R/PlotClustersBiplot.R

In PERMANOVA: Multivariate Analysis of Variance Based on Distances and Permutations

PlotClustersBiplot <- function (A, Groups = Ones(c(nrow(A), 1)), TypeClus = "st", ClusterColors = NULL,
          ClusterNames = NULL, centers = TRUE, ClustConf = 1, Legend = FALSE,
          LegendPos = "topright", CexClustCenters = 1, ...)
{
  TypeClusters = c("ch", "el", "st", "oc")
  if (is.numeric(TypeClus))
    TypeClus = TypeClusters[TypeClus]
  g = length(levels(Groups))
  if (is.null(ClusterColors)) {
    if (g > 1) {
      palette(rainbow(g))
      ClusterColors = palette()
    }
    else ClusterColors = "blue"
  }
  if (is.null(ClusterNames)) {
    ClusterNames = levels(Groups)
  }
  levellab = levels(Groups)
  Sizes = Zeros(c(g, 1))
  Means = Zeros(c(g, 2))
  X = list()
  for (i in 1:g) {
    X[[i]] = A[which(Groups == levellab[i]), ]
    Sizes[i] = length(which(Groups == levellab[i]))
    if (is.matrix(X[[i]]))
      Means[i, ] = apply(X[[i]], 2, mean)
    else Means[i, ] = X[[i]]
  }
  ColorInd = matrix(0, dim(A)[1], 1)
  for (i in 1:(dim(A)[1])) ColorInd[i] = ClusterColors[as.integer(Groups[i])]
  if (centers) {
    Centers = matrix(0, g, 2)
    Centers[, 1] = tapply(A[, 1], Groups, mean)
    Centers[, 2] = tapply(A[, 2], Groups, mean)
    points(Centers[, 1], Centers[, 2], cex = 1.5, pch = 16,
           col = ClusterColors)
    text(Centers[, 1], Centers[, 2], labels = ClusterNames,
         col = ClusterColors, cex = CexClustCenters, pos = 4)
  }
  for (i in 1:g) if (Sizes[i] > 2) {
    if (TypeClus == "el") {
      elip = ConcEllipse(X[[i]], confidence = ClustConf)
      plot(elip, col = ClusterColors[i])
    }
    else {
      fr = Fractions(X[[i]], confidence = ClustConf)
      plot(fr, type = TypeClus, col = ClusterColors[i])
    }
  }
  LegendPositions = c("bottomright", "bottom", "bottomleft",
                      "left", "topleft", "top", "topright", "right", "center")
  if (is.numeric(LegendPos))
    LegendPos = LegendPositions[LegendPos]
  if (Legend) {
    legend(x = LegendPos, ClusterNames, col = ClusterColors,
           text.col = "black", pch = 16, cex = 0.7)
  }
  return(ColorInd)
}