R/plot.PCAmix.R

#' @export
#' @name plot.PCAmix
#' @title  Graphical outputs of PCAmix and PCArot
#' @description Displays the graphical outputs of PCAmix and PCArot. 
#' The individuals (observations), the quantitative variables and the levels 
#' of the qualitative variables are plotted as points using their factor 
#' coordinates (scores). All the variables (quantitative and qualitative)
#'  are plotted as points on the same graph using their squared loadings.
#' @param x an object of class PCAmix obtained with the function \code{PCAmix} or \code{PCArot}.
#' @param axes a length 2 vector specifying the components to plot.
#' @param choice the graph to plot: 
#' \itemize{
#' \item "ind" for the individuals component map,
#'  \item "cor" for the correlation circle if quantitative variables are available in the data,
#'  \item "levels" for the levels components map (if qualitative variables are available in the data),
#'  \item "sqload" for the plot of the squared loadings of all the variables.
#'  }
#' @param label boolean, if FALSE the labels of the points are not plotted.
#' @param coloring.ind a qualitative variable such as a character vector 
#' or a factor of size n (the number of individuals). The individuals 
#' are colored according to the levels of this variable. If NULL, the 
#' individuals are not colored.
#' @param col.ind a vector of colors, of size the number of levels of 
#' \code{coloring.ind}. If NULL, colors are chosen automatically.
#' @param coloring.var boolean, if TRUE, the variables in the plot of the squared loadings 
#' are colored according to their type (quantitative or qualitative).
#' @param lim.cos2.plot a value between 0 and 1. Points with squared 
#' cosinus below this value are not plotted.
#' @param lim.contrib.plot a value between 0 and 100. Points with relative contributions
#' (in percentage) below this value  are not plotted.
#' @param posleg position of the legend.
#' @param xlim a numeric vectors of length 2, giving the x coordinates range. If NULL (by default) 
#' the range is defined automatically (recommended). 
#' @param ylim a numeric vectors of length 2, giving the y coordinates range. 
#' If NULL (by default) the range is defined automatically (recommended).
#' @param main a string corresponding to the title of the graph to draw.
#' @param cex cf. function \code{par} in the \bold{graphics} package
#' @param leg boolean, if TRUE, a legend is displayed.
#' @param cex.leg a numerical value giving the amount by which the legend should be magnified. Default is 0.8.
#' @param \ldots arguments to be passed to methods, such as graphical parameters.
#' @seealso \code{\link{summary.PCAmix}},\code{\link{PCAmix}},\code{\link{PCArot}}
#' @author Marie Chavent \email{marie.chavent@u-bordeaux.fr}, Amaury Labenne
#' @details The observations can be colored according to the levels of a qualitative 
#' variable. The observations, the quantitative variables and the levels can be selected 
#' according to their squared cosine (lim.cos2.plot) or their relative contribution 
#' (lim.contrib.plot) to the component map. Only points with squared cosine or relative 
#' contribution greater than a given threshold are plotted. Note that the relative 
#' contribution of a point to the component map (a plan) is the sum of the absolute 
#' contributions to each dimension, divided by the sum of the corresponding eigenvalues.
#' @references 
#' Chavent M., Kuentz-Simonet V., Labenne A., Saracco J., Multivariate analysis of mixed data: The PCAmixdata R package, arXiv:1411.4911 [stat.CO].
#' @examples
#' data(gironde)
#' base <- gironde$housing[1:20,]
#' X.quanti <-splitmix(base)$X.quanti
#' X.quali <- splitmix(base)$X.quali
#' res<-PCAmix(X.quanti, X.quali, rename.level=TRUE, ndim=3,graph=FALSE)
#' 
#' #----quantitative variables on the correlation circle
#' plot(res,choice="cor",cex=0.8)
#' 
#' #----individuals component map
#' plot(res,choice="ind",cex=0.8)
#' 
#' #----individuals colored with the qualitative variable "houses"
#' houses <- X.quali$houses
#' plot(res,choice="ind",cex=0.6,coloring.ind=houses) 
#' 
#' #----individuals selected according to their cos2
#' plot(res,choice="ind",cex=0.6,lim.cos2.plot=0.8)

