respectively. - `names` and `length`: these methods return the names and number of graphs contained in the object. - `[[<-` and `names<-`: these methods replace a graph or its name in an `ADEgS` object (acts on the `ADEglist` slot). - `show`, `plot` and `print`: these methods display the `ADEgS` object in the current device or in a new one. - `superpose` and `+`: these methods superpose two graphs. It returns a multiple graph object of class `ADEgS` (see the [The basic methods for superposition, juxtaposition and insertion](#superpose) section). - `insert`: this method inserts a graph in an existing one or in the current device. It returns a multiple graph object of class `ADEgS` (see the [The basic methods for superposition, juxtaposition and insertion](#superpose) section). - `cbindADEg`, `rbindADEg`: these methods combine several graphs. It returns an `ADEgS` object (see the [The basic methods for superposition, juxtaposition and insertion](#superpose) section). - `update`: this method modifies the names and/or the `positions` of the graphs contained in an `ADEgS` object. It updates the current display and returns the modified `ADEgS`. We will show in the next sections how these methods can be used to deal with `ADEgS` objects. Creating an `ADEgS` object by hand ---------------------------------- The `ADEgS` objects can be created by easy manipulation of several simple graphs. Some methods (e.g., `insert`, `superpose`) can be used to create a compilation of graphs by hand. ### The basic methods for superposition, juxtaposition and insertion The functions `superpose`, `+` and `add.ADEg` allow the superposition of an `ADEg`/`ADEgS` object on an `ADEg`/`ADEgS` object. The vector `olympic$score` contains the total number of points computed for each participant. This vector is used to generate a `factor` partitioning the participants in two groups according to their final result (more or less than 8000 points): ```{r label=chunk8} fac.score <- factor(olympic$score < 8000, labels = c("MT8000", "LT8000")) ``` These two groups can be represented on the PCA factorial map using the `s.class` function: ```{r label=plot6, fig.width=4, fig.height=4} g5 <- s.class(pca1$li, fac.score, col = c("red", "blue"), chullSize = 1, ellipseSize = 0, plabels.cex = 2, pbackground.col = "grey85", paxes.draw = TRUE) ``` The graph with the labels (object `g3`) can then be superposed on this one: ```{r label=plot7, fig.width=4, fig.height=4} g6 <- superpose(g5, g3, plot = TRUE) ## equivalent to g5 + g3 class(g6) ``` In the case of a superposition, the graphical parameters (e.g., background and limits) of the first graph (the one below) are used as a reference and applied to the second one (the one above). Note that it is also possible to use the `add = TRUE` argument in the call of a simple user function (functions described in [Table 1](#functionsADEg)) to perform a superposition. The graph `g6` can also be obtained by: ```{r label=chunk9, eval=FALSE} g5 s.label(pca1$li, add = TRUE) ``` The functions `cbindADEg` and `rbindADEg` allows to combine several graphical objects (`ADEg`, `ADEgS` or `trellis`) by rows or by columns. The new created `ADEgS` contains the list of the reduced graphs: ```{r label=plot8, fig.width=6, fig.height=6} rbindADEg(cbindADEg(g2, g3), cbindADEg(g5, g6), plot = TRUE) ``` The function `insert` allows the insertion of a graphical object on another one (`ADEg` or `ADEgS`). It takes the position of the inserted graph as an argument: ```{r label=plot9, fig.width=4, fig.height=4} g7 <- insert(g2, g6, posi = c(0.65, 0.65, 0.95, 0.95)) class(g7) ``` The different methods associated to the `ADEgS` class allow to obtain information and to modify the multiple graph: ```{r label=chunk10} length(g7) names(g7) names(g7) <- c("chulls", "labels", "cor") class(g7[1]) class(g7[[1]]) class(g7$chulls) ``` The multiple graph contains three simple graphs. It can be easily updated. For instance, the size of the inserted graph can be modified: ```{r label=plot10, fig.width=4, fig.height=4} pos.mat <- getpositions(g7) pos.mat pos.mat[3,] <- c(0.1, 0.7, 0.3, 0.9) update(g7, positions = pos.mat) ``` The graphs themselves can be modified, without affecting the global structure of the `ADEgS` object. Here, we replace the correlation circle by the barplot of eigenvalues: ```{r label=plot11, fig.width=4, fig.height=4} g7[[3]] <- g1 g7 ``` The `addhist` method adds univariate marginal distributions around an `ADEg.S2` and returns an `ADEgS` object: ```{r label=plot12, fig.width=4, fig.height=4} addhist(g3) ``` More examples are available in the help page by typing `example(superpose)`, `example(insert)`, `example(add.ADEg)` and `example(addhist)` in the R session. ### The `ADEgS` function The `ADEgS` function provides the most elementary and flexible way to create an `ADEgS` object. The different arguments of the function are: - `adeglist`: a list of several `trellis`, `ADEg` and/or `ADEgS` objects. - `positions`: a matrix with four columns and as many rows as the number of graphical objects in the `ADEglist` slot. For each subgraph, i.e. in each row, the coordinates of the top-right and the bottom-left hand corners are given in `npc` units (i.e., normalized parent coordinates varying from 0 to 1). - `layout`: an alternative way to specify the positions of graphs. It could be a vector of length 2 indicating the number of rows and columns used to split the device (similar to `mfrow` parameter in basic graphs). It could also be a matrix specifying the location of the graphs: each value in this matrix should be 0 or a positive integer (similar to `layout` function for basic graphs). - `add`: a square matrix with as many rows and columns as the number of graphical objects in the `ADEglist` slot. The value at the i-th row and j-th column is equal to 1 if the j-th graphical object is superposed to i-th one. Otherwise, this value is equal to 0. - `plot`: a logical value indicating if the graphs should be displayed. When users fill only one argument among `positions`, `layout` and `add`, the other values are automatically computed to define the `ADEgS` object. We illustrate the different possibilities to create objects with the `ADEgS` function. Simple juxtaposition using a vector as layout: ```{r label=plot13, fig.width=6, fig.height=3} ADEgS(adeglist = list(g2, g3), layout = c(1, 2)) ``` Layout specified as a matrix: ```{r label=plot14, fig.width=5, fig.height=5} mlay <- matrix(c(1, 1, 0, 1, 1, 0, 0, 0, 2), byrow = T, nrow = 3) mlay ADEgS(adeglist = list(g6, g2), layout = mlay) ``` Using the matrix of positions offers a very flexible way to arrange the different graphs: ```{r label=plot15, fig.width=5, fig.height=5} mpos <- rbind(c(0, 0.3, 0.7, 1), c(0.5, 0, 1, 0.5)) ADEgS(adeglist = list(g3, g5), positions = mpos) ``` Lastly, superposition can also be specified using the `add` argument: ```{r label=plot16, fig.width=4, fig.height=4} ADEgS(list(g5, g3), add = matrix(c(0, 1, 0, 0), byrow = TRUE, ncol = 2)) ``` More examples are available in the help page by typing `example(ADEgS)` in the R session. Automatic collections --------------------- The package `adegraphics` contains functionalities to create collections of graphs. These collections are based on a simple graph repeated for different groups of individuals, variables or axes. The building process of these collections is quite simple (definition of arguments in the call of a user function) and leads to the creation of an `ADEgS` object. ### Partitioning the data (`facets`) The `adegraphics` package allows to split up the data by one variable (`factor`) and to plot the subsets of data together. This possibility of conditional plot is available for all user functions (except the `table.*` functions) by setting the `facets` argument. This is directly inspired by the functionalities offered in the `lattice` and `ggplot2` packages. Let us consider the `jv73` data set. The table `jv73$morpho` contains the measures of 6 variables describing the geomorphology of 92 sites. A PCA can be performed on this data set: ```{r label=plot17, fig.width=4, fig.height=4} data(jv73) pca2 <- dudi.pca(jv73$morpho, scannf = FALSE) s.label(pca2$li) ``` The sites are located on 12 rivers (`jv73$fac.riv`) and it is possible to represent the PCA scores for each river using the `facets` argument: ```{r label=plot18, fig.width=7, fig.height=5.2} g8 <- s.label(pca2$li, facets = jv73$fac.riv) length(g8) names(g8) ``` The `ADEgS` returned object contains the 12 plots. It is then possible to focus on a given river (e.g., the Doubs river) by considering only a subplot (e.g., type `g8$Doubs`). The `facets` functionality is very general and available for the majority of `adegraphics` functions. For instance, with the `s.class` function: ```{r label=plot19, fig.width=7, fig.height=5.2} s.class(pca2$li, fac = jv73$fac.riv, col = rainbow(12), facets = jv73$fac.riv) ``` ### Multiple axes All 2-D functions (i.e. `s.*`) returning an object inheriting from the `ADEg.S2` class have the `xax` et `yax` arguments. These arguments allow to choose which column of the main argument (i.e. `df`) should be plotted as x and y axes. As in `ade4`, these two arguments can be simple integers. In `adegraphics`, the user can also specify vectors as `xax` and/or `yax` arguments to obtain multiple graphs. Here, we represent the different correlation circles associated to the first four PCA axes of the olympic data set: ```{r label=plot20, fig.width=6, fig.height=6} pca1 <- dudi.pca(olympic$tab, scannf = FALSE, nf = 4) g9 <- s.corcircle(pca1$co, xax = 1:2, yax = 3:4) length(g9) names(g9) g9@positions ``` ### Multiple score All 1-D functions (i.e. `s1d.*`) returning an object inheriting from the `ADEg.C1` or `ADEg.S1` classes have the `score` argument. Usually, this argument should be a numeric vector but it is also possible to consider an object with several columns (`data.frame` or `matrix`). In this case, an `ADEgS` object is returned in which one graph by column is created. For instance for the `olympic` data set, we can represent the link between the global performance (`fac.score`) and the PCA scores on the first four axes (`pca1$li`): ```{r label=plot21, fig.width=6, fig.height=6} dim(pca1$li) s1d.boxplot(pca1$li, fac.score, col = c("red", "blue"), psub = list(position = "topleft", cex = 2)) ``` ### Multiple variable Some user functions (`s1d.density`, `s1d.gauss`, `s1d.boxplot`, `s1d.class`, `s.class`, `s.image`, `s.traject`, `s.value`, `triangle.class`) have an argument named `fac` or `z`. This argument can have several columns (`data.frame` or `matrix`) so that each column is used to create a separate graph. For instance, we can represent the distribution of the 6 environmental variables on the PCA factorial map of the `jv73$tab` data set: ```{r label=plot22, fig.width=6, fig.height=4} s.value(pca2$li, pca2$tab, symbol = "circle") ``` ### Outputs of the `ade4` package Lastly, we reimplemented all the graphical functions of the `ade4` package designed to represent the outputs of a multivariate analysis. The functions `ade4::plot.*`, `ade4::kplot.*`, `ade4::scatter.*` and `ade4::score.*` return `ADEgS` objects. It is now very easy to represent or modify these graphical outputs: ```{r label=plot23, fig.width=6, fig.height=6} data(meaudret) pca3 <- dudi.pca(meaudret$env, scannf = FALSE) pca4 <- dudi.pca(meaudret$spe, scale = FALSE, scannf = FALSE) coi1 <- coinertia(pca3, pca4, scannf = FALSE, nf = 3) g10 <- plot(coi1) class(g10) names(g10) g10@Call ``` Customizing a graph =================== Compared to the `ade4` package, the main advantage of `adegraphics` concerns the numerous possibilities to customize a graph using several graphical parameters. These parameters are stored in slots `trellis.par`, `adeg.par` and `g.args` (see the [Slots and Methods](#slots) section) of an `ADEg` object. These parameters can be defined during the creation of a graph or updated *a posteriori* (using the `update` method). Parameters in `trellis.par` --------------------------- The `trellis.par` slot is a list of parameters that are directly included in the call of functions of the `lattice` package. The name of parameters and their default value are given by the `trellis.par.get` function of `lattice`. ```{r label=chunk11} library(lattice) sort(names(trellis.par.get())) ``` Hence, modifications of some of these parameters will modify the graphical display of an `ADEg` object. For instance, margins are defined using `layout.widths` and `layout.heights` parameters, `clip` parameter allows to overpass panel boundaries and `axis.line` and `axis.text` allow to customize lines and text of axes. ```{r label=plot24, fig.width=7, fig.height=3.5} d <- scale(olympic$tab) g11 <- table.image(d, plot = FALSE) g12 <- table.image(d, axis.line = list(col = "blue"), axis.text = list(col = "red"), plot = FALSE) ADEgS(c(g11, g12), layout = c(1, 2)) ``` Parameters in `adeg.par` ------------------------ The `adeg.par` slot is a list of graphical parameters specific to the `adegraphics` package. The name of parameters and their default value are available using the `adegpar` function which is inspired by the `par` function of the `graphics` package. ```{r label=chunk12} names(adegpar()) ``` A description of these parameters is available in the help page of the function (`?adegpar`). Note that each `adeg.par` parameter starts by the letter ’p’ and its name relates to the type of graphical element considered (`ptable` is for tables display, `ppoints` for points, `parrows` for arrows, etc). Each element of this list can contain one or more sublists. Details on a sublist are obtained using its name either as a parameter of the `adegpar` function or after the `$` symbol. For example, if we want to know the different parameters to manage the display of points: ```{r label=chunk13} adegpar("ppoints") adegpar()$ppoints ``` The full list of available parameters is summarized in [Figure 3](#paramVSparam).
Figure 3: Parameters that can be set with the `adegpar` function.
```{r label=fig-paramVSparam, echo=FALSE, fig.width=7, fig.height=7}
source("paramVSparam.R")
```
The ordinate represents the different sublists and the abscissa gives the name of the parameters available in each sublist. Note that some row names have two keys separated by a dot: the first key indicates the first level of the sublist, etc. For example `plabels.boxes` is the sublist `boxes` of the sublist `plabels`. The parameters `border`,`col`, `alpha`, `lwd`, `lty` and `draw` in `plabels.boxes` allow to control the aspect of the boxes around labels. According to the function called, only some of the full list of `adeg.par` parameters are useful to modify the graphical display. [Figure 4](#paramVSfunction) indicates which parameters can affect the display of an object created by a given user function. For example, the background (`pbackground` parameter) can be modified for all functions whereas the display of ellipses (`pellipses` parameter) affects only three functions.
Figure 4: Effect of `adeg.par` parameters in `adegraphics` functions.
```{r label=fig-paramVSfunction, echo=FALSE, fig.width=7, fig.height=10}
source("paramVSfunction.R")
```
### Global assignment The `adegpar` function allows to modify globally the values of graphical parameters so that changes will affect all subsequent displays. For example, we update the size/color of labels and add axes to a plot: ```{r label=plot25, fig.width=6, fig.height=3} oldadegpar <- adegpar() adegpar("plabels") g13 <- s.label(dfxy = pca1$li, plot = FALSE) adegpar(plabels = list(col = "blue", cex = 1.5), paxes.draw = TRUE) adegpar("plabels") g14 <- s.label(dfxy = pca1$li, plot = FALSE) ADEgS(c(g13, g14), layout = c(1, 2)) ``` As the `adegpar` function can accept numerous graphical parameters, it can be used to define some graphical themes. The next releases of `adegraphics` will offer functionalities to easily create, edit and store graphical themes. Here, we reassign the original default parameters: ```{r label=chunk14} adegpar(oldadegpar) ``` ### Local assignment A second option is to update the graphical parameters locally so that the changes will only modify the object created. This is done using the dots (`...`) argument in the call to a user function. In this case, the default values of parameters in the global environment are not modified: ```{r label=plot26, fig.width=4, fig.height=4} adegpar("ppoints") s.label(dfxy = pca1$li, plabels.cex = 0, ppoints = list(col = c(2, 4, 5), cex = 1.5, pch = 15)) adegpar("ppoints") ``` In the previous example, we can see that parameters can be either specified using a ’`.