R/groupBycoord.R

In CAinterprTools: Graphical Aid in Correspondence Analysis Interpretation and Significance Testings

#'Define groups of categories on the basis of a selected partition into k groups
#'employing the Jenks' natural break method on the selected dimension's
#'coordinates
#'
#'The function allows to group the row/column categories into k user-defined
#'partitions.
#'
#'K groups are created employing the Jenks' natural break method applied on the
#'selected dimension's coordinates. A dot chart is returned representing the
#'categories grouped into the selected partitions. At the bottom of the chart,
#'the Goodness of Fit statistic is also reported. The function also returns a
#'dataframe storing the categories' coordinates on the selected dimension and
#'the group each category belongs to.
#'
#'@param data Name of the dataset (must be in dataframe format).
#'@param x Dimension whose coordinates are used to build the partitions.
#'@param k Number of groups.
#'@param which Speficy if rows ("rows"; default) or columns ("cols") must be
#'  grouped.
#'@param cex.labls Set the size of the labels of the dot chart (0.75 by default).
#'
#'@keywords groupBycoord
#'
#'@export
#'
#' @examples
#' data(greenacre_data)
#'
#' #divide the row categories into 3 groups on the basis of the coordinates
#' #of the 1st dimension, and store the result into a 'res' object
#' res <- groupBycoord(greenacre_data, x=1, k=3, which="rows")
#'
#' @seealso \code{\link{caCluster}}
#'
groupBycoord <- function (data, x=1, k=3, which="rows", cex.labls=0.75){
  
  categ=NULL
  
  res <- CA(data, graph=FALSE)
  ifelse(which=="rows", 
         dtf <- data.frame(categ=row.names(res$row$coord), coord.x=res$row$coord[,x]), 
         dtf <- data.frame(categ=row.names(res$col$coord), coord.x=res$col$coord[,x]))
  dtf <- dtf[order(dtf$coord.x),]
  Jclassif <- classInt::classIntervals(dtf$coord.x, k, style = "jenks")
  GoFtest <- jenks.tests(Jclassif) 
  dtf$group <- as.factor(cut(dtf$coord.x, unique(Jclassif$brks), labels=FALSE, include.lowest=TRUE))
  dotchart2(dtf$coord.x, 
            labels=dtf$categ, 
            groups=as.factor(paste0("group ", dtf$group)), 
            lty=2, 
            cex.labels=cex.labls,
            xlab=paste0("coordinate on the ", x, " dim."),
            main=paste0(ifelse(which=="rows", "Row", "Column"), " categories clustered into ", k, " groups (Jenks' natural breaks on the coord. of the selected dim.)"),
            cex.main=0.90,
            sub=paste0("Goodness of Fit: ", round(GoFtest[2],2)),
            cex.sub=0.7)
  colnames(dtf)[2] <- paste0("coord.", x,".Dim")
  return(subset(dtf, , -c(categ)))
}