R/cmds.R

In mkampert/rCOSA: Clustering Objects on Subsets of Attributes (COSA)

#' Classical (Metric) Multidimensional Scaling
#'
#' Classical metric multidimensional scaling which is the same as the cmdscale in the stats package. Which follows the analyis of Mardia (1978)
#'
#' @param D dissimilarities/distances of class dist.
#' @param groupnr groupnr when known, each object can be given a number to which group it belongs
#' @param colv vector containing color names for each group number
#' @param main title of the plot, by default there is no title.
#' @param PLOT whether to plot or not to plot, by default TRUE.
#' @param ... arguments that can be parsed to \code{\link[stats]{cmdscale}} procedures.
#'
#' @return
#' similar output as \code{\link[stats]{cmdscale}}
#'
#' @references
#' Mardia, K.V. (1978).  Some properties of classical multidimensional scaling.  Communications on Statistics - Theory and Methods, A7, 1233-41. \cr doi: \url{https://doi.org/10.1080/03610927808827707}
#'
#' @export

cmds <- function(D, groupnr = NULL, main = NULL, colv = palette()[c(8, 2, 4, 3, 5, 6, 7, 1)], PLOT = TRUE, ...) {
  fit <- cmdscale(D, eig = TRUE)
  x <- fit$points[, 1]
  y <- fit$points[, 2]
  if (PLOT) {
    if (is.null(groupnr)) {
      cols <- 1
    } else {
      cols <- colv[groupnr + 1]
    }
    lims <- c(floor(100 * min(c(x), min(y))) / 100, ceiling(100 *
      max(c(max(x), max(y)))) / 100)
    plot(x, y,
      xlab = "Dimension 1", ylab = "Dimension 2",
      main = main, col = cols, xlim = lims, ylim = lims,
      xaxs = "i", yaxs = "i", pch = 20
    )
  }
  invisible(fit)
}