`’ separator or a list. For instance, using `plabels.cex = 0` or `plabels = list(cex = 0)` is strictly equivalent. Moreover, partial names can be used if there is no ambiguity (such as `plab.ce = 0` in our example). Parameters in `g.args` ---------------------- The `g.args` slot is a list of parameters specific to the function used (and thus to the class of the returned object). Several parameters are very general and used in all `adegraphics` functions: - `xlim`, `ylim`: limits of the graph on the x and y axes - `main`, `sub`: main title and subtitle - `xlab`, `ylab`: labels of the x and y axes - `scales`: a list determining how the x and y axes (tick marks dans labels) are drawn; this is the `scales` parameter of the `xyplot` function of `lattice` The `ADEg.S2` objects can also contain spatial information (map stored as a `Spatial` object or neighborhood stored as a `nb` object): - `Sp`, `sp.layout`: objects from the `sp` package to display spatial objects, `Sp` for maps and `sp.layout` for spatial widgets as a North arrow, scale, etc. - `nbobject`: object of class `nb` or `listw` to display neighbor graphs. When the `facets` (see the [Partitioning the data (`facets`)](#facets) section) argument is used, users can modify the parameter `samelimits`: if it is `TRUE`, all graphs have the same limits whereas limits are computed for each subgraph independently when it is `FALSE`. For example, considering the `jv73` data set, each subgraph is computed with its own limits and labels are then more scattered: ```{r label=plot27, fig.width=7, fig.height=5.2} s.label(pca2$li, facets = jv73$fac.riv, samelimits = FALSE) ``` Several other `g.args` parameters can be updated according to the class of the created object (see [Figure 2](#gargsVSclass)). Parameters applied on a `ADEgS` ------------------------------- Users can either apply the changes to all graphs or to update only one graph. Of an `ADEgS`, to apply changes on all the graphs contained in an `ADEgS`, the syntax is similar to the one described for an `ADEg` object. For example, background color can be changed for all graphs in `g10` using the `pbackground.col` parameter. ```{r label=plot28, fig.width=6, fig.height=6} g15 <- plot(coi1, pbackground.col = "steelblue") ``` To change the parameters of a given graph, the name of the parameter must be preceded by the name of the subgraph. This supposes that the names of subgraphs are known. For example, to modify only two graphs: ```{r label=plot29, fig.width=6, fig.height=6} names(g15) plot(coi1, XYmatch.pbackground.col = "steelblue", XYmatch.pgrid.col = "red", eig.ppolygons.col = "orange") ``` Using `adegraphics` functions in your package ============================================= In this section, we illustrate how `adegraphics` functionalities can be used to implement graphical functions in your own package. We created an objet of class `track` that contains a vector of distance and time. ```{r label=chunk15} tra1 <- list() tra1$time <- runif(300) tra1$distance <- tra1$time * 5 + rnorm(300) class(tra1) <- "track" ``` For an object of the class `track`, we wish to represent different components of the data: - an histogram of distances - an histogram of speeds (i.e., distance / time) - a 2D plot representing the distance, the time and the line corresponding to the linear model that predict distance by time The corresponding multiple plot can be done using `adegraphics` functions: ```{r label=plot30, fig.width=7, fig.height=2.3} g1 <- s1d.hist(tra1$distance, psub.text = "distance", ppolygons.col = "blue", pgrid.draw = FALSE, plot = FALSE) g2 <- s1d.hist(tra1$distance / tra1$time, psub.text = "speed", ppolygons.col = "red", plot = FALSE) g31 <- s.label(cbind(tra1$time, tra1$distance), paxes = list(aspectratio = "fill", draw = TRUE), plot = FALSE) g32 <- xyplot(tra1$distance ~ tra1$time, aspect = g31@adeg.par$paxes$aspectratio, panel = function(x, y) {panel.lmline(x, y)}) g3 <- superpose(g31, g32) G <- ADEgS(list(g1, g2, g3)) ``` To facilitate the graphical representation of an object of class `track`, the simplest solution is to design a function `plot` for this class. We illustrate how to define such function with a particular emphasis on the management of graphical parameters. The function is provided below and we detail the different steps. ```{r label=chunk16} plot.track <- function(x, pos = -1, storeData = TRUE, plot = TRUE, ...) { ## step 1 : sort parameters for each graph graphsnames <- c("histDist", "histSpeed", "regression") sortparameters <- sortparamADEgS(..., graphsnames = graphsnames, nbsubgraphs = c(1, 1, 2)) ## step 2 : define default values for graphical parameters params <- list() params[[1]] <- list(psub = list(text = "distance"), ppolygons = list(col = "blue"), pgrid = list(draw = FALSE)) params[[2]] <- list(psub = list(text = "speed"), ppolygons = list(col = "red"), pgrid = list(draw = FALSE)) params[[3]] <- list() params[[3]]$l1 <- list(paxes = list(aspectratio = "fill", draw = TRUE)) params[[3]]$l2 <- list() names(params) <- graphsnames sortparameters <- modifyList(params, sortparameters, keep.null = TRUE) ## step 3 : create each individual plot (ADEg) g1 <- do.call("s1d.hist", c(list(score = substitute(x$distance), plot = FALSE, storeData = storeData, pos = pos - 2), sortparameters[[1]])) g2 <- do.call("s1d.hist", c(list(score = substitute(x$distance / x$time), plot = FALSE, storeData = storeData, pos = pos - 2), sortparameters[[2]])) g31 <- do.call("s.label", c(list(dfxy = substitute(cbind(x$time, x$distance)), plot = FALSE, storeData = storeData, pos = pos - 2), sortparameters[[3]][[1]])) g32 <- xyplot(x$distance ~ x$time, aspect = g31@adeg.par$paxes$aspectratio, panel = function(x, y) {panel.lmline(x, y)}) g3 <- do.call("superpose", list(g31, g32)) g3@Call <- call("superpose", g31@Call, g32$call) ## step 4 : create the multiple plot (ADEgS) lay <- matrix(1:3, 1, 3) object <- new(Class = "ADEgS", ADEglist = list(g1, g2, g3), positions = layout2position(lay), add = matrix(0, ncol = 3, nrow = 3), Call = match.call()) names(object) <- graphsnames if(plot) print(object) invisible(object) } ``` In the first step, the arguments given by the user through the dots (…) argument are managed. A name is given to each subgraph and stored in the vector `graphnames`. Then, the function `sortparamADEgS` associates the graphical parameters of the dots (…) argument to each subgraph. If a prefix is specified and matches the name of a graph (e.g., `histDist.pbackground.col = grey`), the parameter is applied only to the graphic specified (e.g., called `histDist`). If no prefix is specified (e.g., `pbackground.col = grey`), the parameter is applied to all subgraphs. The function `sortparamADEgS` returns a list (length equal to the number of subgraph) of lists of graphical parameters.\ In the second step, default values for some graphical parameters are modified. The default parameters are stored in a list which has the same structure that the one produced by `sortparamADEgS` (i.e., names corresponding to those contained in `graphsnames`). Then, the `modifyList` function is applied to merge user and defaults values of paramaters (if a parameter is specified by the user and in the default, the value given by the user is used).\ In the third step, each subgraph is created. Here, we create two `C1.hist` objects and superpose a `S2.label` object and a `trellis` one. The functions `do.call` and `substitute` are used to provide a pretty call for each subgraph (stored in the `Call` slot).\ In a final step, the multiple graph is build through the creation of a new `ADEgS` object and possibly plotted.\ The `plot.track` function can then be used by: ```{r label=plot31, fig.width=7, fig.height=2.3} plot(tra1) ``` Graphical parameters can be modified by: ```{r label=plot32, fig.width=7, fig.height=2.3} plot(tra1, histDist.ppoly.col = "green", pbackground.col = "grey") ``` Examples ======== Labels customization -------------------- ```{r label=plot33, fig.width=6, fig.height=3} data(meaudret) g16 <- s.label(pca3$li, plot = FALSE) g17 <- s.label(pca3$li, ppoints.col= "red", plabels = list(box = list(draw = FALSE), optim = TRUE), plot = FALSE) ADEgS(c(g16, g17), layout = c(1, 2)) ``` Ellipses, stars and convex hulls -------------------------------- ```{r label=plot34, fig.width=6, fig.height=6} g18 <- s.class(pca3$li, fac = meaudret$design$season, plot = FALSE) g19 <- s.class(pca3$li, fac = meaudret$design$season, ellipseSize = 0, chullSize = 1, starSize = 0.5, col = TRUE, plot = FALSE) g20 <- s.class(pca3$li, fac = meaudret$design$season, pellipses.lwd = 2, pellipses.border = 2:5, pellipses.col = 2:5, plot = FALSE) g21 <- s.class(pca3$li, fac = meaudret$design$season, ellipseSize = 0, chullSize = 0, ppolygons.lwd = 2, plines.col = 2:5, starSize = 1.2, plot = FALSE) ADEgS(c(g18, g19, g20, g21), layout = c(2, 2)) ``` Values and legend ----------------- ```{r label=plot35, fig.width=6, fig.height=6} data(rpjdl) coa2 <- dudi.coa(rpjdl$fau, scannf = FALSE, nf = 3) g22 <- s.value(coa2$li, coa2$li[,3], plot = FALSE) g23 <- s.value(coa2$li, coa2$li[,3], method = "color", ppoints.cex = 0.8, plegend.size= 0.8, plot = FALSE) g24 <- s.value(coa2$li, coa2$li[,3], plegend.size = 0.8, ppoints.cex = 0.8, symbol = "square", method = "color", key = list(columns = 1), col = colorRampPalette(c("yellow", "blue"))(6), plot = FALSE) g25 <- s.value(coa2$li, coa2$li[, 3], center = 0, method = "size", ppoints.cex = 0.6, symbol = "circle", col = c("yellow", "red"), plot = FALSE) ADEgS(c(g22, g23, g24, g25), layout = c(2, 2)) ``` 1-D plot -------- ```{r label=plot36, fig.width=6, fig.height=6} score1 <- c(rnorm(1000, mean = -0.5, sd = 0.5), rnorm(1000, mean = 1)) fac1 <- rep(c("A", "B"), each = 1000) g26 <- s1d.density(score1, fac1, pback.col = "grey75", plot = FALSE) g27 <- s1d.density(score1, fac1, col = c(2, 4), plot = FALSE) g28 <- s1d.density(score1, fac1, col = c(2, 4), p1d.reverse = TRUE, p1d.horizontal = FALSE, p1d.rug.draw = FALSE, plot = FALSE) g29 <- s1d.density(score1, fac1, col = c(2, 4), ppolygons.alpha = 0.2, p1d = list(rug = list(tck = 1, line = FALSE)), plot = FALSE) ADEgS(c(g26, g27, g28, g29), layout = c(2, 2)) ``` Maps and neighbor graphs ------------------------ ```{r label=plot37, fig.width=6, fig.height=3} library(Guerry) library(sp) data(gfrance85) region.names <- data.frame(gfrance85)[, 5] col.region <- colors()[c(149, 254, 468, 552, 26)] g30 <- s.class(coordinates(gfrance85), region.names, porigin.include = FALSE, plot = FALSE) g31 <- s.class(coordinates(gfrance85), region.names, ellipseSize = 0, starSize = 0, Sp = gfrance85, pgrid.draw = FALSE, pSp.col = col.region[region.names], pSp.alpha = 0.4, plot = FALSE) ADEgS(c(g30, g31), layout = c(1, 2)) ``` ```{r label=plot38, fig.width=6, fig.height=4} s.Spatial(gfrance85[,7:12]) ``` ```{r label=plot39, fig.width=6, fig.height=3} data(mafragh, package = "ade4") g32 <- s.label(mafragh$xy, nb = mafragh$nb, plot = FALSE) g33 <- s.label(mafragh$xy, nb = mafragh$nb, pnb.ed.col = "red", plab.cex = 0, pnb.node = list(cex = 3, col = "blue"), ppoints.col = "green", plot = FALSE) ADEgS(c(g32, g33), layout = c(1, 2)) ``` Ternary plots ------------- ```{r label=plot40, fig.width=6, fig.height=3} data(euro123, package = "ade4") df <- rbind.data.frame(euro123$in78, euro123$in86, euro123$in97) row.names(df) <- paste(row.names(euro123$in78), rep(c(1, 2, 3), rep(12, 3)), sep = "") g34 <- triangle.label(df, label = row.names(df), showposition = TRUE, plot = FALSE) g35 <- triangle.label(euro123$in78, plabels.cex = 0, ppoints.cex = 2, addmean = TRUE, show = FALSE, plot = FALSE) ADEgS(c(g34, g35), layout = c(1, 2)) ``` Appendix ============================= This appendix summarizes the main changes between `ade4` and `adegraphics`. Each line corresponds to a graphical argument defined in `ade4` and its equivalent in `adegraphics` is given. | Arguments in `ade4` | Functions in `ade4` | `g.args` in `adegraphics` | `adeg.par` in `adegraphics` | | | ------------------- | ----------------------| ----------------------------| ------------------------------|---| | `abline.x` | `table.cont` | `ablineX` | | | | `abline.y` | `table.cont` | `ablineY` | | | | `abmean.x` | `table.cont` | `meanX` | | | | `abmean.y` | `table.cont` | `meanY` | | | | `addaxes` | `s.arrow`, `s.chull`, `s.class`, `s.distri`, `s.image`, `s.kde2d`, `s.label`, `s.logo`, `s.match`, `s.traject`, `s.value`, `triangle.class`, `triangle.plot` | |`paxes.draw` | | | `area` | `s.arrow`, `s.chull`, `s.class`, `s.distri`, `s.image`, `s.kde2d`, `s.label`, `s.logo`, `s.match`, `s.traject`, `s.value` | `Sp` | | a `Sp` object | | `axesell` | `s.class`, `s.distri`, `triangle.class` | | `pellipses.axes.draw` | | | `box` | `s.corcircle`, `triangle.plot` | | `pbackground.box` | | | `boxes` | `s.arrow`, `s.label`, `sco.class`, `sco.label`, `sco.match` | | `plabels.boxes.draw` | | | `cellipse` | `s.class`, `s.distri`, `triangle.class` | `ellipseSize` | | | | `cgrid` | `s.arrow`, `s.class`, `s.chull`, `s.corcircle`, `s.distri`, `s.image`, `s.kde2d`, `s.label`, `s.logo`, `s.match`, `s.traject`, `s.value`, `sco.boxplot`, `sco.class`, `sco.distri`, `sco.gauss`, `sco.label`, `sco.match` | | `pgrid.nint` | both play on the grid mesh, but they are not strictly equivalent | | `clabel` | `s.arrow`, `s.class`, `s.chull`, `s.corcircle`, `s.distri`, `s.kde2d`, `s.label`, `s.match`, `s.traject`, `sco.boxplot`, `sco.class`, `sco.distri`, `sco.gauss`, `sco.label`, `sco.match`, `triangle.plot` | | `plabels.cex` | | | `clabel` | `table.dist` | | | `axis.text = list()` `lattice` parameter | | `clabel.col` | `table.cont`, `table.paint`, `table.value` | | | `axis.text = list()` `lattice` parameter | | `clabel.row` | `table.cont`, `table.paint`, `table.value` | | | `axis.text = list()` `lattice` parameter | | `clegend` | `s.value`, `table.cont`, `table.value` | | `plegend.size` `ppoints.cex` | parameters setting the legend size | | `clegend` | `table.paint` | | `plegend.size` | | | `clogo` | `s.logo` | | `ppoints.cex` | | | `cneig` | `s.image`, `s.kde2d`, `s.label`, `s.logo`, `s.value` | | `pnb.edge.lwd` | | | `col.labels` | `table.cont`, `table.paint`, `table.value` | `labelsy` | | | | `contour` | `s.arrow`, `s.class`, `s.chull`, `s.distri`, `s.image`, `s.kde2d`, `s.label`, `s.logo`, `s.match`, `s.traject`, `s.value` | `Sp` | | a `Sp` object | | `contour.plot` | `s.image` | `region` | | | | `cpoints`, `cpoint` | `s.arrow`, `s.class`, `s.chull`, `s.distri`, `s.kde2d`, `s.label`, `s.match`, `s.traject`, `s.value`, `sco.class`, `sco.label`, `sco.match`, `triangle.class`, `triangle.plot` | | `ppoints.cex` | | | `csize` | `s.value`, `table.cont`, `table.dist`, `table.paint`, `table.value` | `ppoints.cex` | | | | `csize` | `sco.distri` | `sdSize` | | | | `cstar` | `s.class`, `s.distri`, `triangle.class` | `starSize` | | | | `csub` | `s.arrow`, `s.chull`, `s.class`, `s.corcircle`, `s.distri`, `s.image`, `s.kde2d`, `s.label`, `s.logo`, `s.match`, `s.traject`, `s.value`, `sco.boxplot`, `sco.class`, `sco.distri`, `sco.gauss`, `sco.label`, `sco.match`, `triangle.class`, `triangle.plot` | | `psub.cex` | | | `draw.line` | `triangle.biplot`, `triangle.class`, `triangle.plot` | | `pgrid.draw` | | | `edge` | `s.arrow`, `s.match`, `s.traject` | | `parrows.length` | setting the length of the arrows to 0 is equivalent to `edge = FALSE` | | `grid` | `s.arrow`, `s.chull`, `s.class`, `s.corcircle`, `s.distri`, `s.image`, `s.kde2d`, `s.label`, `s.logo`, `s.match`, `s.traject`, `s.value`, `sco.boxplot`, `sco.class`, `sco.distri`, `sco.gauss`, `sco.label`, `sco.match`, `table.cont`, `table.dist`, `table.value` | | `pgrid.draw` | | | `horizontal` | `sco.class`, `sco.gauss`, `sco.label`, `sco.match` | | `p1d.horizontal` | | | `image.plot` | `s.image` | `contour` | | | | `includeorigin`, `include.origin` | `s.arrow`, `s.chull`, `s.class`, `s.distri`, `s.image`, `s.kde2d`, `s.label`, `s.logo`, `s.match`, `s.traject`, `s.value`, `sco.boxplot`, `sco.class`, `sco.distri`, `sco.gauss`, `sco.label`, `sco.match` | | `porigin.include` | | | `kgrid` | `s.image` | `gridsize` | | | | `klogo` | `s.logo` | | | no correspondence | | `labeltriangle` | `triangle.class` , `triangle.plot` | | | no correspondence | | `legen` | `sco.gauss` | `labelplot` | | | | `neig` | `s.image`, `s.kde2d`, `s.label`, `s.logo`, `s.value` | `nbobject` | | a `nb` object | | `optchull` | `s.chull` | `chullSize` | | | | `origin` | `s.arrow`, `s.chull`, `s.class`, `s.corcircle`, `s.distri`, `s.image`, `s.kde2d`, `s.label`, `s.logo`, `s.match`, `s.traject`, `s.value`, `sco.boxplot`, `sco.class`, `sco.distri`, `sco.gauss`, `sco.label`, `sco.match` | | `porigin.origin` | | | `pch` | `s.arrow`, `s.chull`, `s.class`, `s.distri`, `s.kde2d`, `s.label`, `s.match`, `s.traject`, `s.value`, `sco.boxplot`, `sco.class`, `sco.label`, `sco.match`, `triangle.class`, `triangle.plot`, `table.cont` | | `ppoints.pch` | | | `pixmap` | `s.arrow`, `s.chull`, `s.class`, `s.distri`, `s.image`, `s.kde2d`, `s.label`, `s.logo`, `s.match`, `s.traject`, `s.value` | | | no correspondence | | `pos.lab` | `sco.class`, `sco.label`, `sco.match` | | `p1d.labpos` | | | `possub` | `s.arrow`, `s.chull`, `s.class`, `s.corcircle`, `s.distri`, `s.image`, `s.kde2d`, `s.label`, `s.logo`, `s.match`, `s.traject`, `s.value`, `sco.class`, `sco.gauss`, `sco.label`, `sco.match`, `triangle.class`, `triangle.plot` | | `psub.pos` | | | `rectlogo` | `s.logo` | `rect` | | | | `reverse` | `sco.class`, `sco.gauss`, `sco.label`, `sco.match` | | `p1d.reverse` | | | `row.labels` | `table.cont`, `table.paint`, `table.value` | `labelsx` | | | | `scale` | `triangle.class`, `triangle.plot` | `adjust` | | | | `show.position` | `triangle.class`, `triangle.plot` | `showposition` | | | | `sub` | `s.arrow`, `s.chull`, `s.class`, `s.corcircle`, `s.distri`, `s.image`, `s.kde2d`, `s.label`, `s.logo`, `s.match`, `s.traject`, `s.value`, `sco.boxplot`, `sco.class`, `sco.distri`, `sco.gauss`, `sco.label`, `sco.match`, `triangle.class`, `triangle.plot` | | `psub.text` | | | `y.rank` | `sco.distri` | `yrank` | | | | `zmax` | `s.value` | | | set to default max(abs(z)) | | | | | | | References ============================= adegraphics/vignettes/paramVSparam.R 0000644 0001751 0000144 00000002432 12607215420 017267 0 ustar hornik users Adegpar <- adegpar() names <- names(Adegpar) xx <- as.null() yy <- as.null() for(i in 1:length(Adegpar)) { if(is.list(Adegpar[[i]])) { for(j in 1:length(Adegpar[[i]])) { if(is.list(Adegpar[[i]][[j]])) { ## sublist of list xx <- c(xx, paste(names(Adegpar)[i], '.', names(Adegpar[[i]])[j], sep = "")) yy <- c(yy, names(Adegpar[[i]][[j]])) } else { yy <- c(yy, names(Adegpar[[i]])[j]) xx <- c(xx, names[i]) } } } else xx <- c(xx, names[i]) } yy <- unique(yy) xx <- unique(xx) paramVSsub <- data.frame(matrix(0, nrow = length(yy), ncol = length(xx))) row.names(paramVSsub) <- yy colnames(paramVSsub) <- xx ## filling for(i in 1:length(Adegpar)) { if(is.list(Adegpar[[i]])) { for(j in 1:length(Adegpar[[i]])) { if(is.list(Adegpar[[i]][[j]])) ## sublistof list paramVSsub[names(Adegpar[[i]][[j]]), paste(names(Adegpar)[i], '.', names(Adegpar[[i]])[j], sep = "")] <- 100 else paramVSsub[names(Adegpar[[i]])[j], names[i]] <- 100 } } } table.value(t(paramVSsub), axis.text = list(cex = 0.9), symbol = "circle", plegend.drawKey = FALSE, ppoints.cex = 0.3, ptable.y = list(srt = 60, pos = "left"), ptable.margin = list(bottom = 2, left = 17, top = 17, right = 2)) adegraphics/vignettes/classes.png 0000644 0001751 0000144 00002706323 12607215420 016731 0 ustar hornik users PNG IHDR :; sBIT|d pHYs B(x tEXtSoftware www.inkscape.org< IDATxwXSo#lE*Zk][VힶveVnն=po-{H|g_9uj
|L{E uJpt 8P\.3:~wn[cfP5ޱ}[_ @1=#OY?q)B1A>ꃏ#@شB1G fxV-Aƚ,V*s٬'V >wa =<[{'(cL_E왜[1V/w
=>tɕI]{b5﹣&^K%e5~Y8s# peell Ṛ
s!x&Tg_ŀ/ןxbI orceN#IpQ*<\:}zGAL0!""T`ebr9~ӧaocq(((8XHpϞHMMEzF'T*4o ݺuQ "
ZrhMExE.
yF;w=8|䟵B@&MޢZjڣm6=
璒p$$$ 59< }wׯ/ W=*͛'|=<ۇK ۪eKd¶ qAMsqqAmѵK17ѼG /. {uPՊ۶au8v8VuǨT*tSLAғ㈈J/s<;EZ}ڎeAY7,[ TzaqQY/T͖ze}5U-0`BDD$ſBR촩j\rV-[9XS6o[o`2p. %epW5 V+Xy ~q˛ԉaæ .mX۽>>>"GQ{⋯BײXHKOGZz:v yӱ$Blٶn͆Xb23{aM{}U8^PU'$`
.m 9p=}N\̖vck/z$}Jއ '$ѣb8yRp+/\/ XE&raVGfۿCΝ !aC hbZ ~7VRYY8ǟ 189EH^O=^|ɕ8k.JI0ٱs'kd8^9b 55ؾcKuKN ;
?`ɲyJĝ:SӼ1p ~-]wuI^+771nn_(,,SuZJc%VERF#cb Ab)hӦCKxsw1& G őGqo?DDDDt{)Ukxc.ٱ%&-uj=uϙ:L&~iJRN$)&oQǒs}g Me)c|ѦQ m/ٺ 9$ p;UM Lka?{95!eo?j=ifTJK*jDwsv햖ۢysL6Ws8u4&>-^q& >%+aj\bjX;?V¸q&%ͻu60 [nuت*ҋj\"4Zlc1+ ?(mM:p yXEƵVœS`%: 㒏V+vٍQcFcOTTͳx<%WlXl]B~7/rK^'hTV+vكQcFcמ:)r!% Prɵ.F&=0.ZTD)1Ƌ&\k-,X x"""""2cGnvrɶ~՚ѽ;U [q#~}+]yy9^~UgJA{_YV|\:s|J]&ص{^}'K ԚF,txY:c=g3 pbB1wkܝ1 9L8|Doyzxb}jn_}Zj-6!55˖/Ah'}vSҧ}+-7`¸q Xך:YÍ'\a3ad20yF8eǏ 6>S 0/|l6ㅗ^¦
XϱHgkPUU' ~1ޟS˯V~3F YVfDDDDD5fTzl$06G]裮(r-=sK+Jtv&5X{fEIL&3zojߺa@f٢<]U4Xg"td ye龗w뻖fS;r
upRK5=Vk mLmW)MF G Om~I/Ժ-kJrv9W)Fųh
V[:2hȾj/jWUk P74[^R=]kG6Jp}>%>ݲY Q
8,';GS*J"mwVqyhaIQn0P &DDDvDGGKԩ#\]＿qJ%4APPj`Bii J5dfflohZyqlT -7\\\+ס\/Hw#GHj3֭ ZJ#7799j/ǪUc*I*
5FӰ0*tHpO?3DG]wٍٸi^zhެ82dfeDQWN>-). ׯ|-*M_ՊgΠO5BJC՝m;c5^ÕnĹvwwGMwwwAWCii)JJK%*dRC\*h2NQQvG8eYl>GrKJ[ڵH[YUO7;Y6g]݊-r$ w5 \ܥy3zFk< ~oŧ
&ssK+ ୕Iw[,yoB݁-?n/²^UDy|@rK+c._SLbsOUB.xWDǦA/|W \Qej*%iϺ ,p1&|kcX}2ffvwIw֡`bӝ3g? QgK0I-UW,ՈzF4v!B?ɹ+r6_V疿`BDDd ܩS ƐԱ#>iX
vo/BVVG Esh~{ك Q^=|GѽviZ$O#Gp!$;'iCRlnJo#FN2~ IDAT`@Jj* ?"&6&uU[ yJ% }<3m<<<={uɄ?k"|.@fFA8t0 :$1MjC-2B@ [\
nn1~[V{cv
=wGd =v '$ WWWDnѰA5 mD7\ӧNZ
Xm۶h%8 47`\e7$!+ 6 ...z\HIp!IJ6+ztt{t놆
l۾ ~WYgO7 jtruԴ46l(ԥ Y%zЮm;#!!Ϝ5zga
<=<ѦM$ڶm*Ccq.)IM73j,n RK @!٦w2]S?VG-_]?uSF '
i?[(2,.Z2?մ)_o<5ףnob\iCڿ5u' 0np_T|חVo4
7C.{DTTŒZtimT O7q=O8lBo\ qmBv O] FEv8)q)="B:mk Wetm۸ } ,xB7FYu{ e3cK<] .>) Sap6H}[.>{rC4CvlBW`BDDdFIJ2eФIxG**7v,'LH=x`IaA̞^={ x{{kڥ+9 yyyذi6lo=#?n`DFMd$z tؾ};m ܹp"TTT_
Οn]U
B6mjvIHLĞ( 8گ ]ݺ{nxp1;7mBΣ)VY"$$is`RVV9UVx䡇͛ZؽgN^S a 0kLLrn>}0}T_ ?S+Y`s3 ml7l z
=wiS***K 5;sT^ӼyKS[¯O=Q#G
Ƹ]۶h-<5Z[m
%! 3}N|~!