#'#----all the variables plotted with the squared loadings
#' plot(res,choice="sqload",cex=0.8)
#' 
#' #----variables colored according to their type (quanti or quali)
#' plot(res,choice="sqload",cex=0.8,coloring.var=TRUE) 
#' 
#' #----levels component map
#' plot(res,choice="levels",cex=0.8)
#' 
#' #----example with supplementary variables
#' data(wine)
#' X.quanti <- splitmix(wine)$X.quanti[,1:5]
#' X.quali <- splitmix(wine)$X.quali[,1,drop=FALSE]
#' X.quanti.sup <-splitmix(wine)$X.quanti[,28:29]
#' X.quali.sup <-splitmix(wine)$X.quali[,2,drop=FALSE]
#' pca<-PCAmix(X.quanti,X.quali,ndim=4,graph=FALSE)
#' pca2 <- supvar(pca,X.quanti.sup,X.quali.sup)
#' plot(pca2,choice="levels")
#' plot(pca2,choice="cor")
#' plot(pca2,choice="sqload")
#' 
plot.PCAmix <- function(x,axes = c(1, 2), choice = "ind",label=TRUE,
                        coloring.ind=NULL,col.ind=NULL, coloring.var=FALSE,
                        lim.cos2.plot=0,lim.contrib.plot=0, posleg="topleft",
                        xlim=NULL,ylim=NULL, cex=1,leg=TRUE,main=NULL,cex.leg=1,...)
{
  cl<-match.call()
  if (!inherits(x, "PCAmix")) 
    stop("use only with \"PCAmix\" objects")
  
  res.pca <-x
  p1 <- res.pca$rec$p1
  p <- res.pca$rec$p
  p2<-res.pca$rec$p2
  m<-nrow(res.pca$levels$coord)
  quanti.coord <- res.pca$quanti$coord
  n<-nrow(res.pca$ind$coord)
  if (is.null(res.pca$sqload.sup)) sup <- FALSE else sup <- TRUE
  
  eig.axes<-res.pca$eig[axes,1]
  
  if (max(axes) > res.pca$ndim) 
    stop(paste("axes must be between 1 and ", res.pca$ndim, sep = ""),call. = FALSE)
  
  if (!(choice %in% c("ind", "sqload", "levels", "cor"))) 
    stop("\"choice\" must be either \"ind\",\"sqload\",\"cor\" or \"levels\"",call. = FALSE)  
  if ((choice=="levels") & is.null(res.pca$levels) & is.null(res.pca$levels.sup))
    stop("\"choice=levels\" is not possible with pure PCA objects",call. = FALSE)
  if ((choice=="cor") & is.null(res.pca$quanti) & is.null(res.pca$quanti.sup))
    stop("\"choice=cor\" is not possible with pure MCA objects",call. = FALSE)
  
  if (lim.cos2.plot != 0 & lim.contrib.plot!=0)
    stop("use either \"lim.cos2.plot\" OR \"lim.contrib.plot\"",call. = FALSE)
  
  if (!is.null(coloring.ind))
  {
    if (choice!="ind")
      warning("use \"coloring.ind\" only if choice=\"ind\"")
  }
  
  if (!is.null(coloring.ind))
  {
    if(!is.factor(coloring.ind) | length(coloring.ind)!=n)
    {
      warning("\"coloring.ind\" must be either NULL or a qualitative variable of length equal to the number of individuals")
      coloring.ind=NULL
    }
      
  }
  if (!is.logical(coloring.var))
  {
    warning("\"coloring.var\" must be either TRUE or FALSE")
    coloring.var=FALSE
  }
  
  
  dim1 <- axes[1]
  dim2 <- axes[2]
  
  lab.x <- paste("Dim ", dim1, " (", signif(res.pca$eig[axes[1],2], 4), " %)", sep = "")
  lab.y <- paste("Dim ", dim2, " (", signif(res.pca$eig[axes[2],2], 4), " %)", sep = "")
  
  # plot of the individuals
  if (choice == "ind") {
    if (is.null(main)) 
      main <- "Individuals component map"
    
    coord.ind<-res.pca$ind$coord
    
    xmin <- min(xlim, coord.ind[, dim1])
    xmax <- max(xlim, coord.ind[, dim1])
    xlim <- c(xmin, xmax) * 1.2
    
    ymin <- min(ylim,coord.ind[, dim2])
    ymax <- max(ylim, coord.ind[, dim2])
    ylim <- c(ymin, ymax) * 1.2
    
    if (is.null(col.ind) | is.null(coloring.ind))
    {
      col.plot.ind <- rep("black",nrow(coord.ind))
    }
    
