R/boxatt.R

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

#' Barplots of univariate attribute dispersion within groups
#'
#' This function plots the attribute weights over groups.
#'
#' @param data data matrix with numeric attributes
#' @param imps typically a vector with attribute indices ordered on their importance
#' @param grps list for which each component contains the index numbers (row numbers in the data) of each group in an integer vector.
#' @param main main title of the plot
#' @param ylab label vertical axis
#' @param xlab label horizontal axis
#' @param SCALE logical TRUE/FALSE, whether to scale the data
#' @param colors by default set to \code{palette()}
#' @param ... named arguments to be passed to the default method.
#'
#'
#' @return
#' values of attribute importances in requested order.
#'
#' @examples
#' # boxatt(data = ApoE3, imps = attWild$att[1:100],
#' #  #     grps =  list(1:18, 19:38),
#' #  #     main = "Wildtype (red) vs Transgenic (blue)",
#' #  #     ylab = 'Scaled attribute values',
#' #  #     xlab = '100 most important attributes',
#' #  #     colors = c('red','white','blue'))
#' @seealso
#' \code{\link[rCOSA]{attimp}}
#'
#'
#' @export


boxatt <- function(data, imps, grps,
                   main = "Consider a maintitle here",
                   ylab = "Consider a label here", xlab = "Consider a label here",
                   SCALE = TRUE, colors = palette(), ...) {
  if (SCALE) {
    p <- autoscale(data)
  } else {
    p <- data
  }

  boxplot(p[grps[[1]], imps],
    main = main, xlab = xlab, ylab = ylab,
    outline = FALSE, border = colors[1], col = colors[1],
    names = 1:length(imps), ...
  )

  colMdns <- matrix(NA, ncol = length(imps), nrow = length(grps))
  for (i in 1:length(grps)) {
    colMdns[i, ] <- apply(p[grps[[i]], imps], 2, median, na.rm = TRUE, ...)
    text(colMdns[i, ], col = colors[i + 1], label = "-")
  }
}



# Examples
# boxatt(data = ApoE3, imps = attWild$att[1:100],
#       grps =  list(1:18, 19:38),
#       main = "Wildtype (red) vs Transgenic (blue)",
#       ylab = 'Scaled attribute values',
#       xlab = '100 most important attributes',
#       colors = c('red','white','blue'))