O\:S|E}z^Ѹ'k4=ӧNGp]r <;%%% >ۨ{.@7~L~)<3mC%Wl
.w2FZ+aVjKKJ?a4EyjsG Ӧaِ(x͞HE)J!zqL/i^16Zi/?j\
pΪß|B^= &DDDvh$n &Юm[јxϡ^^?%˗*rsMyyz ~B8BPۇh _.U xqI1fh$SOWd2Ѥ˗9jE]DyUuNSO<[.N{J3R5ig`ZеDڴ8{ݪkڎ{3kN4&f;tDpÓ ƈ>uk-F
5BDDDDw&ނ VPge?]c?nz?q!bCϩBھ n3
$^E%z~BVqǚJ*EW.
LWU[ש3}IwhF
jWUg],.}}&>*ekθ7Np=?ok^u[9-rpQPU%6hr ͆|gd z0 ZoIqYV fw288E[լ[ G !CкU+DQh#xx>I 8X
j||< }v@&Mj| гGTҒr9!))nbAtt4ׂKLǪU t [@FVVaT.L&4kuQIOS'UX?mT O7i= ߊ &'- sm Le2[PSs5Y~bkGmNo4>]ybUBxgl$e(ukx%JYR!q"Vݩlj"blPTV*Lay
$`_B8eswQ^ V
{UG`BDDdh@zAff&vGq)$$HeZAߥsg?+8F^^ .[vcCfмysh.;#cLt)v:]Ǽ<=Ѽys4o Ա#Zj\z3gg ((( ~(yr&ܙ&t))f`ҫ[hyI
g̼Lu KbglO~IJI,)NjB tp g"j> 8x 璒$̉٣FWYYm;wꄿ
'hZ?W% x{{yh"ѩc'܁'$`OT :$ɉZR8Z&N]$ټYsDdWdPբ &75BDDDDwaA̓}]ӌVͭDgT3t:Y ԠM.JmQLگv^ $zg8c-7҄-_4(&ǪZOV)uS\BrjըC\zz:&ILEP)φ Pow"e%xzp?Yux@;dzQFQF=~Q3u\:SSI] &DDDv"33S4Xdv ~ 6m @ÒeK nRTT" ;mjb1vnJt>{KlVNSwԿyzxw^;z z!+s6hJKwXf,:x0ƌ Q>Ү۔;nxf>>o)ӯo Hpo"zc
j{0p{%gw___I#"-.@"6.q>兾}`1ޭCqϝ7sⰄ/5sf*ՔJ
rhJK-aCiUwZ۳]%`:P+^b\.-S#?cNe)pR[=Ӈ?2*$̵O0i?k߹M[duMx
`EL $=\꾌-hpr "Gm~y+hZ]r|rSW03k-i? Nyc}sA[bSqۜ1-&GDDt 8}L nŝ:N
4uuԪe+ 3)?ĸWx2B;v`31vOhRiᮻpN#?? Vqc/"@gZZDR#RRD[96!
,LZːLNJٰi# z:M..USrf.pضy{{cs9erlٺ> D|j6oƤG !~ M0Hh%^ nn զQ *ѱOeHsQE WA;!zcח17Œм >zvٍ d ghH .8VMbQhHnr-I\.Ċ P
{vޅ7gvj&e2ULn'=' ©c&$& M7fIB ) _bRk[~}n9Ţ˄{ 0 O0I/JK0Q*$۹YvלM o! 썕FYۭqAry#>L)p ř@ IrIw(wUO092nbv男k.H -AH;BO
AG,uQ* j $@tƌ _ ={ŋEΟ? 5iRr9Ccr9jUl*wԖRO磨ȡ*v⅗^?klY3gѹS'|縐/*oނzLySL2]]]`|,uJ&FW^+2= ;tBM 1A.]k5Ń xyzMBBb)VEEEmr ~ǩTjf n*#UI\ݜڒ^y%t_|%ӝ2dKm
>#/՝g^ SZBF* R*xH>b#U..L0!""")r"GfsM-UUJ @[))Upc{D=G٬";$dw( 0f-m/p]aWR_vwI!\AjխA?($=+6qWXe߅t'S)ef^7 J0X V7I
K{lh,=*šN9%c jAU&KNtbw*&ѩSgh
hٰ|
[7heq6,bbcE\]1b 7n _ *b{.L'`1]{vN6Ip;w=C
Yغm+ ;&ZD`ſW^k7\G ~Nw#*j ;Vdg~I+;ڵm+ڞ 6n\-ۤԩ3 NRT5]j{| &R+ٺm ƍ Gm{ zZٜP~}غ} ;ZYaي܌뾶mŸ0{R>ֲw><7kfzV$V}}n\[5BDDDDrwI2E?L.۟I 7%Os? RǼRFt> ;Qu49~RNiËM4%:C' L
'SBFvi{n|>3J\#@7RR5*9eUƵ_խQWJ!%iUQkU( F`fPR~Y-B|/\
۴!퓧
i|*垳A Ki*V0U L """;\]ѩcG =&wuxGФ`Ǯ]1J{+.qVU :d
##99HIIEaM0Kҹ3 4"-= ٰZURUmhg>*i
i *
Ç
a`ِ .,22ԴTDEi tURɶ ̙tCurss[ ]{%V/^Dv$ŦfI ދ|2غx/u
ݻup)I
HIIAzFrrs ɕGDEU`sNc}}}lR4
+*gjTPBΕLn\ y-EO
@RI\zWvn~IѾI`y& Ǔ|ϱL~ \`٨ީ<4U4.}=t&,лBJ\ŒFܱ냿oB1z1['ڿ6^WR).ؖVT)}Ԯ7_sCϸ%e4[>H9zb.iD/|(œztd ytݐ tOۦjFI9`BDD$` L ^}u\ Js~FZ08*9Ys CEK.UD ҨQ#4j 2 >>op,YR6 ^={ ~@.]D㋫ IDATGwћ{ P5(~5b;<=o]r~֬Y3 nI6n
իcxU ;z4%A3nhӦh /'*J4HjEK;ZF>}0yhhl6dS;.≜ee7YHH䪿n%U)eӆOpQ*֓})7V c/J9
d=)eP}
LMѩ'>:)U#,V]en..U{G {8)g=gt}~a#os5 ljho~G7iW۽Uu)ՖjEi"dفċT3.vwQ^u!mS>%/7Ls6ђrJ M}~Ȇ;?2CǿlDM8 E$ECDD$_߾8` wF 1Gc#pWP(_P8Dm;p M@RU2y]X$l'a!26}o0iD || wg@5{zxb8G=H{F'.=Nyzd4 U7y.i|2
'S
1Bc+2i}a 9|n 2YĦ?V
7ȉhW].ygkZjUlsd { Weqsp Fܸ4+-mOo_GC3lii{h4-[ʑ{ Pp*{3s|9g~I~A^ڜ&.E^ڒ2GZĦd-.\n5R"`
^_Ln9qMFZZV^}hZ),,\IKO^y͚7}Y퓐(͝#̚m̮ݻs$=p 9r2BT(4`tUEeӧeނRXXhuW?whdtU
{p7ٲޭk$~5lC˯V
"] 1[cdϾnzIMM:cb$:&FZ7J(];w U^ej&*8(Ȯ\/̓'OJBmڽKv%<"Bd2INN֯frUoJiMMJW_MXw& ^={JxX`_zFd 9zZWxVҋh!164kL>\<yyu/$Z,M4o///hf2nX_3L$:&F m۶o//9~ /ʦ0gggyk hQzCk|EDV:jU YRM&&gVai\+[OםM=]k,?ls-,-Jkqe+B}{fBϭ;?n0lƿ-\""SR?GL+n-pl8 qbLfMmMfSA.DgۛD4X(o`C뛥dJ+71=?j۠k l"/}Gn̲e;WXT$ n]݊[nl/ԱJfVuGIIlܴQ6n ͕IIri 2xp} ۿR`DTJppJܼ\IKKc6mѿ_4iD >yk r}IP`xxK~~Ʌ $9%EZ֪H~?DbH&,,L^o 9s挜MJi!\l,］D&MݮpHNNT"\TcÏ͛ٳ6cKŃo>@V|IIIZe^{={D7$n&
QHn^Il\M+
v ~}0e}vǥA
b%%%r9s%%6m\g?lݾݦ*n#>$V? h{O/?٠eB{P
?J(}lTϬ2 oM=gÞwF6zk[zw.+GT9GV왛z /yuѤ*Ϊs荦K6v8y:=/
uj>Uʁlc'& بiӦ쓏ɧ쵪~*Iɵ-))t:U˫BIBbU(*Ǝ<$ɒlɚ&M#=,)o?^" Q[5T^^. PVDxxx葵nY\wޕiOMB`@t5yV$s߽Wi>nx,[ܦ(ٲm oעdd~&DPIVoO/INNU{d!e(o/Z,? g̨Qr߽6 wfژoyo.Lfs߇rT'G^0mt+9R?
1 ݛZB^-MiovpNQX z+?+.G|RQrϠ3J|=LDDggQ>L[Ϊ/M=]uzXjwk{,Rx~]̙ɭ>rvT_ /&㇆[
z58 G4~'F\
ARɬ3=^U53g
[4[nY]k/(C\ gOϫ!Ν:ɪVHv|=z+VHV|ء[=w-wh%ϵGʂAݫ,A 珍Zu5 9w=j/y8n̢mI}Z[Z6^zOk ԩMx=qudV-[K #^wǎ#|]R)?|G
9mvT
.m1] 2o3}Z־wew ;7m!>n4u]K*S'?Ϛ0e-q~1w z{ TJB08Ƶ ] w>uڍуçU(X[BosO3?Y[_<㱑 P(tSlz\w{WL {GK<]K61]o;DNsw-"vVewiѪ0ke^U7lmמN
bV*& .ۧgnZD" :[dg&g[кj宩SJDD{_??YK6m]PSǎ+ %$$DRSk Z"%%%W]L6m?RU #~/00H4_k.s6f]9::왳?Gao"vsuMojK.˚?W¢J۷EDЦy}>P(W_Ol1F爛[V.x}K !!!2eѵU(W$4$T>c1l:jHya qww=z(i .o-Y"vkmW[HY{廕+%? Ċ%ķʛ.whY_LGEX:? rIFu )N Rڧ啴-FhҚR::r5 gGf nv!uuw'YD~?'5I/v/9 [9 2'Uu8Y =NfɎ'.t.,k7=Fd*
Z,pP
]!>nGz 8rv'd;FRTZxx0^n%bEDt5N <\vQ%r`2LflvP*e*DRi5缵M _>JRzt.cnj 'JyZNNwF[k[s$`L ? /ԩSb0 &
aCG 6'"rY89pB
xyzJd2xРz ++K~'ٻo >rk)]t.H lthxr"ᄜ>sƮ!44T
Ç
o\ ;d}j[PٽS'W]eZ<"B<U N';vʔ\pw??_ Æws+=h}{B! ڶl{BB||lYjx urrM/Gw߱:>?{V={G ~j
ӟxǥ[5>)3fͪue "*f9sFFGKl|8qΞPO " ,!1Q>]L6mdy߯LFc}eޜ6ʫJBCeIWT l>jڴg ^3gH^~I^~hn&!!!Ѭ8]cM&$$&ɓ$''G Fh\||W"oFQ%--MӥPJKK\4U+Lzl6KVvIZZKii4m*"kc4$!!A%??_
D5"ZKxxi "b0<EEUWWqj%4$X7-,un^Jnn(
//O feV4 ~z =&)-ji =v-2L-_
.֕Z Z#j繎L&$&&JjZI~A]<<^v-6orƟ=c(d z@ =&bxL(z WÆs62 zsĉkq 9 Xhh8w8 p="` (
yd rQ1LuGTJ[nP cG̍ P ;*GS'
&%RV^.&Iv IDATZh]]E&MHv]dtA e6 5mZdc/ @
& D>Q 5 "` *& "` *& "` *& "` *& "` *& "` *& "` *& "` *& "` *& "` *& R7 ؾcl۾_l@~ҿ_^ _ 6 /C ( `'W Hh PL \MK u4c~ ױ#Gvo| uLFle eeo% ""^ mL ` XE V0 UL ` XE V0 UL ` u:=|NsűsydOqCͷ;ႧhAv!ZmfhR
\*=\
sM .& j4㫭w&Vq oy{M_L&mͽZΞ7)j_C<[&g[q & W ,7֩EKoc5 @
ߝHRf9ErZ"ʚι[\#̷78 Oh TL vw2mZcճ'!u\VairWLOE \C v8Y;6F\DDJK囯Vi|H+ `Kh9~fHaahwjſIjUZ,ή:vvVi h@bBlK=z `r
ʪuTN:*WPc{_Qkl S)N ~ QnJ0/@%` &0 *9u|{lLF9|h >M<<|)AoBY9IerQSL_Β闎=|qE ZR*̍ E
_-Xk_B)ff-eނ5KKK
4 %%: OM=]u ꊀ
d2ʗ$Y"I}> , ,EEy!Ο-Ŗ6Ν d:>ֹEea:Qk45jI8] ~ .:2اl?xd@RfAR`4i4 !C߁&쨮9N\qy [tZU!woY%sNwdӑs JtEz2WD̎je:'!'i}.sl;瓐Uq Z4[DZ>_lm>̀RGo4i ժCWbklbuM+n^7x JB营8dz:e~prT딺\V:fod6Ԫ|/lT=Oαl{S}K:A,rR
4 BgP@omT<ѬrͽZj ǧZ{ٌLgQ(o/&,~FDDo4T _ɦ-np-"` @ZՆK(S%<77/qsһ1MgzYpT$D"":Hv^\\gnX 9+wݛU0d2kj w.[ǜ18 묫 9EN/i7͢XAN2ENߴ>Lv߿Zku f5%
K%)`܆g Z~C73CgWI:'lz/59p1Guns{E3u7'Vqc\""o\2}AyEڀIΠߗ[G;zgP]];&ovtD ^- &-Y{=YP'B?MүFu>aؙn}:=*5woj^Ħd= m̟ eWŦd-]q6w;h]ا J,]` "K-zNQeu'g].z58;3$[ \-L h l_Ģ}ȝw?'
+͛/>[P9MBdY^EWW*g [c-Y`LDـɦɾ~!:g{#^Zmq^f]&Rqym}
F۶s3_h5U*"WyTwZxg3nTJ1*bfP( ?3u[_DDl3_ .Q*啯d6[\5:Þk
89ݝ <\b T+YgWy]鹗K*eZU9&i;~?|j%ڨUW'ӞNݜ OU_au+_RMfX>/
& 4k?*mIO2e"Nܜ4Q _CD߷9 KNvCZi$窎llɐ,Q I y<͔ W||ͫLN\quuo&vV&yyV;V)^MݮYyydHNnWCZEl t?vɬ&h; wT?͋v߫7|q0cicUWXJv$/zܕf UΎs:M;;M{S=vpB)ffwqR7V c+)789Z}^[z*$,\oT{+ R?q=:[j #xl&h-VGDDg0*~؛ָe0WP 9ރn#"R\+.婊c/W͡^-|a~"PlXT{W:AG*{&_OvʏNwx4)IN獾@mtN3/sjbE?}BE.~?ޓiӱ'*& xddmOR6t9o͎_~=""q]phw+ 3̛=X-Ǟ?V) yݥَ[zJvvT2 J ~ܛ%69JDic*?47{X-VΟ Zxe@C#` @=+--~[MvG ΕKZ}a4e_-$>nEeSz)coWݢV99}p Td RZeѯ>x϶ ww&{ߔ~mjklHEe:AZx7~d!E'=G2~9pUC_C
K{Vnr Z1}3Z Z|j~=p˭qfpGcQCѼi}Zx,t+wOTKDθvؽyU'Mi[Vo&m^z{E x`4UTl9sw&W Q
^C_ާG=7T-L/] . rިyhQ`똘STًvjaz+*(͛|C#~<իe@^[ǟ;: t-+ +hV:ӳEӥ&EqZ*D9Eeκ'
VvR2=2Bє77 M~aWhBQ^T.lrT 9Eso9TD"dt[o|sW<5 {5ۤ ؛b6 m~ߐ ODWՅKj2gam[?{UߩVcXaAV~వosCy|QEYDC;hz C[1|Ę%iYX e=cUpqr8rٵgڻ { >âmȩsK*3ŲlrtMҷ5?gd'sj Mjۏ %Ξ {lude]?V^~r3dwȅ~`6K.wz͒Y+KI3o.:} xk553nRT:\]n:ng]3O[y vmq:]T: թKk=ёU1*meM=7856{ -V Ըn^͕R
])j=
Mr]2c܇ FYqtmjU{ekJj N-w:[>;WmyE-kÄ-ޯܖYPZED䶨ߎ
& #,WisvJ6iE\N^抈+]/b$- ;ʹxy[ kNB!"K %$xyCJ%#=Eӓe-Yb
m#ݺӵ[٥cWWԹ5WWw͔x8gqeXuGiߡt%~ܜtIKK$9q'iGϗϻi*-Zt "ig89UFåJJ
gɋ/}/
}{rw$,4i&Z888IfyHOӧJL_U9w)y~v'"RVV"|TJJ.VV(ҼyGiSD+}p&
m-#"")rW.wrr!C'W7%zٽF ꛇ78_fq䞚[<$5xȖؔ&
ǝ 'balQ749Y s/k5q)+ U3F?n<Y+B6.DDtmMG/n0~˷]+pI,(Sq|!M] bk>tEIF>?^-Xu[nOU
Ku+Kt
Գɯř2:#I~;ٕDj -A 0 $Xvb|D6i-lK ,lߝQek V#{ а
Qcqj)Sg?/\ў|ED.Bn4]FO\Rzz,%z;zV|jl6ˑ߿weh4R [&MmĄYִ*RbP6uP(jF:6iӶ(*O>&| {v_[V!lmEzɽ(! 4ӕF"Z_@L4W?s^ n_SȤ&eB?wַrڢ.ҟ[d2kDDfQ? 3V<5nVckYFAI zpqf}ch=\oc9Oնrѿ-!57RD$`~:6%qaN >0Þ{]'w2w],ޞ.NU zɽ~WTU *Lo h;x:=ހI] p"` @=JM=cѼC#䟯"\Ff?8;ط[a2c';Ehɼ+ӣp4VQ*8881HHH+yyݗ֙%v*l8 zqppY, jZ>}F 7="M_s*
yt<@K^FFZ\cȼ+'TdyJ&o,^/Ml^WDD{y9doѾngu
TKz#O>Zĺ┐_r)`_bWULJm?!58LD>;` IDATo4Mfhv|l0WJȹ]ɽZضR)}Z5sA|O6 v6 "4\g0L&ghRIe:ɧqnQs zD zTXhޚ_#P(ް.ЦMw;>iڎ mUy >5>fMQ2 K~SDD[Y
\J[l
\ɹNwV%|_ .i{%T*L;FJYY$'Hx okx~~ @jHfYrss̙A qtZFIfVEDS\\af\HM"YbDT(T*APm¢")/B weRR\|qLB
(Jqquj(te4888TcB!jJj˷GzLU*j5b2./ \/*B&7˜v(*ӵ<~>9SW|uQ[͞_mj4/>2 8_d s[c6Ko"|d6WּgR]r84fåN /WMJF~ɥ2ѷst϶91\߬.fC Nj_jQ3TEsՂ.N{xٱ/qTYDȲMGI/2N]Tgsfs\ `KۂԤ2`b{W[>%۽aY*9lY%w\C,Oz ]%`r1Mvo)n߮s֯ʱJ \`uDOxC^%`RT'YY7!'ҼE*[$& SdG$&IoҦu ?.Sj1Ȅ㫴t:1zEI*/̱G '~cX]ګ/Tistt,2,ګcǎ|}һex|oμl =*S*C+
R
̟7On wC:NzUs}|<7"T*ҡC{-.x%Yn]srmX]O7 bGTO>ͫޏ {bgf͒#~/t:>zT1
0v<41 ~Q9utmҷ?/'F֬Zmw Xi֬'N3gX2l*mz^&1Ţ#-.<=cI:[uV-~Gu۶*m,nܴQ}f+ %,4J۩ӧdA5}үo*mz^ xŐC䮩_zeINI<"Bf͘i/]r -}Ǣ;e-g͘)AU;$Zlw;WϞU ~YQQ}dℛ,/y[.P-,,TxqQc48.f,BaXE]f_m)#/orLFunQ>x]*Tv:_N-.?V--xPZ^qA q2 2GGvm&;euѧg&d蟒]8:%peݧC|v
nё
r5\)%Ucд^
UU3Wy̯\PIuLbBjϏk8k[ T:=/uq'Ŝͼ>j2vǛx\keL AYy]N "JEGYG _{JҳHپKmqnRh4Hii%KӀp5Nj51r*mzO/Ψ_''CMZeg6XDD$,mə CRKTq+P($44TBk]-I(ЁR ݻ[UݭZjHU J ""ΖkҺukv6RYmߦMV ѕWi
4 PJ`|uQ $=\ڱ,8k*ڪ3խ.&E ˫\RXhyRW;nAAE[Q5lDD223BjW͑sWi37/WΜ=k^+h++/oQ})dh[8t֢S
kNNI#GVi32,X! 7Z6LmղbчY\_u+"" s*m5]_RrlbzG,;t6(`U:vYmItLtEKDF(
XLN冟.d "U&g22 n?;d5mwK eQl*%n@APdAvBn:?i6)ms]\ys4f~M?