    if (is.factor(coloring.ind))
    { 
      quali<-coloring.ind
      if (!is.null(col.ind))
      { 
        levels(quali) <- col.ind
        col.plot.ind <- quali
      }
      if(is.null(col.ind))
        col.plot.ind <- as.numeric(quali)
    }
    
    col.plot.ind.total<-col.plot.ind
    
    if(lim.cos2.plot == 0 & lim.contrib.plot==0)
    {
      lim.plot<-0
      select.ind<-1:nrow(coord.ind)
    }
    
    if(lim.cos2.plot != 0 & lim.contrib.plot==0)
    {
      lim.plot <- lim.cos2.plot
      base.lim <- res.pca$ind$cos2[,axes]
      select.ind <- which(apply(base.lim[,],1,sum)>=lim.plot)    
    }
    
    if(lim.cos2.plot == 0 & lim.contrib.plot!=0)
    {
      lim.plot <- lim.contrib.plot
      base.lim <- res.pca$ind$contrib[,axes]
      base.lim <- 100*(base.lim/sum(eig.axes))
      select.ind <- which(apply(base.lim[,],1,sum)>=lim.plot)
    }
    
    if(length(select.ind)==0)
      warning("\"lim.cos.plot\" (or \"lim.contrib.plot\") is too large. No individuals can be plotted")
    
    coord.ind<-coord.ind[select.ind, , drop=FALSE]
    col.plot.ind<-col.plot.ind[select.ind]
    
    graphics::plot(coord.ind[, axes], xlim = xlim, ylim = ylim, xlab = lab.x, 
         ylab = lab.y, pch = 20, col = as.character(col.plot.ind), 
         cex = cex, main=main, ...)
    graphics::abline(h = 0, lty = 2, cex = cex)
    graphics::abline(v = 0, lty = 2, cex = cex)
    
    if(length(select.ind)!=0)
    {
      if(leg==TRUE & is.factor(coloring.ind))
        graphics::legend(posleg, legend =paste(cl["coloring.ind"],levels(coloring.ind),sep="="), text.col = levels(as.factor(col.plot.ind.total)), 
               cex =cex.leg)
      
      if (label) 
        graphics::text(coord.ind[, axes], labels = rownames(coord.ind), 
             pos = 3, col = as.character(col.plot.ind), cex = cex, 
             ...)
    }
  }
  # plot of the variables according to the squared loadings
  if (choice == "sqload") {
    if (is.null(main)) main<-"Squared loadings"
    
    xmax <- max(res.pca$sqload[, dim1],res.pca$sqload.sup[, dim1],xlim)
    xlim <- c(-0.1, xmax * 1.2)

    ymax <- max(res.pca$sqload[, dim2],res.pca$sqload.sup[, dim2],ylim)
    ylim <- c(-0.1, ymax * 1.2)
    
    graphics::plot(0, 0, type = "n", xlab = lab.x, ylab = lab.y, xlim = xlim, 
         ylim = ylim, cex = cex,main=main,...)
    graphics::abline(v = 0, lty = 2, cex = cex)
    graphics::abline(h = 0, lty = 2, cex = cex)
    
    
    if (!(coloring.var))
    {
      for (j in 1:nrow(res.pca$sqload)) 
      {
        graphics::arrows(0, 0, res.pca$sqload[j, dim1], res.pca$sqload[j, dim2], 
               length = 0.1, angle = 15, code = 2, cex = cex,...)
        if (label) 
        {
          if (res.pca$sqload[j, dim1] > res.pca$sqload[j, dim2]) 
          {
            pos <- 4
          }
          else pos <- 3
          graphics::text(res.pca$sqload[j, dim1], res.pca$sqload[j, dim2], labels = rownames(res.pca$sqload)[j], 
               pos = pos, cex = cex,...)
        }
      }
    } else {
      for (j in 1:nrow(res.pca$sqload)) 
      {
        col.sq<-rep(c("black","red"),c(p1,p2))
        graphics::arrows(0, 0, res.pca$sqload[j, dim1], res.pca$sqload[j, dim2], 
               length = 0.1, angle = 15, code = 2, cex = cex, col=col.sq[j],...)
        if (label) 
        {
          if (res.pca$sqload[j, dim1] > res.pca$sqload[j, dim2]) {
            pos <- 4
          }
          else pos <- 3
          graphics::text(res.pca$sqload[j, dim1], res.pca$sqload[j, dim2], labels = rownames(res.pca$sqload)[j], 
               pos = pos, cex = cex, col=col.sq[j], ...)
        }
      }
      if (leg==TRUE)
        graphics::legend(posleg, legend = c("numerical","categorical"), text.col = c("black","red"), 
               cex = cex.leg)
    }
    if (sup)
    {
      for (j in 1:nrow(res.pca$sqload.sup)) 
      {
        graphics::arrows(0, 0, res.pca$sqload.sup[j, dim1], res.pca$sqload.sup[j, dim2], 
               length = 0.1, angle = 15, code = 2, lty=5, col="blue",cex = cex,...)
        if (label) 
        {
          if (res.pca$sqload.sup[j, dim1] > res.pca$sqload.sup[j, dim2]) 
          {
            pos <- 4
          } else pos <- 3
          graphics::text(res.pca$sqload.sup[j, dim1], res.pca$sqload.sup[j, dim2], labels = rownames(res.pca$sqload.sup)[j], 
               pos = pos, cex = cex,col="blue",...)
        }
      }
    }
  }
  #plot of the levels
  if (choice == "levels") {
    if (is.null(main)) 
      main <- "Levels component map"
    
