R/minmax_mat.R

In visTree: Visualization of Subgroups for Decision Trees

#' Minmax matrix
#'
#' Identifies splits and relevant criteria
#' @param Y Response variable in the dataset
#' @param str Structure of pathway from the root node in the decision tree to each terminal node
#' @param varnms Names of covariates
#' @param interval logical. Continuous response (interval = FALSE) and Categorical response (interval = TRUE).
#' @keywords visualization pathway decision tree
#' @export
#'

minmax_mat <- function(str, varnms, Y, interval) {
  comps <- strsplit(str, ",")
  MMM <- matrix(data = rep(c(-Inf, Inf, " "), length(varnms)), nrow = length(varnms), ncol = 3, byrow = TRUE) ### min-max matrix
  rownames(MMM) <- varnms
  # nodeids(potentialtree)
  Mlist <- list()
  # Y<-fitted(cond.tree)[[3]]
  if (is.numeric(Y)) {
    length.mat <- length(comps[[1]]) - 2
  }
  if (is.factor(Y)) {
    if (interval == TRUE) {
      length.mat <- length(comps[[1]]) - 3
    } else {
      length.mat <- length(comps[[1]]) - 2
    }
  }
  for (i in 1:length.mat) {
    nodestr <- strsplit(trim(comps[[1]][i]), " ")
    node.varnm <- trim(nodestr[[1]][1])
    node.dir <- trim(nodestr[[1]][2])
    node.split <- trim(nodestr[[1]][3])
    var.row <- which(varnms == node.varnm)
    if (node.dir == "<=") {
      MMM[var.row, 2] <- as.numeric(node.split)
      MMM[var.row, 3] <- "<="
    } else {
      MMM[var.row, 1] <- as.numeric(node.split)
      MMM[var.row, 3] <- ">"
    }
    Mlist[[i]] <- c(node.varnm, MMM[var.row, ])
    output <- matrix(unlist(Mlist), ncol = 4, byrow = TRUE)
  }
  if (is.factor(Y)) {
    y <- paste0(comps[[1]][length(comps[[1]]) - 2], ",", comps[[1]][length(comps[[1]]) - 1])
  }
  if (is.numeric(Y)) {
    y <- comps[[1]][length(comps[[1]]) - 1]
  }
  return(list(M = MMM, y = y))
}