LIlnu1Y
G'
xʎiû|6]:+?kȮMS Mc059s5əyILno۰Nu8 *ƪ K!b5Vr%mZmx49oۖزI%댗LZIJXiR59êg +et PeU*" {wCGqvK~KIR\V,J积c4[ֺ_Jdžf"""""""\d\[yΊf_QA@B^צmO˗`кU\{22RQBd2B w* EE;]`[SSϡ8.ֹ q# LnwAuUYMA^qAohQQ牉[DG`رmܸ1V~KǕǼs];~8 ZwwPXjKsrs8tQ)&aʤI.ͽ{0=̵Zl69~iBpQ<幯̚WfviQ0z(migS w 7^7^w*̃}Uο`4mǴb4S&M3O>e٢jѣm!jmyǕ)͛UN})
jcV"4E ހNg ǎØN+ : ÷ \6MëecC:%k
橢(
" X BzTvnto S s6mbu^=&\u[ߧ
]g^0wmYiֽN]}-'3Hl[5Y;m zV
Q8eh &*aXW.\Pf: [(3c{< C; TUX3什]n*d0YFjd?Y:&J5
Fticn5,0!""""""W_6luz^h7z+4˽\
wZ-Hz t3ؽW >7Ү^EjζILfnOx{܈(
LA@@@(rsӹ~aNF?w\H:b-1n>.DtJ$zݦ rq. WRN672W&!,4̥ Aݪ\;.pyߥJa۪\Zqbcc fc{RmjڗGw;\L* 3q QVz#%R
>,2nݺ.J ,~bvndj䬢Nj]U i H1S';7z%m5uhCTफ+ fGnzcί3ɥovt9ұ ߿@!+2
UFɥu
?x<(USʤv|;\h]ifKk=,0!""""""@D,^z su, 9~
v/UVX{6T8zvqVT; Av3e9."RPκ8 o;tlqN*{ f݁ڷ`L_hحYTNx%A|OBc}U ?)l{ad2[ԯ3F/\lKos10qw֙b*UstU:`K.s6gn_E S/r ztKQ¤@ch>~ [EEVEE|\7.͗띅V ,
TTʼnܲ@RԼ-7-.I%š~އG%@߄'|
{Ot6-ocamUtͫ%>^f_Bآ$v% `jT !XWB%]c+@ uݢ+ y]Q2LIm~u |zb->;bx(Vik("r"""""""' &N/z艃[\ Mu?Rr* J-d2`8sܥ?BC ǟ|?deV)rH[_ l?D-;[hڬ=|UP rg
7 bm?_VIq8x[|TAO{+dZKh $bծ3WĴFsC(U*nҵy.]`V3X-\xU]ϫk&gu w=ghZ Gr6Wk0ٵ Ry]r'b"]wTQG&g l8y"t.=lP c+/JCda%;[,0!""""""rby8umq&osfZﭰ{֙\cCTDDDDDDDŶ?] JccC?n%0T Ej9 Ha^aa^M>Ε /
w?w)'g7n
:FՙMo1 &. Ć>CkM 6jNnҰo Z\2n_G/U{BFԜcA:?T"P:3c Zl\_n?CgVܙ年J9u% bU/~cArvQoζL
y#R߃
juykbB)=6-!>Ңzϛr>Ԕ,t,9p䨷=g4[\ns`N<ݨ3XÊ{ mb?zwDg YŶNͱû4
ܟUmkxJ%c8ˆthkcz&_}b]W JD">'jӓw9%UJb+7 _DZcB·{z+dvW.<һ?CgKT10 8u
w[ Am0E)uтn_sߒwLȒ`nfE_LWyɯz{s&_DoDZ.t_SdI|TIDIVQU8R-WcBSQQp """"""" 'O|nhiUyQa.^
mZ;o85b1d2ŪuSߵs=/<_@/ )o?fw:יBqW!w`ᛯP6ItL&~^Sr}nQQTl*>P'5xhfw֘ସ$&oӇ/DٵY#NUȤeo̍2 4}-u+]\R 虁Kj7[G~JvЎsR[WU핚[<$HYA7 Ui;wAߗ[EK7gh^S;..Q)^ cBnsL|vPI
4mV,fjdj{ku Su~Y&8TbM\˕!YE^KA0h zyd+
$ۡ
Q}""""""}Xz'fͰ`7tLTIW8t+9w Ү^(geNJĜcOPŚ|x>*V IDAT?99Wt,Ō?X˗5 7i͜o ܾc_VxaMFz2V~TE˖] 6~DZ]iU
ڍ ڂ܌J?{
W$Tj.Tj߭ ʆ*2}f} z_ٵ٬gRa=[ 7U6W"1!_nsߴ&ȯ2IHtg}vW6W"v>}=fi#>*-TR\\ En t;%O7S_ۆI$W\%fG>lMWnTʣW)& q\m:w̌! ^ NJs"
nCDDDDDDCo~~~xw; C"1a2iX,/r% ӧEqOa h{Ǧqr(*ʃuz[nnCL.@vAh_ Μ9eK{~~N;wK? gt y*Jh|_#yGwoN'8Urwfew| /@
p}Ji7j1i~hߡ6n6hފA)矽]; rAR1[l#<< {,/"""rhV)n! O~יI-4zSh֛F5@* 2DR/M?n`;[ W)3sM]㞳tGZd XJ/4G!npxЎ[Ćڊ y !~p]Gz74<ȕ< 0hؾ ! t /;3 uFsb K%
yv hkH{ToJ.ڶhmbY tFfO"\뭐G zvp] 1!#6='+H*-8*sm[-<6}7i7iy+7 o49kF g0[fL*Q+ ,[ճeĉC;V6 Odz]I%\Z >n
v'PhYEQIrYnҐq۞ IX\3h Zѹ=wG9t v ~JHrv^`KSy=' ;լLЃͽ9D"ﾇtG+Wb:"0v3>$6m{ $4~AR8 EHOo >P^tXiƗS%oǁD"`;b*\7obǷmw;rsm^qgСS?Gl2"##{vݻ~uqװ'fWxwc;[%tÝ @dd# ɧoF"}FكʼI332&=] tv6a _
m=}>
/r3̌9} tt^zS/*اV
n9~*sZ^w:WE$'*s1дY;DD4U"7'
GlkP\|x ?>"MZcQG^t~禯:}} ?_Ǽ{!rڶA ş|c?}*OU=v&m0vL2Ԥ
w"i4زlA9ughe`¤vM>BqZ] Fqj+5"""""""I< '`NsǀSO#bu0i"|P{,_ť)]Ll<_F.< A"'8tL~A.rk]Mm#Tm=IZa(̙-:vWsA*a+_o{!|$""""""n9?Z56iC= Hs ފGvر v۫8-:gѩsGBaM WVL VܮbG4DB\c6p`[/sZ_rX+оC 5-[vq)yUtr'~Y!N: (Jtt~tАh8 )^v$դI̛GoŶ?j˝ =zޅZ;O*;䫨BaM|
-s
tM
u'{*h4N5l={ݍ J ;*q+܍hXϽ }GsMNJm^ Z!'' +FDDDDDDts8rx)f /~4jX \~QD;~At@~~b1C
JFsL2ғ HraPM.^8<&"0( QQqsłԳ(,AQQ 88AA
3t(ZrYhpFlvX ͂ } ;iQm-K'y:
"""" LӦيK$ \ȥ⒚ "7DF6CQѶ5v"J4 HpS'""""""DjdSc=gѷw fEDDDDDDDDT?n XcǏyƃ ,0!"""""ުV5kl&M`Y ,0!"""""cƢE.*ʃY/,0!"""""VvN6&O
o.\F<Q"""""")L&L:9ٶسǣ_߾ ̊~b ݔXG 7ރ_,0!"""""Ok`OmqqXY_,0!"""""ʱ1J»߁׃Yo,0!"""""FNN&O ƍ{83""""""""&DDDDDDtS0͘}z0#"""""""" L[Zjm ר\ x0+"""""""" L^;?x}XRa`VDDDDDDDDD7QISh4 Ai Ό"tDDDDDDDUa61e4dffbc~_k9ܱI֓ a Ko.~ }Vs8}*O%9<&DDDDDDT[ }mܰaCp$ VjTDDDDDDDDD """"""WN&N`-:5mZ+ 6t(ڵm[+""r&6&)-&DDDDDDT/X,L>
O>AZ95
x{b 8x6ճ'&Mn ,0!""""":ooocccͷ m-Qm'qDDDDDDT:}sͳbr{0+""""""""[ L/()a0ly|"""""""""$Պӧ#=={?A=ѭ&DDDDDDT'-^m=z`I ̈""""""s~۰_}m Ӡ%""""""""~2GDDDDDDuʙgksmcR˖# sIX`BDDDDDDuFAA&N ` ۸At46m%"""""""" dNnmgϞ+α^^X|9=s?Z VtDt {1x`OADDDDDD bkE-=U
4_P""""""Պ/+Wb? 6̃YF!"-pp3k""""""wV={ݺv)S= Lm|qTT.^ TӯWNnbwg4X`BDDDDDD=gm˖#(0ЃYUT&\4&&;""""""E՚"L2:;gZ`VDDDDDDDDDTQZxaKHMM }{""""""""""WjŊîݻm]=&DDDDDDTڼ|m o/T*`VDDDDDDDDD*QJp3_m{yyae`VDDDDDDDDDQ)..Ƅɓjm9fu ̊""""""V/</_ ~!9҃YQUj{ ;wƝ;uK_`FDDDDDDDDDTU,0!"""""j-X駶qDx8-^LX`BDDDDDDŋ9{6DQ (
l9BBB<U/ #"""""jSVcIh4+fm6 ̊*9Hbឥc Wx*'"" QE/2SRlnj=F{0+""""""ǒ:2t쿃Rh=y LZؾcmܱCG fDDDDDDTwU{69JyZ\oD9!~G̹'$""*""""""ao݊V~la% ,VQx-CU뛫(VlzV~DDD'}DDDDDDtC.]g͂( \K!44Ù jQX`BDDDDDDU֨1a$5j[l3Ѿ];fE7B- Ey4Pz
@@`(}= y LJDQČ3q)9s{=-77 Glljω=P MѼE' \yEQDFF2.^8 N ;+ 5j \A0*Bņ۟%y9""""""?[m;o3fx.!rK^n~Z۷R|Qqj^MֿW7n #8Ͽ2Sqq\p/ @S;_u:D|N)ņ4=[${+d֒9g~ Ps"""O` m࣏l0,_r܃Y+DQ5 `d2T8jum #Gϔb`|.""""Ra<QUܒ3_mg"˱|y83d=+wNLL36G@@(ӪH: nŪ "uwl٥G婽TJ%\%@ T=GU{ņUr ((|rRQoZۤ婽B
T|6',0!""""""i4L< [̜1 :t`V
ɀ_{ v-8$F]pNcGw`u8x`
$Hݠ 7ncG(IDDDDt8 /, QT"hRiB&)g68P\" 3G>ܫcx{ w wœXR 9ͥK$R4C&ѸI[4iZ 0}G Ms 8c>GoBT2@w4yN[i,@x-w>=(BZ6(&sdpb-'R zġ O{* &vpSZCkgν@kh[<#8rBf-&_c \؉9G-0>Ԥd%-$@U' h(x1~;4G75sT0- 7YeS' 듐N%Vn>>\ZC%cTRuGȅQVQt/U$&5Pk*ڞL}|՞sO ev+y L% mܮm[̞9˃m [^ex* qxןU8W*⋯CPV\DDDDD5m+!/~}hnXz9:OjCf֕z%-WK[2]c53:Y;%e}U*5m݁W-Vѻi xaD#kgL&&
yOGWg4K5
5l9?-4-!nNHW /fyv͗<&DDDDDDT ;w?CCCβ ̊\i O.H*rQtS䲅y_뺗M >^}UJydQ 9Zo4G[E^Q=u%7׃f..K%*dXx <bw *g7XmǪ\I?' IA]LHG $4mDκZ̼ NmE? ,>>~pMFF
zMEO%MGA~6F=Lρ4hܡ+ˮbB8 O% EZ- EppZwC=RT9G""ٹFqahw$6j˜ wV1+
Tm<ʻ:5,olZwӮ3Wk|U3~1C1\v_o4CwC3g~| (>qyje{c>lqD ͣ~zVkVl]rvrRFl8|q* .M?< L*Qe =jǜ <r2ozs{Φ
.=S*=ѯu<%g-.} IDATȮn UvMrvOpNJ ~%9 sZ;];rxx{bѰXw w}n^"32O b2=tmhk/ q /cWJe4Q{ jCZ^,M* SX㈀C:4N|gTOX\]z L/lE;}d2ykV >%q9, Rr/4lMBt0>bŅd\&/ כ,Q:$-pogwbKU|GjQz1U\tڝweT"û4=^5[pBкt,>*hsC+.gĉwu\iûlH}*6 _њN=7|v l9b&Gqw'L*),̖vvDY" ڽm ˖APTg_7o?YYxqh
AA}m "n3HOOƾcU(,̩o@MHJ: z>}AP(!Ve"q Lj
E1:}.+\wmOv1\#ǡo{ p*Ղk:|A'%VDDD7˙Lf/JQyj}<.`FT*] hw$._uJ)1dwp%N5+~D
@)kƨnG?Eg4Tov<}cAeN7 nikRm \;u%w0_*[3
tVcZD8SV:֡Qovi(8@ZֲDjN{ &'줦/j9T*ŲKQJKDQDcv1* qI X,X9
w`G K qq xiݘ8 7/_g_6m{ں!!QzhРk.38۷( m8
D\r >7~]]UB7t`2#bm% H'.Ť)+TnW_,p+_""[EyL1,E# `ě<ݨ6rsiy_yX}ߺLVzw '&wOeUM h߄:'4=RյZ7+
)[|T\e~wnpB&yiTf67:V)/ ghT8D,0!""""""N 'ҙ kN
`n x(c
t]kA<]xyyʞT*C^wIӶOR)f ͛w|q8;[t$H1KE|Ю%8$_գ0yp]lMw9g""[I"q\fvmbҐl5Ʃ+vE
4{dfUz!0?`3zːjά9|RmӿmR93 &/g_JqaR{JX˛q"DEQ]8@g4,`d.Fʘ+:Ic07- 7"Po ~-VMDu Lٻ(5 B
$BｈtAJQD, "@HP)6HWK tB ֹ C6 I&<;a 3 8 kԨ&K~:
-6Gؓ 2u N%*|2 tc[e`r>|1pV49~p{
ժ7FZW.WΟc bӯMVőGqQ֧W,LVw5ծU+۷a5 ,LΜ;fշmقU bqFsgϘi`h0b>zcd[0F[w*O0 >|4;{LףAMd97Do&Lj5hnv7p"|L?JV d͘=۾sw#00OWq#}?&~oTOo)kne2k++;m_[byF?7nBʕ
j7o_A)&Ksj5n=?艱c?ss2 5k?͝;>7|OOkn,] 2sd2erY*6ݾs#Oj5U6@e,xƹq)f 2 A =m{ f5#Y6ls|p`rZObZ 2L.O ARǥ 5dK)^vGk+uaPW|= 8ko=̏M\Jz7CcXsztg۳7=M"Y#4:AwBUj$YhLQwq+}}]}'>eΎg~¸O
qէY2Ȳ{ :e"X` 3X\ 8;c% Xd N< *ryގ=/XX37n5n[$Et_Jaov].wy^xG F
&7E&V\ W/ ^(?,Q` Q oǏ
,ZOOשK$Fǿ>*[ز3~O!--(;
?1Q.. )>>}#6ꀳgG`y5GR[u>}8Z|6xej0鴈g
:_QRArJ^
e֮m[kk:CVjԨ5j4GJBa\ &ҹъiP>ԯo\o//L<Ťfff3x 6 郏1inMѬiSV(_ +4 \#;>3imZ6inEߊu&577ru̝=Ç
3 N&ͭR
ٵ;x~5?2y1c\%$+= =իUÉ#Gs+dbJv6sOS4 4Nڸu /\O'[uL&> :q,Y?j;Otr^Jx m,oUnCAde`CDE
&DDDDDD%Xzz:GbOS -54^&$JNI\xF-+vv[%g12JH6F՚Ze`rDE={, wZz?3 :xm[c5|A_/V:\ LAx:vJ: , m,'n:5UKl-̓˸]hqa;uAJ.vVک]p .>pݺwOm bJUi|xh=VK8 ff
Vf4㥮
/,Jv~it,W *jvoxxГ/V q;2λc=Nig~
Gbe0YŴ>/d7\l0!"""""*&M{]~&*>"0롸wXVL&P0qIr XjjX] ll#ȋZ[ k 9 da[Ԡ:ORh u>es'"""ZH -JIdS6rYj){~/~Bs?kT3QE G')W~dK#̸^g=ܬr ̣#3TYGG
&Q|= f
yVZ$W'ϻOD$橑IleF#uWy
Iv+, d;咔j [U/ה;XFe JVnmaGJ7$$"""""7/_nRZ5LbRo ' Kʖ-={B.=~67AP%JzPfF_xǶgX=бTTes~ߪW.+QIuQ -6D2_e㧭j<5M=m{ޔ 5?
ygg>2=$;rI8
SPƱR(2ϫ!UUox:8NLSXIL/:cAgzUq SչqO
qoD^=z2X kܘ~-$>yo_D-E !/_'zP̘6
E b IDATLb#pAWw FRR\\Ը< (>M͚w1FJs{qdd(dp) @Ռ qQF;ztBV\T頋>s~H_$3݃] v7$?OXd/Լ1RakW9 wc)YqJHUMRe'Y{6Lvi' ,.ŤTVZ9ǥgC Ff v;hM J|ϔ ECѮAa=#X[m/C5ʺnXKh ZspW\[`Vۘ D$-nCDDDDDTBw_q\JU̘6]DS˕
N aӽ`ܺy֠/3PVsV0N{q^QX:(k9`T":IAFz%66h >[ ?
Z|Ƕ4qzV&3b
(=Qv/IvY=bw*?Ve/:j6 E')lߵd2BsNo'pW
72ច?ØB r9_ 2CVE*""|As ~]j.ya
W>{fx'>/£G7i12 z20'''-& R^UZP- ıu 1O
j3Onպ`OSDDDDkFܲk.p9Σ}QX3Ue2M`<378X[ܞal,Ӌ.K3>lq~Xz#m,Bz\`]\ juf
yv5X= յNu}{dv;k*/j:0 rYZVsA0SkuZ!L[!Q"""""7XJOHkѬi zݻ >kʕæ7f&+W.