    xmin <- min(xlim,res.pca$levels$coord[, dim1],res.pca$levels.sup$coord[, dim1])
    xmax <- max(xlim,res.pca$levels$coord[, dim1],res.pca$levels.sup$coord[, dim1])
    xlim <- c(xmin, xmax) * 1.2
    
    ymin <- min(ylim,res.pca$levels$coord[, dim2],res.pca$levels.sup$coord[, dim2])
    ymax <- max(ylim,res.pca$levels$coord[, dim2],res.pca$levels.sup$coord[, dim2])
    ylim <- c(ymin, ymax) * 1.2
    
    graphics::plot(0,0, xlim = xlim, ylim = ylim,
         xlab = lab.x, ylab = lab.y, type="n", cex = cex,main=main, ...)
    graphics::abline(h = 0, lty = 2, cex = cex)
    graphics::abline(v = 0, lty = 2, cex = cex)
    
    #plot levels of active variables
    if (!is.null(res.pca$levels))
    {
      if (lim.cos2.plot == 0 & lim.contrib.plot==0)
      {
        lim.plot<-0
        base.lim<-res.pca$levels$cos2[,axes]
      }
      
      if (lim.cos2.plot != 0)
      {
        lim.plot<-lim.cos2.plot
        base.lim<-res.pca$levels$cos2[,axes]
      }
      
      if (lim.contrib.plot!=0)
      {
        lim.plot<-lim.contrib.plot
        base.lim<-res.pca$levels$contrib[,axes]
        base.lim<-100*(base.lim/sum(eig.axes))    
      }
      
      coord.lev <- res.pca$levels$coord[, axes, drop = FALSE]
      
      test.empty.plot<-c()
      for (v in 1:nrow(coord.lev)) 
      {
        if (sum(base.lim[v, ], na.rm = TRUE) >= lim.plot && !is.na(sum(base.lim[v, ], na.rm = TRUE))) {
          test.empty.plot<-c(test.empty.plot,1)
          graphics::points(coord.lev[v, 1], coord.lev[v,2], pch=20,cex = cex,...)
          
          if (label) 
          {
            if (abs(coord.lev[v, 1]) > abs(coord.lev[v,2])) 
            {
              if (coord.lev[v, 1] >= 0) 
                pos <- 4
              else pos <- 2
            }
            else {
              if (coord.lev[v, 2] >= 0) 
                pos <- 3
              else pos <- 1
            }
            graphics::text(coord.lev[v, 1], y = coord.lev[v, 2], 
                 labels = rownames(coord.lev)[v], pos = pos, cex = cex)
          }
        }
      }
      if(is.null(test.empty.plot)){
        warning("\"lim.cos.plot\" (or \"lim.contrib.plot\") is too large. No level can be plotted")
        return()
      }
    }
    #plot levels of supplementary variables
    if (!is.null(res.pca$levels.sup))
    {
      coord.lev.sup <- res.pca$levels.sup$coord[, axes, drop = FALSE]
      for (v in 1:nrow(coord.lev.sup)) 
      {
        
        graphics::points(coord.lev.sup[v, 1], coord.lev.sup[v,2], pch=18,cex = cex,
               col="blue",...)
        
        if (label) 
        {
          if (abs(coord.lev.sup[v, 1]) > abs(coord.lev.sup[v,2])) 
          {
            if (coord.lev.sup[v, 1] >= 0) 
              pos <- 4
            else pos <- 2
          }
          else {
            if (coord.lev.sup[v, 2] >= 0) 
              pos <- 3
            else pos <- 1
          }
          graphics::text(coord.lev.sup[v, 1], y = coord.lev.sup[v, 2], 
               labels = rownames(coord.lev.sup)[v], pos = pos, 
               col="blue", cex = cex)
        }
      }
    }
  }  
  # plot of correlation circle
  if (choice == "cor") {
    if (is.null(main)) 
      main <- "Correlation circle"
    if (is.null(xlim)) xlim = c(-1.1, 1.1)
    if (is.null(ylim)) ylim = c(-1.1, 1.1)
    
    graphics::plot(0, 0, main = main, xlab = lab.x, ylab = lab.y, 
         xlim = xlim, ylim = ylim, col = "white", 
         asp = 1, cex = cex,...)
    x.cercle <- seq(-1, 1, by = 0.01)
    y.cercle <- sqrt(1 - x.cercle^2)
    graphics::lines(x.cercle, y = y.cercle)
    graphics::lines(x.cercle, y = -y.cercle)
    graphics::abline(v = 0, lty = 2, cex = cex)
    graphics::abline(h = 0, lty = 2, cex = cex)
    
    #plot active quantitative variables
    if (!is.null(res.pca$quanti))
    {
      if (lim.cos2.plot == 0 & lim.contrib.plot==0)
      {
        lim.plot<-0
        base.lim<-res.pca$quanti$cos2[,axes]
      }
      
      if (lim.cos2.plot != 0)
      {
        lim.plot<-lim.cos2.plot
        base.lim<-res.pca$quanti$cos2[,axes]
      }
      
      if(lim.contrib.plot!=0)
      {
        lim.plot<-lim.contrib.plot
        base.lim<-res.pca$quanti$contrib[,axes]
        base.lim<-100*(base.lim/sum(eig.axes))     
      }
      
      coord.var <- res.pca$quanti$coord[, axes, drop = FALSE]
      test.empty.plot<-c()      
      for (v in 1:nrow(coord.var)) 
      {
        if (sum(base.lim[v, ] , na.rm = TRUE) >= lim.plot && !is.na(sum(base.lim[v, ], na.rm = TRUE))) {
          test.empty.plot<-c(test.empty.plot,1)
          graphics::arrows(0, 0, coord.var[v, 1], coord.var[v,2], length = 0.1, angle = 15, code = 2,cex = cex)
          
          if (label) 
          {
            if (abs(coord.var[v, 1]) > abs(coord.var[v, 2])) 
            {
              if (coord.var[v, 1] >= 0) 
                pos <- 4
              else pos <- 2
            }
            else {
              if (coord.var[v, 2] >= 0) 
                pos <- 3
              else pos <- 1
            }
            graphics::text(coord.var[v, 1], y = coord.var[v, 2], 
                 labels = rownames(coord.var)[v], pos = pos, cex = cex)
          }
        }
      }
      if(is.null(test.empty.plot))
        warning("\"lim.cos.plot\" (or \"lim.contrib.plot\") is too large. No variable can be plotted")
    }
    #plot supplementary quantitative variables
    if (!is.null(res.pca$quanti.sup))
    {
      coord.var.sup <- res.pca$quanti.sup$coord[, axes, drop = FALSE]
      for (v in 1:nrow(coord.var.sup)) 
      {
        graphics::arrows(0, 0, coord.var.sup[v, 1], coord.var.sup[v,2], length = 0.1, 
               angle = 15, code = 2,cex = cex,col="blue",lty=5)
        
        if (label) 
        {
          if (abs(coord.var.sup[v, 1]) > abs(coord.var.sup[v,2])) 
          {
            if (coord.var.sup[v, 1] >= 0) 
              pos <- 4
            else pos <- 2
          }
          else {
            if (coord.var.sup[v, 2] >= 0) 
              pos <- 3
            else pos <- 1
          }
          graphics::text(coord.var.sup[v, 1], y = coord.var.sup[v, 2], 
               labels = rownames(coord.var.sup)[v], pos = pos, cex = cex,col="blue")
        }
      }
    }
  }
}

Try the PCAmixdata package in your browser

Any scripts or data that you put into this service are public.

PCAmixdata documentation built on May 2, 2019, 12:38 p.m.