?{ZDͣX00Z$'v
j'ԪVmOy:9>[ZRb,fN!!A !rhUQpTXwH= 5gy;=M84T_Ge- k3ƃ\.w+ ս]7; TbPZMp@ l-̓M9\W! usf.i>8r)~N[>ӱnYխ-rR)/~ i:"6#^fqw Qёݺ~}Ĉ
ĄqNp I VfMZ\s=~ͺuXӏ@Cfz\q c*AFPvKxyWFBlloƕˇq3'fUf>8|pS_xI u꼕n=w-f2;~lV0Πns< "" qݡTԽχP~kP2!!MunY-7:`࠙hа- ,N®k=ot)[AK0v mԾ tA xt18c"22#icPϻCuOÁ} jffhӶ7:Y ^aqQ; >o6_PRNukLd/a""bN:FITk Tk\r 3Ey|wG֭pe58![kTk,R l,b*9蒝Eh ֏IVz4N<":Kؠv ~ u˙M#R*5:\qr{2[óNkbdٻGo>^*MuTku
LmaHs.WIjޏ:.UK^cY.EDDDDDD
74T\3ϐ0ݻ'OFln% JdEkA(]+Bz3 <K 新sBzz
kee{{ DFX;Xj58x` x
RSaVK;F-_w{]u*DuSa9ܯKN(ۦF=u7YfU4&u"*\"
tE,\H ;::bEH/qM0h.ݫV,+y_
c hа}4#sldhպ'V W5{HV>s \]K=R)q%\8 GHHPK Ǭ
:ӧ
v?B~c|ҺMl kڬ
9yP(?z>d<
IIqu ř~t%DDDDDDD
&DDDDDDo[q
\EKdEOcXd Ç
ôS_*a5w;4lV/9qppAn_bq NNnTzWŒeqqu48Wu :Ou̘h\{{g A_͆LV|MueM$rh>ɋ=cɲChӶw&9qptE aM]e!""""""zn]!>^*
? hQJ'bMP`xu5aO?B A o_P%Vvܸ~
QQHNN@rR<)=W~룒_ (yu711O XgO D(\QA` kjbΝRŊ3s^if#0˒dCDDDDDDDDT`BDDDDDr*-/"`lll$LUtT*MCG5333̘6 =u0ћ
&DDDDDDȨ(
V X0w ʕ+'qa#F굫by&#""""""""zs5V?zbccڰ_-%LUt?C ,J*5?BU%LFDDDDDDDDfK L7{ ۶WIܹ{ ߠ6""""""""B"""""ğmű/̚L&aqY|k~W2&#""""""""*`BDDDDDvΟ/"`lmm%LU4v7~3 b}vXn %LFDDDDDDDDTrI r
1ch 2 E%NV~\/7#L ?r9?3ADDDDDDDDTT`BDDDDDTi4 m5̰CV*|z lk2 F$LFDDDDDDDDT2=o.].۶i!0QKW0v8 9zTaڥɈJ.6S[m_[c ̝=2LT+!!F 7hòKѴI ll0!"""""*߸Ysc;[;X ;[; SO!yXaPJ eo_+oR 7XjJ ؉Qh4 LysGd'm 5bccŚ/֭ZҥKK,g:Qc Q1j?f ڐ_M҅*dϜHKKk
e a ;a Q12w<\vU z
*aµs.L>
:Nӡ͞KKK %DDDDDDDDTn]!H m;cʴiBߦͰ.T!Zn VҠ0~XrRL45BB!u ""**?d1 `BDDDDDT,Ysc[[[XF6z̚=[%d2ƌ/>\`&DDDDDDDDTRHb9 jf9fWd/== cƍűŚ9̜+a2""""""""" L$j11k_
mIp'$oF`X 4nXdDDDDDDDDD6Ih|8~eK
GXX ǡb kVFe?? )R """""*vڅ8._ ̝U֭[J+bӆ?\BDDDDDDDDx^$"""""zMܺu 3f666X *x'Og ..N5j;<==%LFDDDDDDDDDy""""""GAR93gRŊ*xwáT*Zw:b5o\#
ћL DDDDDD%N"""B
%:o/ajjzA1n̷d"""""""""b QZh.^$j#FDKcغ}Xd7[|ڿɈUoi8.[,ΛؽTTbر8~Xٳ靎&#""""""""W""""""p+(gX *8q:ln)"`5l(a2"""""""""*l0!"""""*dq9zT*X=c&*U0U
ேɓ'bkWF%LFDDDDDDDDD
&DDDDDDHaرk @wޑ0U y_3 qb ֬Z
ww 4(JcQ fmm
cQ """""Bhb\xA 7o GH ?ybqX {;; ~QDT
r ,u """"""*`BDDDDDTHٳ8Ƣ!%LU0nۆfBﾋ93g\dED 3Q"Hl0!"""""*ıV,
t
Uz}1FA&+~
vQQ"""""HW3Pre S:Ni3f` bM.cر~&7 \ADo8DEI
&DDDDDDIc̸xL}ܩ^RĨoSb; :HLZ}0-~_D;yb֭^DDDDDDDl0!"""""*HK.ŹqM1_D...C[n5\
ח06ů&ʔ %Lj b! {=Cҥv*T(a2"""""""""*JDDDDDDݻw1yTqleiprr0ի~ĠJ*ش60l0!"""""zE1 bmƴZ^c0`ЗOHkM4w77 `BDDDDD
tz= ?aO.;KlFfjڥ ֮Z
;[; T̤@DDDDD: X3gϊ&c&z5W 4
0 FL&(I
&DDDDDDw>L2XpoHNӦbbM.qq߾&#""""""""
&DDDDDDp}L:U [YZbҥpvr0Uh>sFYZZbyh߶Ɉ` Q %%%aH(J6mTTVMT Cb?,_zuI6^Ƿ!,,Lݺt0UૡC3U ֮Z
Qqm NDDDDD$e+Vl%Өa# O
KU6p"""""" 8x?^ {zzbɢEP(ʻG`܄HWZfx lmm%LF$O^?ܼ$LTrDE>]kűL&A3$LDDDDD ʽ}R ߪut{I)''ouFw[fma*SI`'n{1v8F)3Qc ?x'cKKK,_ gg S?Ĝys|}Gn1m:+"!>
7n) IDATлHII@RRl0)" 18tp8dl0!""#%]uwZkE`;+ T Qdb4UcG` Ŷ$4&qtR5Z
&E~D|_;+ `B᫇DDDDDDHNNƈQP*bmɨQ.`r'/#$JDyTb_p?{r K᭖=ЪMO+W%Oxz x(4b WؼLDDDDD%Ro=wUBc&++k*"ʥ{ +EN""(z c'GhhXc]Tyj1yTo1O )Z
ۀ W!99>'&`?W5=ʔݻ¹}x64u~Wa2q芀%LDDDD? 19] VR!*.o`h/ƥ쭯 {(gl0!""""" VSa?v&55cFٳb /@6m$LFHNCpΥ\d2㜫WSf4k%yW. ƃtZ%cssK adB[sBl_l"*`BDDDDDCű.Z B!a*EDcp @:u$LFx! 'FYY٠v]%*U'RpptNCbB4p%\|w\^rvNgEy+D9sDFűL&0
凗]cgR!""""",<|M$aaaKEd C,9^^Xz
*//a22ERR f菸H:t0 NF
rB)7/T\ݺBBB4vXC7CQ9W ߚ 8tp ؔ#ʨL_|;n1L&ZYvW۽w6و^Ŵ>/ u"*\l0!"""""$9%GDZZX$ԬQSTz*DVzu^\]]%LF=V,7j.wƄ?by:>b2MGr=yhCրw%冫|9TU˸uwZ^f]쬴 ?{/Tوe Oǡb ޓ0 >q}JohFdd];"i ڂҥ}}^ ۶L.q;:DDDDDdh@DDT`BDDDDDUp q\^=L7^D۴y3.^/3NBH*.T4'fRs 2 z@zzZ Je
KKkxff~mVgHJCjJ"llT
nn^P(
FƄCKڶR!YRSV+ac {{';2EDDDTԮDٟL|jZ3sNvOR9kB.Θ_IJC#qЭ8|h3.FmpBajk/Q^| GG_ϗ(_*ji:uZr07Ϟ=¦? ?fjyׯ 1%aтrߠa;i ןxXJu_x|N ^9*]й> 5Wjpg)MUgvd2vRW)sipw-M5WyAψ:ƕͪo<_bא&i OlLߩ7j{\F7r+~,.Ƌq5o+ : ljV FtMIoLbUe){}W8}UU{-SGvd2"U.j_^Ni[4>yicg;0u=Oh ) KJS̜Un\YkzvyK?o/S˿{84 = ћxJJDDDDDTDݤ!_+@(Urh0igXS$Kd {bTK nD\p 9*';nѯYM kJe*._:˗c1.g5= ƥF%6_dEf <&M7# ןؽk ZIC ;עql秤$WvM7O
eѨ}77oK+k'ѡc?:xt:eZn{KDDxhOF>K1[4UPUИn?V9+1MeŪ 얙$+{5k9_lͧ[wJuhuz fo;cvUk$ViFwznÉv%~SF%x1v|
F8pbblQvudb+^myw?K`|ARURUCc#lnόd ,.Ryitz O9
L4Wf{r `zgq}fR'r8DzDDDDDT⥤`HￛڵjI*w))2k+++,XPTDF||kR:%z$''`~96d_1o |_WlXksIf \ǔIt|Ѩ0sz K2;n/Ѫ}&5d%NԼE7:cFǶ\W. 6h.dhߡ+g$""$6ӕ{\ 6$V\?45gLל TDP.\^/<ނ]^!IVnu-$jpn%E%#+ -5tݡ95d"4)C>aSK"B{iVewƱR-;o煺9GnvF3EbFƚ\.Kt=]ᆋUm|j2ߣgN}ˋ5:AsBP`uokmST#|>됒 =0%?HHlϢ$zR] v
t J, R12!Js11]-3wռ\UvME htz|rL`cqagj2<>/*)^-mnaߵ3:սٟ˔ =]< mR5Vnq)J$-"!qNG/+
\
Ii24WDz؟^.!/_j ~->U%6 u{%R?H蜱Vau`l0!""""m՚58v8[.&N a=|_}5Zٲenj+WNd*ZłaA_] w-QFm!]=G#%%vʋ65r_LAFh z;ƶˡ=_9ERbYH̞#Oܹ}Q\UC-0hxA
lۺ`5m-G6=!e [{zǔi)CnPϫ\>}5ua 7(*N=pԽnZ.gֽ]{oq*{oD (w]zyy@y˖ z嘠L1z-Z4/˨S[k3ܠ/0"t^$jP ĜUy@x\1{K뿎h3urSԢ{T9}?<6: \pkoi* m\v=oj
LBc^yjS5NkL~nB)#'1$EkeDDDDDD_=u
֬ xHo_G%+VĖͿ y.].qq2AWyLI$ WqIxX06oo]2,:?b][?sgpn%* ,. HFmq"#_? Wo_Wq x~'a _7nn%Qѷ v6q؟珍#j~DDDn6Vr*&I 9䭫>[%+w_pZN5ZD"NЧ $bεo uo#"dTֱD0}or(. m8o RI J$A Gg 놴اB;s~ q<-;ao
R.Įst^5 te !cS{U\ %SOw u z}K_ Ծypb9h*LVD9V&hs(CֱO3lT L,0ƠT* drvQ <X
농7bf D]N܁4X$a07TEk݃6!##]30|fz_'hHYgyH1pO03xm\9'Sd+ٹt-[/]<>T*p߂X= a0.Kf QBJz{/?~g-Z1i/'[\z b 2aIm?,LS\R}jr+kˆ
*<
*xF{Hcww:6i8_훖~[UIYr$tٗMmԚA-[J gyM$nLfz.~ ^aq纽m}K
s크/)Uj
\R+kViWê>",0!""""NRRFĤ7 ZnG ` =
n]b036l 1[.N~zf({Zw33KLMMƿ'=>wR]O jAA:Kv_"KK[i#q~t @,[aPfsۿy=## Oxo\rqqH$FzDDDawd)2P(U֯bGvԟueE/=xBanR EP+EDt@} uؘs?3n?
Mn5VngR.9+n|,'MJйVeDկؤOS
zܠZw{˴wFT$k,0"tNjagHTd6OhIIWx}˿w!KD ;{*X"*"""""j0y={ InУ 0j,sR;lPL6
"xc> 8>va
4snNr3ԭΠs$A̯A 6,&"4jً͠Q"_0G Ǐn^A'HЬE/AQ ;U0\8;-QADWQ|JzyQ o▝f8ͯ<
\sC)gԪGo?/--
R->1TPf/>u Q1CѷO,-mt9~'H_DmZ
h]_,ed*Ɠe=_,!kZȥ)9-d u Rm[F-*{E8٘; !S%aqR.mΑ` }T֬['NhU*W &1eddæͿjbӧMC`TwU[1{c789y. xx8\XwOU0NJJqysm
f>zB ܽ#x-ZΓS"th(*W1O
;O}W Y0t7iuP&o=;?|ϝJA@e/'&T+i9hc${/ A9Jtwkd&*3Ip 8Z" 2/,XTYOzf[/T,'8,. ^uk;.MW( cMaqi,_R3vtt⅋ Jrq$$&bԘ1|&fffaF̌-B5ok/_Bo}P~ mխ ?Z ]]0] lLMMakݺĦ_kXfmUJՆT \r5kĥ5cG_ ?M} {{T$O$""/ kcC4l8^sZ^?8pT
':ۘ_pO
W
$4*h?(2A.9]Q|G<w[Jns|W%ZTzluU/:s)J:dNLTMۘ
<=zOItK# ּ|]|g sq\* Z &Tjp1TB2in^ G%h.&)] an}9^]ȞޫO_+<@7-5s.>t9q2ڱwH^JT>
σkn`pv*x 1dP>j*
j
ǎ CHh~?/Ǿ IDATcGgF H$H=vCŊs
fΚu:~~hղV|UH,H$X$WW/{ĤR)Ə 5k8~DA,C$Db1 vՋ/Zhߦ-ʖvW*X|9}
DyWRԯu-["&6VpwGN[t =v O>\S$T.G}¥Z ݺFx*44 x
6D /agtl/V(_ jԊ!Y,34n5ŋ
t݇L*CΝ>zn<"15k+Kᇮ###q9kԬY }
\S$dɒP5N9$6g][k[*DxI&5Af͵ ^e[ͅbbcq}}},LTE煇;J,~1T~
4) W57,,L{
|+TKLJӧZsTRݻZs 1s3s/^\kW4-SFkn||<➞Z詩 Ѿ ALR!$4Tk5 BN"zsEETB
_~vB)2-C<"&Avʸ]Ѻ KSX9KBjzBur W]G"e&̛y)CAk o xx˗dt+ >MHX>˳RSSVRp<
Y`i ݫ˯
L"##knk*0IOO_֚+t 8xZ5k,09s >$,0w>6VKNZ&!!ذqka
ZsdVOW#G k7n`ђ%ZfMj݈~iutr*0IKKÜys[gut>>]&;vt\tIg U*W*0y Nњ;o\L2Yknu̝?OU` ||
LN:۽Gɓ5vysmւXrr2oSO۴Aa5: s >zTw}|WZsu>$|ܜ 1~| -Ww _RRZi57 _%_[К;旓Bb$bQynO\LECݬǙ Mt>_F̽Xg>Ƕq/ro1ՃW1?ܯ^OBcDw{Sm|!xq+RZ&ٺr&Rr0;/-eTeDw0Vl6hʟF$5) ǡ^Cc?vA6*^+Ƶqﲃ_j9 f(yueԽv H&uJ,0!"""""MVɓ4&ֵsS#fەW1rh$$P+кqF#5-
[m**o{{W; Ob[d!:]#SFF{SXIr
7S'wꕣ {T\|ׯD_[J9/(ZMԌ ۆ B$w8vTaZ&wpͳ ! ѹS'MV-ZBy-ܴb˕Ǧ
Eޥbr kVҊ{kw ~AAM&RoY6iH..]Kyc ժTъd2sK(31G#1AXsg~U|u0uvk{y9R oՊ* ~8a7%W _c h-g{dkTGkb 6DLie2ιU*W֙={!:&juw Zh%S&M!|Ιu?JՀkj*}g[+9'зwׄjs*nӺ
"# /QK:ukwU.}?w)tQ+niestpD 4TvH*M+
|tѠA+[NsMJҥooEn'ffQ+&k뢾g ӊ;Y511Az{>Z
x=[Igxb:o;X=OPV!"Xkic1kvY&Y %(9-SRDBanY\k~3_4 ITB Lh7|Ƶ]ׯrO~.='rݍ1:܈
ѵ+{drv]w桯91P"
[wQ.
n]\TrJ< &3v^p)%Ŷ+j0ul!2Qn ;~L3)&1# >'#=͍F
ayZm))
صAxDDY6^-=~t뾑LVZZr3.%%I E\ܛw]GZ
~/[$ /'oxnƃWp2Q*ؽk u\`iب 1O<
=[W3'3U]9rbB(WW#\\\t[beeZ5k5W"]kbŊ9>+w7/{{{N9111{. 4oLj%DeܺCg''|CbYL~};DSwhҸ&pwE#rI''m۴A6m=@Wxq+t1 E]ѽkW֫[jURdIޙ/Tl4cY D˦{MQuY,[\3vppE!ݩ6_l7]0_pG2i2${(RTw`?VZ/^h wwsCDd$D"1 GK5!KRy C?`*҂-(̅&qzo.kX 8::PtMXkj8vd "#_ =jnJ?1R17'
:mر[&ǎmvt@j
5O""$lѾщ{A"%;Z]+jw_+ D Ӯ= jm+OCkŵW(UVY_z2t.JLM_ Q} J6Sy@m+tE]5&<.I|-M(9kW[]D ۾? On8q'+T* \ kz%a=
iYo=OI/ j5LV ݺA-s.? m GkKwJ"""""*Az _ gg#gZƜ0g<
ˑÇcڔ,.gǎ CnäIZ%>*`ֵy}13@* 1t`WÃ=4h[ 8 k%"+sX$^p\V?aZnYqaE :]py ^E'ߣDmi*}R~K߆>cԳӱ=zPEᆭεoqݶ1zy9Y Tjm 4V~U+ =v3ZdeqVRΗ&rAq A %
o&:I*>Nfw}sDŽkm׳[R6zA cgj%T9w:.2 *R *lKjJıYcGT1:?3U~qjg:$$&ɠ,nd %+xw\de*X$PչnuNHĚ7R3n3v\ (\* ҭH}KS&'qNЫgO#f$ACБÚ9V.[.;1ˍ7W_!C/8вE D *AR˗Om˂\[#55 ^ϵ' sg0[7"a5JuV߷Ѡku4BTN GJLl/!Z# o[Ƕ
5kSIDDdlsLafF\GtiWM7DvNHM/n|t2u ,{n[WεK?of5v]~ oTruP'l>U`Dfg!|sTjuɍZt=}ETgaqm˺퓙Hry\&DDDDDT߸G Ռ+V&1#PW^ iF|R3x CG G ptt &bйS'lS%b˧}ܽs22SШq7ŃxI 9+>\P~;Aѭ2pL{@"ɕ# ,,;*9S-4㰰 9Vw-zzDDD dٷJı GWqXfF~cR(HDr^YCYǩ
qw}|
&I#]&ɲ.y9~eb /fgk,])1[ri`!YrLV w(WVft. UuoSRe ܢ
%2 L8w<,[cEFG }mXJxyao38< ĸo&Gp13 @=!J%PB-5^SUذn*mtݻVa
*O/s yM}
{JkWOk[/`eK#]L221`DGJ{m{[7O nr#7s+)ʵRdf|+˕)8*A{遟cSDPژT*qDjФn}D-o8c*:?JѫĚYcVfMV\* ;÷߹̽Q8}_i
\uϻkPCЊ0C 0s~/
<"
*x5P:RܒWKި>\$AQn?}< 3*nɍkw+LkaCq6?3*7#XGGŋݷP^wՕH$X8o>\]]k\ƨ1cU\+-[Τ*V^{Bt]9/eW :
7pQA|U8sz7v :u ƥKozE꽾JDTTHSSb/ronn\ߗ;uhb \C0bBA'>"))ug=b7o/^vZ `ovW>yꕣg:4j
wwzo:kI2(YOvXgbosw=3wdi֢WmEDDWW `)vq@߫>I+ *uկ̱ ~9)
q{'$fxgv=sQuoS ᔐw^|*+Z5jUP`>-ae),.ZPZC]QR7A&vY"}(*{/{wѾk\loz[=.jnook!O{g2&JZ+Yac.3G Z?NoEk^1{#&ugnpj@ة$ tEy{zڥ]OWJĪb/=,<2!*0<˘2q)US3 ;Jۣ/ZbV]^?mei/cｈ \zsZ}8Yex[ LAVis5V.DHYi[|O]Sp4R{Į~?Q}\X`BDDDDDޔ ?;>6bFo&,Xjl?)&Mxl
I$tC;CD"W_OGbeo.|,* G7 nINd8W8J||4vZݻVxnR?`*N 1jq`/xH3ݳYPEFĉ? IDATƯYwj֢N)(2X$&j*aX,0*8DDD$e + `&3y1L@l zv%cSeTj4(2]Pd|3t &OQ*qHwT5[Ԥ+1">nc)>w#}KJ(W* ,\ڽnMq2/EJ] 2!尾6ʺec=}vDfQq& <|H3PӦN5bFT*̚3. 9
S'OfqI HNNkѲMk߸AP\"Ѿ];߳ӦL-%YlKD}abC^*Ƥ)wk\,3, \ :m0C*Wխ{_5k5mS|ML7JY!`ҔhԸ$ɻO0Pjo\_\*yjkFZ2#sLcTۉ G=5KS 3۾&ߘL^y :
}< N1F)RXZYO2en~@*N.$|n'W}|d^SVyLpKӫyq᭫ hW)#ZW=9D3v0!""""B XdflgkCnmg8r{x&&&>GtMiiinzDh oִ)F 2K8psƗ_MCCpaܺu/!--%s$,Y}qpsuT H Lfb ZӦı[dDX͚D5q~P*hffa~uZV8n .ߏ )){bŊEJQjT" ''w˰ `ai#Lnܶa.2^E~,(N&3EjШqWԨ_ 'X[V:|.>-n8? YRuu Q|
AW ҨS-gQa ~O> \%5CsL$7[-*{ ۾ͷ0Cj 4 ̂[*ФnuI 1:>n!?4o=Nf1}"A9mٓV
X$p1?ߢ:պ vזP!ԛ ?Ϸ[?.9GVX.2d= ,}R҈ k +4&sheo1YDjoi=|+pkUV Ot}M$7;380m{he5'k3~Q{M~ϫ&HNSe(U*JβY^ɤ/2Jd\mOq{B"m!""""B˗WOxE`5]˸71b(\f_ ,^ 3bfERĮݻrji_YN 9
|}smkbuP4d&44sHGXXb8drXXŽA*5n#4$11Mao
WW/X8V" bc#d03'\]>#dd!.611ልDbb d23[ nn% Miiu]$jbLA_k gNՓ "0h2hP@DDFVQ _8i|,#M9ه&'fXMR䉵]V/iPDAtq퉻% 2}xǚ!-'/$4%:!LX*qb/|< z鎋J&$Yg(UR3I,6J-& ,F`}`x}Hl}R|g\]5֧$"2 ;QQchK `qF/. aCCXb%*/o̊Ja
^*U0f(ԪY TOە%?8;3v@"ŊEe~^JptSsgdh8eDDD?J8Yp~ _X2nj wߨ3 όYƐU cT&M~gq鱠$%D&DDDDDTL?<|P3n߮ kČ bax%qɒ%f*x1bexֱrah2#DVcXaaaDDDDDT
o5Vi!*
X`BDDDDDʦͿq/! /]cUVʥ`cÛKXl)n߹uF -[A$ʕ-͉#wqzT3k#fDDDDDDJgZ;V*iC̉c /]E4c[[[,]r[[ǤSP(4Maِ1Xl).]u
CFN! EIIqظ~ Vۯ\J% """""2ܯK 5V~er"*
X`BDDDDD«W7PT Xsxlظ.Zz쉉}1 >ȣGd2{1 8vLf숨HOOEFF: >>
Ob
y1#"""""=ř%gHT*p3_T[Vk4Ewǘy,0!""""/5-
ƌAll&6nX3J>*
[51HF/4JNE ,_ 0 }>
333#eHDEM3pַyJyWʧP} ^ms_X)!r"*X`BDDDDD~ hm۴%4|8v&fbbBv팒SQUkVcמ=P*cfffۻ7b !}Tm7vDDDDDʸ(` h ˕+h\0|(ܸyCĒEM#==]pL& =0hװ7RD1A.CкM?D02JB4oc΅` X\ 4c,]˫W4t(={9;9a(W\S%$$_7a#%%EpL"sN6d\]]!} kwo :uKqxxzC.77rvDDDDD:eݶG'X$R9[Gեv s%,0!"""")$$F]D,9s>^pL$MV1|7RD1[};
""""" 5>i >,0!""""'--
ǍEtL&6zhԫ7_1!))IQ:/Y
kk|ϧ_ƚuuQF=bP,0!""""ǟ=uVj|c0Bj1YOaTg \!!!Zkת#Gj*FȎ"""""*P~f\L̘>=Xa=/]*7|ϧQ8t0XϵW董PN #dGDDDDDDDDDb W^5ckkk,[fffJY?c_ib"ǎŀ_[ ٩3trj$nYYZbŨ]vQ%˖ Z\\\0t`tQn` 3l*% 51gg^
e˔ɷ<
'Obe8~1{;;|@|駐FȎr LȨnۆ{hƥ1sO1d0DEE rXr\]]-$88+VRǬ,-Eק/͍!6\~ Ռl|+L8w<ƌ dMV͚Xx %$,< ݻP(LMM`ccc (",< c'*b̛=ŋϗwك?NRZlY3gB&KELl,֭_m41T ݻch (]zz:F #ؖfİh'klrA_߾vD|ɡ0HLJĦ_7c!))IpL"C>t(܍"99iQj `?ݻqf1h<_WRa̙kߚH$·'_߾y~a?l_6"..NpL$e9l8J,i _ """"""""s,0!""""|cNػ7f͘CRSS1oSL&ì3Ѻe<]P({YZ| F
!;"""""""""2&Qqf͙[YZbْ%077ucbc1lpܾsVVXx jլkJ
ÊU++5k董PZ5#dW0D#""""""""XExDƌ BV+sfφWy/0h<
\]]f*ӵ :Zǎa:^GBz]e;
"9:8;""""""H\FFƌHMlİahԠa!##::Z+[ V\g<];{ ,[{k .Q#yFȬ`ޭwf4 L(͘3nݾ7m k9{c'GJJ&VVm,]VyvAvu,Y_:aCCvFȎ
*Q{vl߱C3.Y$f"(ܹ{MRm?Ϙ TgdǒeKq9cNNB fMZw?ARib۵?ĤRXJ%PjUaȑU#""""""""JDDDDDf9~fܨAC :4Yr%VY-};j4D"QWP ?qK/ÓOU(_ FD
},X`BDDDDD d矚W3{6bqT*1mtܽKy^NAs,Y wj +YF -Z"""""""""*X`BDDDDD]L9C3677Dz%K`eikk`8s&&1wl4o,)Hn˖+Z1lt\,!
L!B!B!BtP(ӯ >>^1lڴnSbLj19w¦QbDDFbڵ? ðYXX`СC)CWBF!B!DB!M`¸%6O1nxb+Ukaog_bѷX۰O Bd3110d]r *2!%*.!B!oT`B!B'w>ն+|>Dƿ|
~b*֨aCX %r }KNN͛qÐd}B}aacm j5 |1w4B!B!Q !B!Dk/^E"V-_ss??YytϛKFF="77O G;j4 !UѴY}A]z
L!B!|>+B!2?MD"QxWX"Z7ofF ']{` dءƏ WWW=eH!B!B!B!BH
~Wũb vmۖس'T1g_> _R) >M[6#%%e8~| !B!B!h
L!B!Zt ݿnܨ&MPqsrr0qMUX(⅋Ъeb/
ǎ XE9 4_ &LD !B!B!
L!B!:yvݫvvvƒ5nRRƌ ^bVVVXj5j,0 cͺY&M|!B!B!B!W_cmHW¢XGD`Ԙx*\6 w7b/Wrjs` hѼ 2#B!B!B!BFii8y2*)ָ<U\֮mև;wbUx)g+Əۃ!;B!B!B)T`B!BP(eʯxq A 5nKmTVJƍ|Rk3\
+W ƌ ݺA !;B!B!B)YT`B!BXlܽn!&O\1 <J*ֽ[7>wgU UkV+}6669|
##OB!B!BJ B!BXN9w+T H?ǏCBOB!B!BHB!B --
'OX,V7 E/;;7oRŌtbhTl T*e344D}0rp)CB!B!Ѫ= IDATB);T`B!BBįSwT!SǎE/1)ǎת֭Yի;Ҕ ;w`מ=aֵƎ )CB
T* gpߍm =v 32͙Yia#@j
!Bɫ#3RŪN8^9E}TcwC:v/ٽYKkQIM>l{:Z^:0N}D);T`B!B+pmvo?T1rX₍յعX=b`xЮ=Əw77=eH( Ϟ^˗w̌TdfBC$2 S >QFcE ,ӓY*zǧV/YYix.Lmc[N|b# !BD'g
~6SPۻiw{ q~}BD+|KOar$rC|@d+["Q?b>!-*0!B!+w?ܡv*@<֣Ǐ0nDbUVņ5kaci^od8?ظe387mƣo%=dGD ;qdf.77HNW &ߠ#Z Z 3!>O aa 99~sg]`9`%"#_b*xPLYYiHJX,idd ''O=fD!Y\Zѡ[^O#}ߦdf}s$27;21㊾ LH8r'BJXa]R}i!Iϑ 'f!B!Xpp0f͙6
j
X[Y<ŀmtH$UYӦXx D"Q=CTĉXakխ&OKpp2%X+Gx}̈́K#"^bʢkDiSbK[O}3U.>X 3"BcBӾc:+N=--GZ?3/xUg nR[m[íh? !*T`B!BW*== ~ QEJԱo~h!J*ֳw;{6O08V]jUa҄ ha'GH>
oۆV05@fF*RSN }<~]{_v3Dr-IW]ê5kξ^^8~ZjɟD%F}B ZJ
M8mU<| dR0 Ox<> ˹".6B΄B! 9RN߹B!B!BB+V[7Uï?2 ӧU1@ggJ,גp}\ ?sqq1cmN5@>m8KcqIߣWI/%7M #=L&):>_
Ojw* !қB!/&ۚB-8j3d # !R*0!B!+sylݶM]|y,[Dl0駟p UL$a%hִi[ /^իY4 8:8`Q{O#ip~/_;Ί -[y< K[ < ;
%NN}aWP(s4"B!Cv =~+WB! J!BW$̜3[-
rrX[k=FBb"F `UAJ4ߢ
ŪkpR gF ~}B(!;Bv,G[o {{g_p[6.
,,lu,B!Bv .!B!B0yrssUj*ZQc"66VsuuŦuRELL ֮_SgNCTaA4` LMM!!;sjd2)+VAb|`h(ĘqJdO]bb ^Ԕxf 66T. J2^E||422!WR;vIJMGHIGZZ"LDfqDժ
`g_A!-5^Cjj22Rq@*HB!:901eOlZSTnP2&F" +aIjFol̴_sP2Ug Vy6+#Gj V¤V|\TZe-]ɖgs 2+A(}-qJSQI^9 Pm.2Jp ӮsK_Rޥd 7\x혖-xJKar5W7;z3[|U>S,ɕBSct 0q%W7 !lT`B!BW@TbʴVu >IUV
֬iҐ
7ȱwB #,--!!Nǥ
>O^ H~@ Z-У8TXĎ
'!mgc[͚nGA$2+p̄hL-ٳgw7~_p߂!xX_Ѯ @pC ;O _R8om7T\K0x gOoGD|xI
ިM{["Sh>~?4^
wtԦ,]xiĽQi9M cwQ|iH rG ku{/4#wD%e)/v z7^:X>
}c
s2' P*O"M5KںkzlК}F&mk.z0 *f{c3[) )[+Eá# /륛?tvȐ,q-%jj2KOޏ.g}vf]l ,z곏V˻ }V/3UA
L!B
^oPmSSLѺ 0utH fjެ9-^ ccUiiiزm } {Ĩ#aog )OB,a4z0Jܱ TP <~.]G
;aOIÿGRA'}6߶JF|fx,qqnk\L
apz :" bѱ << :n ж SX6v`䕰-o{Tyf
yB!k ,ȕʝdÔ,q,qm\!Zk{iX /clg{FrA
yX|z/4ϩ=) eW`g*IA2S~ኡ-tXRYp7 ŦfZpvC7i
R c"urSM.W(ǔ+" XpvwBfJƔNl>`˥ϭ]>%!.-h:bAC G'
xn~~mer s }>?BHB!Ep"6m٢vtt%K!{^VqI^zJdegahש#*.ֵ+Μt+诼%< O4PIϑT⧑楙^KO.IjK4g\aoԾ2$hf0!B! &43fRma就)/0XlvɊO7G*\!Ho~lݾ
iikx< ڶi cS/RR^622w-=ey9q|#._b/#кhѢ'\*bq^se Wyaox+&uѨqg ) x6.ߋࠇ Ķ`c[ uomm_/$Xbk 83v_Gki ·R]@&"*5._:Y4S/t4Dc*-- #23mmˣS硨U \ 0½{f?$ EǨ@gF x:ΟۣggC .qvXWH!k h^i#>,:~oRV'U.]Jg w6bR죓3 _M.5RoJBH
=~ϑ]~j
7}^mS5Af[^7| O,WP2ᖀg|^nYS"3y 13~ФRCZT}]:'%Kl`3ۅ%uceKjrHQ;vK@ɃmEX=_4T!qDG_m6%-Ԋ$Ӳ]qjZ嚎ѵlL ^D
ˑ>s5;pCSإ:rŮs_AdmUطBT?(wVP
GBX5qۥWդH HOf< tƻ5w*<̌c\AQzr=מּ-220~mh]W9 NH۔13G=M kXz,G"w ]x)]-720Ѳ J&WF%nn*r`vی XcSHhB
L!B\7 b2Mh( !!ރU
ݸy}I?аP>OOOX (BJVƼD[g!IU< ]?G#t>ndY]ZXV,;;HKFCX2oͧ3p^ +ִYw88j {eԩ˺w[7ϰ>3B.u
jVe"u\DbFCn{Wnȵ-2S8"ٽ>.RWr4-- -aM9W K<>yMZE1.>md#X:^:U!~ ګvQ*ؘ]T#^6464x _B>Njmd H\9?ľu= bbz=B!B aaa>k&*FFFX|lmm w} < AԬQ{ws
= bx%k_''QmF|nrs3V!K~wio귇P(ҩϛ'톍jV
uu,ҩ'g?0kL%ۚ5g/#I XB!_L>`Muc?QX|n_:B};5[\s{5iP(^Zzb9˺SbeUs#o 65tZzSH" Ib=M#4ܹN_x:f[nǒ3k_RvjV埞#1̕ ,.j3Ê־w1z;τ|ͨB!/DfV&LlUl^gcxSU˖ضe+JW_c8d0?`_ Ǭ3q)t>!_>ccNLR>!e}+y P7 ɉDVV+V7: PY'8̗ {TP~>X!K*%ym``9647!%ҦDDpHr~ o>knҩG8Ǔ|f;v7I]JUsUm^/J.%
=DBz^t^sN9fY_E|N jgΪ:NwBb2#N>tSٖL*ny%&z,>=j~ !B!`fLGDd*ַw ]vލ)SB*D@}br upiD >a݆8y$J4` 3S|F ld2DFE!<< aǝ{w!H`hhtj23IJgK*C`/nfV{ 6Z-
^,Zi,X5{7ow`igu-ɊIRs |CʕQ.w6'O"=#ڵmi{uDFE_ =8mCB1wN :ܜKLJĕ+1zqd8sqɀKĬ)%*QIlf=t+NH\E cxx#xhɚ*^h!gQI>-L%21r3aqpƾG6Q1Kmt:IrrtN=QU]nhS#(3]\ wy뜢edO+ (%LPr664(җa[s?Ҳ êZ .yc)Y2ad/R~~鹒?S,SáH?pM,W*!z@&B!|._7 b|gJ>s&ϟS ֥K噔
7#G ͅP(}aa.eyQ$"UL)a7J5!訠R9֗B}HO(,C"Qj*1ѡ?"vE~ EgqeggJy>- 7Ϝ BP`0deeQjwܙur9'&"<";F6ߓ\.699Yc aaa߷wo!(7Tc$<}%Xvvƶi.2 +j gχ&_͛go߽ =T\|ء#'&5yr^_u5<~gbff Lބ q
L4mҸ D&?9mB#&gj|}
Lnݺi*0y5vF݇S` XRʜT۶qJX`r4
L4>M&h@C.?eDdd<ϴ1/~{;8vTA7R1/bo1l0#}5} *Vƪܢj
D"/>xk_p
lm4}A<9CC7j?1g,Nzժƶޚ1|L4ֵ+jԨoNZ sgm+ps622c8y~s
ᬭRn`U&Fq1C_i}PFt%Ҳrx'+E1)-
eѫ; V\iQ5$- G9<:)pg
^$7NȩP2VJG}=]gL^|>@ f-(灕aX*˝ȈRPJ~i-"B<@+rRse ;LA iu:-ily2KkɊ~ŧ\dT"kI@),JQX`Xb 9&P !cP"_31 3g"<<\ݫz쩱}pHF UwzH$8x6o݂T2hݪ&^]&DR6I ظO\Cr/Oxxx b6mBXXKU+b\V \tC-~M .Hc%&ǰWBB u)!>:إԽFII'D%V \GӲ` 18τBȧ,Y}זJ
ofO< ,vs۷ot>=6f_}qiF?"OJzn{ Y|]J~f)mDEGKn WɱH% \:τBȧH/Z?`ldÝQmjzx :z !X1T4 "}waHL}
<Ɗy{H> UۣѼYsNN
Sd/^!8a2y8y<#O˻xwѦ@L b[$f4Kɖx3'uʥ@4*>뺌sIDO. Um_>ЎLpv5u]jN9[^fkŧ< -su'*)(X_{ 'S(Kz]zХϫy3F:mkZlZetr; Zr9TKل#4, \њH$jFȨLfm.s*TD葵0Xh~mcLҰzOx*z־D 9=O-
uk43?^/< jlߢxݥs: s{TJv?֪Jr NƊyxVEŊ5 gaa:G))q:d]S'u~/T*ppRVժ7ʧ|ڶ#|a[PO ;?y9{Q}+չ%cٿYm~}\l%."Χ wޛ:W GĪs'$ǐ@օP2Cƪx<(Fq~"#TlN+۞
hvUs d ylŧ% 1C2ǬM7 _ 3cV()Y%v Q !B!a0c,}\^gзw V;Di3g `￣˷'Orjܽw|;z4u')LR6$QP(E "$ ptt* zx*#a
ڶ CC ld2 cZx=9{f. `u `Uطg!Vd2 V DY+Vx*vWR9I-9no /tu*0qZ[ꕓ1}p_Sc abաàڶ}:7m
Pb=GسoVL$2EV ǚEhݦ_^!i:~M?^.) LN6fz:ݮ泂tK
Wq t ,".-NKmj
|,=GRCUL;z; zCzz۔={Xҳe' | P2_̮jԍ LMwwBb'~I
Q.<1 3c^vrTFzi`I+oeqz# 2u}
\|9ȝЂz `zľ TQ;zufs(*)/ k_PIfN>0 vv\sc4Z+ B!B>lKիc6?i" <|ahPxK'a5gkkQ#FO04,F$0 8GÒ6HJ*zY q&xxx̬_c? 2vӧc.`%BB \
<ݖ$zoG^xH5i A,ۡ <訏O،L 4")OB|4֭_Riׯ9W/;A,X(sHNz}'Ș'&
֭aa왠j5Cf =f
e緱nͯhۮ?+ҶHK }x/_1LNN&vlXqw*hq-=~oef1|~WaǶ"C̬eD"SxzUG蛧Z[ xUB~=qEz^v}W^nΩ,ƥս7^?b{qH8gWnk ϐt?_o `4Vƥe[Aٚ{8XKg ox+a T2[A 3 挦>3_oxQr <$eHa}s,4>MbFrMT|N02,>GQ]cm~?&#cHbFno:n,Tۃ*]O;Nd-#d pmyvbqx= IDAT%m Xؼ˞ÿX.)V!>~BiwNZ_:c W\MJpwe̯r\7^Mm.>vlC=^47Y1+ؘ=T/0Ii_~TT
7S LAcBB!OkװfXl9kYX ;5É=6]__u*EDFb8{ n C`!| QQg!GhX XT#$V`%Â 1{X7VVwp5D"3de!## o:9/(tWyZ~ w !#FY"ܾu 7wMm${d=DEdb諾9!
q2DADٲAf^頋#m|҆tSqp~t`i^ LL-q
Je2 1hГZc 3_[7/]Vq 8yb7w X ';
W~jJ:֎7U9xسKęӿW |5?J~ܓpdd$!Tz o,G_/XXڡ0 ai)/Xb_c }7hk2pq#݇ *
iq dW{ ǰtQ
gg5 70F/Qҧu m-Ht!2%bdC}aq~$Ha]PRf[(BGr:,iyV71iG 4,0حo%ܙuoMRVwN8kkn*T9̂RMjp~qk.(-˙ÿy?{V ϓ_\ozUw;`KÜ"Eiznv|HReTBd@O}ۿk [^sg~ZuriaVAqb;E}*B(zk"K|$fnBӢ1؏^9ܿv)v̨VR#}ܔ rދQ"=/٘g(U*b̂Uӻ։|kvSM4q4M5 MΚ1Np2hgn`d *ʑ+ Rs㾯#}5/Yנ{ 8_=_5Ԓc7xKC鹏,i Ub KLLIJW@xROO_~###1w|dܩe3///l~^ iiiؼu+qڸcccL6
3gssZ@QNnneb㐒 ӕ@ V'{m= Y?~Sc;m/JaKA X#;wŲ;j,Saa>9LM-zc2t3ÖMP^^usg{kgwvX Ŝ`WoAjXZ77 ez8^9^yXkxtkZ'-زi)Ο;/%%))qYi)xe5ÆOk'qrLDDD85 ,Ty*ʼngt¥5ŏFL^\Z=,
";f/\UV(Y7co|ꋢr{j5D}3 kkϪy[ȅ;OLCV)ʔoj%=%=:_(<ݻ N P}p-sG^XTVOG^ /uRФzHasoZͼGXȔתvN)W,&gugf zڐ:Jh\Ob7^Ly
:n SEEEXTyeѽfҿbC^X=z`
X
ܤI3k6lllDT*RRR N///6`JB&Oycƾmj QS@GG{-Ti LJ -ˑ x]
pppot zbG֠sD"=<:EL&&zػg=Bn$X|vdz ) NOO ~ؽk t>PUf ;{t/%w;[yE%e.'rvӬ|rO$_LWR; loa\V߱-ak٣E)9 '?'lx ^s?ŵ@Yn6?AM%OwDi,O{}o>G$7§&z"a`?^ q9_msFk_/E-((.mP]@PzCЩ9~r_t&DMƀ Q;{"&5ꄧ)S4 : +Ԏ 5
>] CCF$Ï;wXkH$¼WɩOCͭ2YE8*B$ P(Ğbx{+x
cuxPYYc+Ŕ߂4n_]\|@ QS0`x?oBxek gnnxP 8@ /b(dggX^%9>`ߑ2?=HnB(Sږ)Vjkh4KØ@7۳=eq-QLQesTjyȿφ'حbr*ĸB P2]m=ȜQ]Ymw;DwG< ~Z;ԥޞ^G}Uy:4F C l/ ȉ{ -|\slmlv Co?gy-nxRVA yIZ|. mgk+|].NV@_לCo?yZnW/xFdZTjo=-VwYJ-(WڣH?c NH'ƆiX;"7.yvS?=>
Nv
?1q>+otIigbfKqiLiV517n= &Лu瓕+ߠ.C;&v8 !+MގoƗ*J%a`P7~/6ʋZӞ˗.)P\\ wķ)xAcpth4Uv!G, iiiZ ]BK,X0Uvpnߎ_Z]%B.AYi)LLabbP;2<rXYC__ ^STJ rQ(5V&"*f{q/xlii 2all3wEY@ s fz=;o@ ̚suYDDZՉuKںR\OH*0u4)v*t2խ}`;V\Z.j
8h+;FX{9ZH hҤLش\snKך [Zӱťsm
D>Ǽ\x=>2&uQgQ[`""""vذqfPTXz~~@7_ 3g{bϯ{믑U}Cd"4YaJ%*G*G хĞ;F,]no kkǶ.23j`P(-,`nNFwk2jos """uooNym]KK06S
toجvޢ"--qvFwt먍jkN[ADcT:;$%%aP k Be+VG k* ؏-۶!5e%[:>5PLVхJ$11Qkԑ BuNNN8^""""j*5Bxpp0lllQ˸b,~5k֖/]^={"// /ƍ1ffӻOV8|6nYBB]db۷OFKOOӼq,$Q6wGۈbqDDDDDDDDD11`BDDDgeZk O^ 8P SsCXppE~(UѕOWtpr/.,b""g0 F04B' ;tE@ wzD/0 }}DDF`Y7m'?/__}7yرu! //O3^$.> TT @ GGG( Ix{;f~"""""""""cEWKGjw|&L*#9yt4?q~\7 VMnE
|Uk>iyY&fa^ {6^Dye7 i @* .\ku_x7PXXW/|zzڀ.UńH$j{T*!..Z$;;500@'Mjظe1`BDDDr@J5v*9wh
;q"|zn
Iy_Y17)"r2
!GIQYcK3 ֍9w
UDRa= 4d`5ҹ3o[˗k:aD"[W]{/+Ǝy+?5':&_mڈ'OVgcmٳfcɵ0(Q( !NxM$A&CBZYYAy7D"ԄI\ b""""""""" 0!""F9VTJ5NyGuЧPT@QY#VEqBΝB|ʯ5+}2z}F7SO&ߞELG3E^vQ}.b=%q@P,\9Ǣv%Yk_Yi#F&04Ffj~N @jZ>[
011Ţז //OsQ|&ܰA_|| YIIIظe3obnffӦĤ}F\|e^gTF]S(E Tn' """""""""j_0!"""Ňg Vgtg\>>7&m1xU)C?\ÿGf9JxsF9 Oܫwddo%xq":yAy>A:L
'xK9n8&6sDj_v@ ZF\|1j]3'7;⾱.4q"^5vvv-H2$ (+ީlmm%KU%L g';uDNΞm]5@_ל9m]=\0!""")qzȉAp[ï+"֬ f=j4l\qj]9Qu ϟ?\R֚ >7
NM|g8CӜvޭ5,BZP _} {Uv)??"_J x'1\84S;w4aOH$[E ^bMܼiDDDDDDDDDDLH' FANf['Đ #hL FD뼟V=\_Mڷ{y6Lwr®$#)& u4%y@O_ԦqO mr ζxLqq1 8SJ{#oي |ݷF=,ӳٟ#11Q3&>> qd@-LYёīr """""""""OZH'hm K (U J
KqPFN
j:k"b٦mʉXJ-ql(.îql-OU6!S ;! 9fc0䩀́0dh c}Z~ݟ pwnj3䈲2o ؎j<._sl
4lIT*}]GGǻl*F{'"""""""""1`BDDD
ZgFN .8{0\v|ϭv0 G3 {&G~zhće9!#9
rq۫t8
K7=.(rJYY_ܯϑޯ 0 -vݻ^rĤ$| Wp;%1{EYj5/BroM,@LC룓f bbOO4QC1`BDDD
vrjU 5[%Z+Hφئϸ=Lۨ%D{/쪵kHOb<=wR 3ZFJTv݇xZ
ɍjԹzzzǁj끁xm" ?DP !1 , qqwߖԌC)DBa3>5&DDD j5p|-!OCO_ 1l ͐^9L_>UlN~3Enff-.,
ZaeՕYK\60z N헑 2|6o?-ױ)Z! Un{w-//E 1jȦU.$w!%%* @(ٹI0l 0`BDDD
z9 i ZkU# 0lB ~f>)5BQItť#%*5n`qVU6g3r, v0è]1lB {vBRtǔ*JR5G\ZZƑGZ > CR!%%qɽ@INnn!<={jI:yzȰ}Hw Q ߣ ٣T IDATݽAN5 ;YV$'C㪍iKG frtͤ&N77 }{
p`ýdk鋰`XfʉFק̔jk V{T*T6sL=1rFKJJJ/Ut!C"D"KH@iiikbX,x[,3k""""""""" &DDDT"ZݻءsW'JӬ s]Ḽ7>7S}35v3`r|OTJf[@ 3\r@( E ()*ktV[U?? T*juׯG=633.$q#--Dpws T XXX4DDDDDDDDDD`cu`ŕ!{cp2 1.Iթ s+jkEmPI{Tg8Y6|[gs |'~
ij_CV @'ŹrRDbRRHYhcb.$x{@__%""""""""""""qs1!rN '_buԠZcC[QQZk :ǵ^(_Qmج#Q*V*RtMZ ~?#LMM&qtpxyn7X GsQsŇĄ}ں""""""" Q=3y#Ekmcnm#qpfv0). &11 {7>]w'X%f**.]9|AN:Xfջj5 :&VPPP5{
|><= u#055mH#;qζ.1`BDDDu:G!ҠsG<0I蛩
J4ܒjku>jm$-z
Dp:&00Da
EY
c
@I(/+GBbҠV7{ICXXXTiy
"OD!""""""""(0!""Z)U7ڠq00lط={yٕacoINq5sk6fEr@HSF͘YJªOW=Qabl\hk^!0 ̞5 ^b1lll&D3KaÀ Q%/CVMxXeiw< _c8?100joC2N%EeZ
ȤuF8U[K.rfbb|4F @V ֵk߃kPNeCkPP[@DDDDDD)L:7QZZ
J+h5Umݩ4. at
v[k[ *km_IQiSi"S1zJVk+..FQQaEfV$mafV=CDmg@п[ADDDDDDDD8 t>N!k*\iMIљɐkyҨ*PkX^ڨ#zcjzA n~%U
# J1mtƶt Q$/C. P}R(˛3 H/%"=)EUGϣ
*vR%Re9:_'N!
UVL
࠵ tBYYYׯm #j g&Mիݼ'""""""""""Gu@QQZۆF Y{C-labiZ
{ Ͽ1kb'>䩀V.]:A(@l[y#.b>17S:cZkt9:Hw77B
Bj
J4L
J%~ڹ? gH$J$ D`@ LMM^DDDDDDDDDDDa60n#96Kkmk pXf^)6HVnAiw ,FnέVGCq隵 G`3:].N&/ϡ Xv@ ֽCL oHe %5GUxB!<==X:H$Aܟ ^ CDDEDFj:8ֶHOATЧ>/k5uRq6ھɋZ 4Ok?q 1)i t8˪~ ZƗo BIai#N!Z\D"DM\aVGJ8 8x|Sa)?oGTt4 Ã L:* ; Pz ƸI5Ʀ(28;t݆ɐc⃐^Jo˽ۭkhBr>9 +Ua*u 93ܼmfYå#QHYKianϥX`?c̞NKDpqq۷1ܡ [wޚ۷!
B*
E4a(**r!4, V#_$(0~>""""""""""z0`BDD(U*Tu"u6A,510C>gT(sK`fe$><~CQZR=h11_ws3œլ\Ld0132lXn]Hho]HĺjpLӧs
|xcZK7RbNx {TOzR ~,NZܶ p疕910W`cc7WWbG\ T
iRDEEkvf@I$w'oQǀ Q BtK? }Fuiچ> 0)+U0<1G gUJKq֕e*(PT@nVi
NE^V `x,^ ^Jĉ_C}> c~A|,Zk4Υш
^#ab L-Sqȸ 8BQ\ج&թ+W/8Yp eJzrlޝ
ĨA66&5^H@=5\ aZ*zJG`lV.:N 2x0N<"@ ߇ BtFgoo00jo
)r3 5kjp_Wds?Aju4q5^tal^gP};<3|_ysM- llZBauK]ߴ> p 7 .bXۛB E*BJ|v<öAWzZs}}Fw9` +- PVV+EEE8v8FUB$ Ӝ 5^vJp=HOO'4uun@Ȱ9DDDDDDDDDDq1`BDDDDFhֵ{ɉ5x{"?鏃^լŅ#>, fOM -1n3,mk^9{Z_bռHӠsܼmlSpT%ʬow|~p:6.?gWqZ Zꢧ/=1y.Eb+`Ӗ-u _vmpDeǒBݫV,[?_݂Q&&&ݫzYՌֹ$+7jC H.@" t;ҹ D"Nу""&J]&*' GFumЧ&@E ߡAƦ: FVt1oӮzC/n USq1)IE@uYvsi^NfF6!^ gO2e?p ii)++o ӧOV~Ȩ(+WbҔ)x駱db4 k++ 4
ҬBP)R)BC /xRDDd""#} #CC#(h <5㒒%//].]Z~===# 'N ,< RIh(kFff&9u
:YsssCWIX I06Ո_vj2!fnf L:8/3^JO}R>[6@sww+3_4ko1ٳinnnXn \W#22Lv1q$L|v"-\k+V{&### =wC*E:Rwzd$''Ç "^^^A"==Ӆ5KDm9G@?%""@v0BNnnVCͥHN/(Jj()*Ӭ0jRP TH
5{, 8RSS /ϞN V0w|z
ݫ7~ݵ~
6mDvv6 @Ra=ߘ\j51`BDDADEGCRiڰj)eJ5 @Gͳ7TYwV & ̚s33|JjT*|ǐ 0Wp%Z11}.^X8o>,--[I$3 *:GD
TOd P2$;'gΞřg5k..H$ B$055mg""""""""""H0!"" ""#4oI7Z^S L~G/MZ:e
tbz%Վ߯ {7mԌR*_CX0o >7"у6244#ݺn4kr9Bv9 :IKK))HIMc TDP`ED"-> %pssk!"""""""""j0!"" T*5~HHHlÊ57i[B-_߾f]0 CB[5'ĖpyZqqq}gΞŅ10 !!!uBC/..Ƶqu͚$T ]<Q[cHHLt6 ?_?L:n]oŬ"++ pi] 6lI
aמشe TtqN %lߺ
"Ə 6nį~RPw ~ o/[><b1b1 7 P^^hHC B*Ell}{\t /]ҬC"D"$0LDDDDDDD c[htoXf[ؖ5Q` IDAT&D###lXcǏ rrs1WyFѣײﾋ)=UkVҿjEEEaY5b$&ܚY:===# 'N ,< RbNh(jƝ;)sfݽbDI066nǡ& &C.]ameQYә1imژ ;gWeMDD60!""zEDV177KVCDmA__ Y~~߿ /cW/0x ӥ پ'k}D;'O̹xq4:k6LMMY::### =w!4,!PHC ju_RR Poooh I>>?fjW +py{ v!Þ_FWTvr4bcC&@ ߮F]A$oRTiROjNOYʤ };\8{Dm&đYK?v/!zhB|035ŏ;w J
,\W~Gnj#AΝض};r @ii)[?p K-P(lyXZZb@f-=#RiXPxRBTt4 t9
X,@ hg+iy-d0Lqn89{b8z?ΟQ7, %u _R.DDDDqV o\[(oogbDDDubY &D7@˖ [m / |M}}}Kx'~WߡRUt |_veѫgf} G# TtHLJ҄NnIGBQ[!!53S3 n$.zP(7q#MVZ 7c>jG=l ں""bP( a ?_ Z
J? r 3^ިk=VƕW5#"0xt :yxxq FbbbZ$Dh(!/˸|fAUFH$X[3QNǚU Z~;{Wt6 |aii K[(JTHC. *~ . ""bwwͮ ֙v7-"u;^"ⵝց^yh4Vd Ņ`BAAAA|9H AATcP$ ze$c-OJϭ-Ќ=St6Țu vױnz$&%i]zѣ1y$0X,<<< r#8$!
xPwlܹwwնձPZ^011k"*P% 3#I6|6\\}JcЙr(ΜÅ\]}+ >>AAc1b3>7f$mz?AQ `BAXX8l6nu'ɗEr k c]0 *
_;
1\0Ŗx.U.]о];?x D v5s&! cA7hmDy]IZZZs$%'#)9W^ 0L8]_x~!>Eim,##c̜
u,\u:t uJEsڵ]WW?88Qă{~AAA`3"17Zt1ln5Y$A:^ 0! >[$ A#p8:}`:ALѽ[7` >X~]
Y\.&Oc-8s6SFFf ¹_!CN`j͛EڶL (8 IHrsgh4ALl Ξ+p8pwwχ:u]]QN^ޥ Xh \kN6n=F÷~k_3 ,:AA%Z>y;KӓU (/`BAXxx==h ѸQA6[ aƷB$.8pYDb[Zop!lo1l8VY/^hυ`1գf Q1ѡ}{th^ۖwuC
TwRDHHBBB |x{ 8)vboo_C!?uڧN_]\l)D9 =WuPUɺ ƗĥVV>((Rp|Y *cgi*=A &AQMQ('=dȑ5rdeo TFwnL692U\q 3[7m}qe߸)))s/_[0~xL0B#*=ѳ{ 3"]DFFBT /J3<}LfaawYN|}kkJkW\Njn 4Rak&M*lQLkEG1`2YeZO"˻x>234H$B0 ʺ6<<ׯ4lsVqZ[v!0h"??[{e+I$*^7HOORYp5@P07Wԯu}
|Ǐ.#:#!䃢(0,C ~m ccA \qF\y,6-#G$sTqXL13<_x[zφxjzP|_BeVkx|.;ĈYߩ擅lMU=JzVyy Reb2\vf
A- =}C\j
dN5de]3*%xF"{Ta%+e
5 f
\N kձRqӺmqUM| {N4}#ی<\AҘ)a1E<+es9yv5'u{`%0(HJR36_|*.7=_RWT[(rXyF v ΫiOYƮOD#&AQM%&&j3 g )^YÀ4u*ܾ S
Yg ء#ߏ]{vk`r:s?{"OƘL&깹 P(Dpp6ɛ7h 6[[w''>>>#
(
\`00hwUG(}qisJ i HOKkq6ͭp`2Zݾ.1(\ Ϟ@Xh jUSMCQ$fXh}].V>+!K~ 1i)
7RĦ
MVKw]&vRU{}oũ
<%UtRd H!8!sܖ_~߫ц .<۪P}%6 *i:8eLjnlܟJ~x\|gC 4{iȵm87\?&ud /< %BKwۚ=+n_uNh1/{ӫGWRBv^eP5K?" N$Z<=1V.EgAD5A;p/A8e2 /^ sXZZV:F< Mc߅0o >zC}/AT. ?߂l$#b"]i`$%';Gjj*RSSqu NN:>WgD1^2K _6u*fQXNN5RS?==, ߬)W!F{7x df&s
'Yhؾu6 _ƋD.aoaúP*_bbtFAėa͠/\Tk, GQ`I>aI mɕWq6? }N_FQ
Nh¼N6>c5?}o$:uuid"q9ba
^ť 84YF:A226d EQ ?(>3@ش<헖MPٿIϳ*Tp,AT)`BATX Zj } cgǤS]p3-<" cƏîmU+/#F`UxܫׯQӫ7fjTغD111A&MѴISm[vNNppBw-aeU5jd!++٩JArR,2gJ' lu*Ƽ96)YYF V60d
UMJmPզ뭐G )Tu㰘PWGn8ʘy25Ŗ(TJu- Pk(/ֲ 4<;f汘LL)Us5vĦ?SQWGԗbr٬l CQKTT(555d^Y/`ȹ V&fRRe%Wk3}y"&AQMG|HINa(kkk߳SgLGPp0 M\ FcCKXi# y\q 'Nĸ`D2WT{5j]۶h-9%ﲜ 44V0Zp' d֖)Ghc677 O
$'\]}1u*89{JE8rh-^J ^\mk;CAhּ;-L& {qNH$ ʳw+7 WrSH8ub;N 7Lf5LJƝ['q8rD{d0vbt8<^fJb<
;O[ tj
Ř@TY{3lu0P|4hpRw 3{5
7 HGQ)$+gs=1?=MhcSo3&nc 5N|Dg?v {^|iK
EqvUO H :yf|e
o@?B'e쑫>y6_S5 W^}[$r8٧땉byj~ADiݳҦн6Bimf|nhk:Nl/gw64AAD5 !G {v·3)))=n,v \z`w^ܩv={@& d2nߎN fW 6Qj u
]7oNxD
-dr9^|/_h|}!dƅюa`pq(ӫ), w o G,RL&foGT}6m_iפ(
o'=|j^ޅR) X ,-k:N 7|=j -01c{1盠c!iNi) `1_Sjnl^2pD urzRHNu(agy쾻JWߩS /)Y^3-5.fijج>onG=>yv^*
E$rU F}`(6!-3_N~m41kL] yٯDIl5ۑmVr|ӅTjiƅgckcy)=I5JnkfugJ״m^Ӻܾh 罧i fɹYi,MӺ5֭AHرt`s>]Ys(`BAP^^ RS?Ԡp'&A1ܶ ۷ YYY;qܶ
TFFFft 0 7mKRRR0g\ >r ΅$>Q1 \]]~} T*""#uCFÓB0 @VӺ&ޕu\(J8feeWW &ǭt˻`0ѼAu66Z[3U(&LZ
5T$+_2d`jAƼ,`M~1AYԹyW]u#꒒++x'4q [=ѽA͢zX{Ҧ{~1-Wܾ&MG>ٺXj5? _ _ڹ/_`ȯhd{ >/l6>6d(~_O8w z[eVL6
ڵE>1g< 8x_LV϶ ( EDPLo=TGhm={E21fܢV4[*DG}~eR1oddRl0
AAYԙV0-!)CFç-;x[] G|&iѹ|;ɵS֞{'Qh5py*m^zϊ$:Oc 'vN:~ۘ a Y [ *se-^>7dThHv9?AUd0! j(ƍq!8$Al'%"%$$ !!A-dnݺoAy zfW_E"$4i%\ M[2`ԩzzY[+RSjc_JCij3<Ei*,u:`XP hK"DF]2?AAl͙V[s#EP֘;; +6^u OR?:f7q,|,Sk'gxvTi<8,foZ)xA5vuVHߝ,!/G,/Y!>Y5͍ 7Iy4Gf
'e<+s]ÄZ˔
QnylVfᠦ1K z$ AATC pQBo a2ybL0QۦP(0k8wtm~}cɴ@D[o~z'Q=uV6e
mނ7nؼq#&O-Z@ (>B 2*
'OǐЬUK5+V¹Û7ot2TrB}B""ю-lagj<~*jK4RbqΗ
=8++c88ԣ a_u9́ mB' >9;5Lnҕ/Ux$f?/<| ~w4Jl5`hh(3+?I>=#_J;[~US IDATS#ڐ9~ ƔV>]
NU2^,H/oUBtJ]a4
\_ϸ`UQQ?A_ OzAn̙Lqf 7.^ PGk||w99P(^I,=T4zS
II 7o C}ѰQl;;Vkh xy7C=P߿m }Ig)eRk̓E]E]oSf|ncMm=N KqT.EY eK[oKu4{X+cyddlLQp qi&s9 ҹ6W,Zd2hfx 1C>Gvn$ߔ PU5|k@U)J7Pֵq#2>ڲi1kL˓},Sj!Gkvקw jhA`BAՌRDLLӃ`%M0,_EaD:y'_߮ 6]g#c՚
Ӟ{cwʊ<Cb0prt#
PՈBppޕ։Җ(J$!I j|>J%LVH&&:m$tv\FrciY{{}{Cxؓ;s 0a9:}%~y(C L7c`XpqG#7lo`9e^O߉xx obCha -3leU
wE
JDQB!mlltJ|)Jdff`|1`0 Fzeff|B̉ hl6:}e2J%T~R\! 0L\N&&F;/ 6SfBAAQ2>95֞c;8/Uxgzgu#(8>O;ޤԷ閅G5}B~1e\瞤٘?s6 ״ .U߲I4(}@5$ 0ȕ7|.; 'k5ōw'{+ipz_}gAGL J {z~A :&&Xb
-۶B$b +e5±GpYlU{LS|
N1F)H⿉b , ZPINgw toEy3p\z{vB (Sph; ;;MZFt+DG {盠~hn 4RL& sĮzde <] \\}ѤiW9棲TFJ}c߁gŅ^"($$#ǎљco~6\; =r$ϝ>j:}[ gϝq\xTcǰzv}38QoqרYS]_~_\_&Mo ` VFFxNG`:OS{)Ӧjd2`%w^
:vЙwa5sN_m|9F
yF+Wƥ+im<.._{Il0x Ğ;@/k4w sfF]imJ[hI~0y$9}36?_?\\͛q
ZO=ߧӾufԱ祿n7ء?No =1f(ER2A~o//hN]{v6;{5 >zD}]6-226n;nXjْ֦R0ouvCi_rFss~Ç80pa: ΙҘ{4|04iܘ֦RDϺuV:V\FfŅVW DR'ɽYLC}n^-E8E hᓏ ڿh߮ßB$SFm4
l>(U]CX$e%e]}Y.6f5p=_tU`%^x/>rVBWB$o@ iD꛳u05g..U)ArgYT~ֲ Qz'4j{˄RGJQgyvJQ/26.=>.=xʌ
wiл뽑mJ
Pb1Ԝ\}/3Py6y6ߤc,[ G5 .,d7!gAAŋ;+Ʉ[zA|>{? wEH%)d21 t
ډgi=8)q\e{d@|7 L*I ؽgv##c7 WW_ 02gs$_ӽ)9jܤ xp YY)Jx2 ?0~⯰u*? ԨQ sGq~DD\uMQLe@\Y[~ m7߹\ruo
S`10ѓ]P
Ԛ6e L
~?Q_jNP /?O: ʑ.Ht4
-c{YYYz u :}\99Э*SzO"훗h4|5__>#ug b`/5-M/P</Uxgzg? mv_Sn
gL̑(Tjoſ}Mtq$y " jd58? WWW*AO>գV._SKLL
p͛59sFAWff&<{ , S-Gvz{4GÙھn=ڶyX
}=vvƪ},},^ w2 >^@,ePp8oI$B}g}bڌhԸc
;~l#Nئ=kK_aH ZۚudXŹ?FpѹpxgO#88 1ѯTq
\V:{Wgxß"7l>_Gfj;] L4ӦLmg(h(
rN *JLLEߗez75Luj`###B!mNPx<U9Ph
d07-' II*(( [bM_ .\L휚w =0( sssXZZ'&&BPh4|,8 d0 4(h4&&,9"G_闒>B .px;'yXʓyE$4^.>pq3T A-Z
PmX ҙYCq0 9)A/߂JEGsШq'xzS!`0` b_Ҏf9PdU,KoҷF 55@O& }LMLuC`_G NۗJڵcYF/051-6LQ }}E]%j~X^F.iM& ͢Vl{^XOl&!45zZqφ.OpO,}Tum &j
^ĆqER7}N(-Ce0щvuS\)_ aT+*o&[! V YX^`લoʗ**{TJ)j6WKRvF9"Y}
ET5n-mjiA7 `jU/
I:$?29Qz\Ud#,){䲓^- TGD PL {Ҕ cb4w!Giڤ)N =_^BXѫOo?x "Oyėӫ |SO\EnnFH966\}fPW/Ҏ6`Pp %ԟQ$Hk
A8pxk_5.`w{tvu[ј=Olv:(
)ÞA{ \"Q
ֲ89+w?] Kb1>o~/CY:[/F 6BZ|gQ v5K @TJ?K e &L
MMT{a? V&b
O% #鯿W,
UCɕe)3>֭Ao{ yߴ6.E+$+wkHy㐵i\\tը:̴45*H8$zv/ OAD5Ozy &ATznng~ԱœO1~$v"0L <_1ciiDž"֬[;w*}oQUx :$$cƏGzFdHb8{4:v@;6> | &O蘘* AT^p~^VjVޥ:sh ཻ
#..qoB
#,Rfڰ?q9J䘙YҎ 0 }laeU>-imQFh~? KSc '=Яn2tΩ}դnJe*:X: /&+ *51ZE&gZg[QUb W#4 E+$:!s5Mz^CHJZ Fo J2 lH0ՕWq6*ͧ/_ԵSmREk}-v/,B b F#>Nu< ̥}u\5dMA{(* G9"!@?`ecnB֣Tk,[ ];e0 Z,B b F?d0 KZ5,,wn4nDbb"FتҲE /\ G5 9z=CV ś֖% 11>sBXhn Ѿn\?Z=F1\.ņ!KU5-GV6l/.KȈ ≌|Q ?'Կpz 5uH$r,2?łA ⋡/ĈnWȨv, ƒ!KvLr{"2%nc!W~Z0S EKvuܔlLI *G,/Q)9ƃ֟\]7z6nl{K o)
m^̏M6PQ)S
ݫ KK͘#{{tiLo& ؘɔ*G+{'egH^<6+"?0ACQmkTAT `BAHxd zATSS/iZۖ1#$hb21b0\>cF᳡|Z
ĽK /Yؘ%#++!/li4jzy u3O U-.{ -66*6&//[7¹3-=S zPwsc(s;q:*U` سs RS #cٸr]=6|(3ޜ9%"7=;n8\RqVZ[fs 'AA| (TK[ ;liȣ_O>95}en"Lq߀gg\ygcȸdlZ4&uZCa˕jp>_S TKL͚/Uxo W_=txfcnP9bۘ9b6ӊՁ_]<6[h9ÿZ}h`զ2oOcƩs s Rte.\}Dtn͝gg5J O= hTbU.~>ߗD[rɱHN;73`
C ŴW/˻`0aiJlP3ٱHgp A!kۤR19Rw}eg,ӼfU1pN?@A S p9^'arYL\Ȕ*K\RkEױ4۰֥o/Uxu_9"Y s\v<Llc&+zp@Dwz+s8,TPYe3rԱ4sSsu H~LT8hCCQ\q\qZ yقCC 8,eqc,M-X8:&=4qޚq JPZIޅ˹܈{}
(U3+>Vs9o Z"W:k(E#Z;$.L&i(/i}38xp٬,#N<㰘RRm&S,r, vGv orkpBkpB-_M8qF vWk4\Be)QlJN}J zT\A! j"YĨh?&uTxv
E2{jCrKX_8Ԉ?q؊2yڹC[z,-Z*(z˕"^oz(ϚE)Tj ajCBgzuTULp(PȗJHw8GJo. `BADDDID 0 il J¼!0d*a#F^?pqm\,[G ü9sѪe*-AT .]G}Au8^Ex(p:C]h%GDFF3vz LF~^LLaiU h~ }kCĢ<<|i4;ȡ5( Fm1lQT=X ZGabb `Yʔe! -S{zev Yky 07BHp BB Ml43hѲ ARH+d)9B '[e\RFznWc#Mu./Px;X Ml7GK,ģ{IZi4Q>lSho%ٷ)]
jD$gSE 17ީ9vLy=QT}D1 e][Sumij T 4{iij;!?
Vs٬>]/ *u:9PoYyby̘: 4.OӺ2 $esr
E) u:y|-MT ya 솎P KxGA%;xV]Ca9{֏0n\m11Xf- <9]{'8Q wEL m
m\.ExA
ٻ(6 :KGDQi&K{oQchر{%jb{)RewυuQ Cy~p{>;&!5%Ʀ03 \]?BEj^9-~۱P{\ZC|;k{x~Es$Tkk{ؗrG5uőHܺyCsD2.U9 r{EDFsεǑbSݕjH
eQ]Jѥ\ =yo2{C{} JpMN˰MSl26jB&e3+Ӡ\.h,7&p^[/R RU6i*tZ24@ c0d{uY` QVQA}(0vL|ٽ{ED$}cu>t(ƍ+b;s .^Our {#0يT<[?;i;v'D"
&D6pձׂ"G>01|xaNz '""* ?ym. OwHDTht K-AkÜy?u+15mԯ'C\
WRa˶hӮ-vSa +W'R""""Ahl۬BDTܰ\R*]4DDY˖Xrgn1}B٨!ѯo_ =|=uLkS˗ ݣ;|zc"*j5֬8gc[?XTDDDDD%mS3>^Hq 6~YL99D4DDQ#_Vg{a¤:+4&VVnL'ԯ3AC`܄Y!&Bѭ|t+SSdЩRY^#""""*}HQc 55Ud
>e(WةӧѾs',g$%'c9+NnǷYH REFu0%n\?3V;wRP㷃[ ZuVsXxjw=|qM8x-hF 깮DD%\ DDD~~~YL<`BDEWeOOlݼ
Edd$ vZ mCIFF6lڄbܘҩRuO bYcP"!!q>OKA0{b-~;0@ nj$TF$jE\r[RU>{mjǺ :7Qqw=
QQxiؾe *|;?xbbbDL~4` > .]tIbbb0K\~]ĔD _"8?|ksIٲ3o/ʗZ#$gD%6| ɋNa/ZD*oω7xP"b
&DDD\K
&DT q,m][ @ .bfG:ut $< )!}V~/PX9NN>r>.9lo$""""\(g26
s}T#QI-r
9?ck++8*%b"ckk-6bQ} iH k}PY܀96'x> ;Æ I(7|A C<} HHEbB t[BcTZ b&""""*V
l7^ rh%"
^͛5bcc1` ܼuSdchhܩN#ITtf|-zU56byL{.b"chh_,EĤ$ > /]1Ya?vV͚:cG)VDDDDDDDDDDa Q! F=
&DTd2̛;=R[KKKècɓ"&0UT[xB8:: =~ :v+*RB"""""""""c Q!e{ CCCT(b"'H06dIDLh ѣabbcZ "$"""""""""z76bYL\]!DLCDTpƏ _M 5̜ !bgldÆCЮ=$v, L> ;wDLIDDDDDDDDDvl0!""*==DLBDT Y~ կ- `Xv>CR7w.vn߁Ws.z닩ӿAddH ""BJ #'L
~ӊUpbSw^ժamX<888h #hӡ=V]tSeb Q!Si== G LMM_Qn=P,Z'OW M """*DEQ!'cw~(HDr χw.ZJJ
F Ϝ1Yrrr²ƦI\/{`? KQq""B&0ӓ
&DD9%JЯqn=V\ի[6ںFwMؼu 222DLIDDDDDDDDDEL
? c\J.DLCDT4M4 cG jL1 %b'J >@;EKc IDDDDDDDDDEL
??c(F ӦA" xǏs`
"'
ssL4 CO?{1aax(N@"""""""""*r`BDDTk W8DDE0 !fey2S
cŲ_~\]u.]ݼ1秹)!l0!""*DA@-r<={ cFcxH 0b Q!}\tiXZZQlm3g0bhUtv+vj茝p /RB"""""""""*L`BDDT<\ӃlۼK֮\ACf7{ԩj5v
_ov,1[ Q`BDDTdm01 Q\o
1W IDAT8d2cWϞdڱx̝?qES`BDDTH#""B{L#m,?o πx &b2XZZbƴoon4jHg,qQc SX$oADDDW3v+WNDDD%Cbb"F [okkJaZ L|gΞŋJr9zQG@PN {ĎAD%wUDDTPb """ ??c333899P(X&|. "_@f-T,rB4m5o;wb5HLL T*lݶ
1c2iX 3:& J#Q U2$""*=!HDLCDT`Ѳy me\ 7nL|r GУ[wFqqa xw˾",x h4Up_ &""""""xC$DD%!.YkkII6b8_ b
͜ݻDzt =aC1 )!'nCDDTdddcOO\2sssv
bh"'6[SOc%xv?ۻ7s3sEĎAD eӏb """"""b QaJ=LD#H0}4(X PTz$%'k."', Z4o&|m;v` jܸy3 .]tIbcc1/q5SQ^` Q!}\FFF"!" ￛ8`"XJT
5;֩3IH b Q!eOOnCDTxw튅C.at5pA1Y6%KT'O}y2>L
OQvhY,+Lm۱3g}F#b鳖-qh|5n LMMuR7o~6!l0!""YXxǞ \0j'Xz
̵`䯑!bC_SǎH$ڱh̜ z7or
&DDD"9f QUvml\VVV'ObرHKK1Yeog??~ۉ5j }Ǥɓ &RB""""""""" 6O 666"!"Z*nڄRK1dp$&%pZ
oق Qg؉hױ#\TѻHd~+xzr"K%l۲NNN[71` ƊUk >pcG5>>hӾ >ADLIDDDDDDDDDob 6xxrélYl߲*ik~~7p """DLVaİ8rڷmD {of@>qmSQVl0!""QbRB? {xr [6nBմ'3?hp(U
mێW{ )LCddH 56(% W0!"*q:ԭSW[ G bvl݆?̓.thUkV#--MĔDDDDDDDDD%LD}lll
ˋ>֮Z4bbb ܾsGdED"AmC9|ciiiXz5v#%bJ""""""""
&DDD"ϼ R)LDTT_~A֭ 6._1Ybll1FCh:c4w)!QODuwQ^X0o>y{kk5v N>-btX4vlيjUݺ}=3"*JDDDDDDDDD% LDV=d Qq J10h mMTb¤8xx5j ;ӏs`og MvXtSrTA CTj== ՠO6'Zx%"*j`BDD$ cT
w77Q~6d; A 3'$$&bСb+rLMM1nXtK~^'Nh0id7|3uT,bRz->>w\;ĄHLEjj2rYBa.Q U6@䵓=< ZUNNϰKUf85I+M5?/(Ujik|CQ DDD"l0)_ LL$""A/f~C_W,GRR"&M(rl2Xh1 Çڱ7oGء;vvv"&- ^ /W:G@B| csg *VǧͼѤigY&ɓ{x tWA`BA][(*U]+\
NtILHJ˨时2ijrtj.Ews3ll0!""1!"*:k3S3|=uv{7#1) ͘ T*r¢vZ㷝w`?-_ Vfط?N7
~}0j)EF`ӆٸu#(`g
>#$[/Y4 pEfQ
9 ͚w1QlFBEGOKHQVy}\aU}7bf"b?sOOQAhѼ9VXb՟wc7ӠRDLVIRtG AtIehߩ#N:)bʒ_̘%
)Uͽ& 柗/#r$8흯""""*R-\BDDEW0!""Ax!W0!"*ԯ
>0b($$$ :v )y}effBXd N> } 'ND:u͔)`cg^; ֬
*Co̳r]4j?KW
ƍq< 3L{NN)VHHHs7`N-H De$Fe RHSae֚UXy
+6 8 AQ Qزa#̹>j$V fff"',ڜq[ڱ ҹ3Ə 7zպ+)[ @M|D
7ps_Nc_tN .Pѥ\\KU;gt>G813DŎADDDD'H2̌
ۚQԬX*[C'>'TqRDlF#v"
&DDD" ELCDD 6oaC z Uaee%r¢^z.ٻXؗ/ c1b0"-_◥cK*WS}`j$ 7}_~fή 굈
ZZ.<=Ԉ{Q#; QI `qJ
c-쬭ݿ ċ(#Rݼ#ЯN#IRr/]]:?hۼ{D\j`n.J"sј6}pY%DDDDZ#HbMQx7p +%DDTTq"""enqJҥKc-:|vA;?6 9&OݺaEXHH~5
S&bҢŋgt/1F] >RIGx,Z e__xwn3z M{ 9Z֡p8:V)Qw
ĝ!l縺@>SPJ 7")OS[\Vy cHN|++{xU?
QMIMM۴a6mYwADDDzh#Cמ] i'o:X]ZwUW+g$vƂ2iͮFtKN@;=&em̯5ZV550|SB%ed[ZukO*?-|-=S do_%y¬F;-km~OQ{vpn ITXo݈0h}mjzUg o0uiJUi hc7Rc:!-5(}9>L
X@@ !A[ دK /o@ZUxi!ރ]V".. h]8r(F
^={B.QU8nL&GzhƮ?~kyS[aV/;
: 55RRtoBѷ9q|;m Z\\Ν݇sgݣ 댧&eyMLR$7}u??8t \ $Il}vAy{ n=BcF;}lRZ\U?MJȔI-5O#q
h%ϵGʂAݫ,A 珍Zu5 9w=j/y8n̢mI}Z[Z6^zOk ԩM}Uο`4mǴb4S&M3O>e٢jѣm!jmyǕ)͛UN})
jcV"4E ހNg ǎØN+ : ÷ \6MëecC:%k
橢(
" X BzTvnto S s6mbu^=&\u[ߧ
]g^0wmYiֽN]}-'3Hl[5Y;m zV
Q8eh &*aXW.\Pf: [(3c{< C; TUX3什]n*d0YFjd?Y:&J5
Fticn5,0!""""""W_6luz^h7z+4˽\
wZ-Hz t3ؽW >7Ү^EjζILfnOx{܈(
LA@@@(rsӹ~aNF?w\H:b-1n>.DtJ$zݦ rq. WRN672W&!,4̥ Aݪ\;.pyߥJa۪\Zqbcc fc{RmjڗGw;\L* 3q QVz#%R
>,2nݺ.J ,~bvndj䬢Nj]U i H1S';7z%m5uhCTफ+ fGnzcί3ɥovt9ұ ߿@!+2
UFɥu
?x<(USʤv|;\h]ifKk=,0!""""""@D,^z su, 9~
v/UVX{6T8zvqVT; Av3e9."RPκ8 o;tlqN*{ f݁ڷ`L_hحYTNx%A|OBc}U ?)l{ad2[ԯ3F/\lKos10qw֙b*UstU:`K.s6gn_E S/r ztKQ¤@ch
L/lE;}d2ykV >%q9, Rr/4lMBt0>bŅd\&/ כ,Q:$-pogwbKU|GjQz1U\tڝweT"û4=^5[pBкt,>*hsC+.gĉwu\iûlH}*6 _њN=7|v l9b&Gqw'L*),̖vvDY" ڽm ˖APTg_7o?YYxqh
AA}m "n3HOOƾcU(,̩o@MHJ: z>}AP(!Ve"q Lj
E1:}.+\wmOv1\#ǡo{ p*Ղk:|A'%VDDD7˙Lf/JQyj}<.`FT*] hw$._uJ)1dwp%N5+~D
@)kƨnG?Eg4Tov<}cAeN7 nikRm \;u%w0_*[3
tVcZD8SV:֡Qovi(8@ZֲDjN{ &'줦/j9T*ŲKQJKDQDcv1* qI X,X9
w`G K qq xiݘ8 7/_g_6m{ں!!QzhРk.38۷( m8
D\r >7~]]UB7t`2#bm% H'.Ť)+TnW_,p+_""[EyL1,E# `ě<ݨ6rsiy_yX}ߺLVzw '&wOeUM h߄:'4=RյZ7+
)[|T\e~wnpB&yiTf67:V)/ ghT8D,0!""""""N 'ҙ kN
`n x(c
t]kA<]xyyʞT*C^wIӶOR)f ͛w|q8;[t$H1KE|Ю%8$_գ0yp]lMw9g""[I"
Y|]J~f)mDEGKn WɱH% \:τBȧH/Z?`ldÝQmj
00jo
)r3 5kjp_Wds?Aju4q5^tal^gP};<3|_ysM- llZBauK]ߴ> p 7 .bXۛB E*BJ|v<öAWzZs}}Fw9` +- PVV+EEE8v8FUB$ Ӝ 5^vJp=HOO'4uun@Ȱ9DDDDDDDDDDq1`BDDDDFhֵ{ɉ5x{"?鏃^լŅ#>, fOM -1n3,mk^9{Z_bռHӠsܼmlSpT%ʬow|~p:6.?gWqZ Zꢧ/=1y.Eb+`Ӗ-u _vmpDeǒBݫV,[?_݂Q&&&ݫzYՌֹ$+7jC H.@" t;ҹ D"Nу""&J]&*' GFumЧ&@E ߡAƦ: FVt1oӮzC/n USq1)IE@uYvsi^NfF6!^ gO2e?p ii)++o ӧOV~Ȩ(+WbҔ)x駱db4 k++ 4
ҬBP)R)BC /xRDDd""#} #CC#(h <