R/create_rank_table.R

In jasongraf1/VADIS: Variation-Based Distance & Similarity Modeling

#' Create Table of Variable Importance Rankings for VADIS analysis
#'
#' @param mod_list A list of random forest model objects. Currently supports objects of class \code{\link[party]{RandomForest-class}}, \code{\link[ranger]{ranger}}, and \code{\link[randomForest]{randomForest}}
#' @param path Path in which to save the output (as \code{.csv} file). If \code{NULL}, defaults to the current working directory. Set \code{path = FALSE} if you do not wish to save to file.
#' @param conditional logical. Should conditional (default) or unconditional permutation variable importance be computed. Only applies to \code{RandomForest-class} models from the \code{\link[party]{party}} package.
#'
#' @author Jason Grafmiller
#'
#' @details The function loops through a list of random forest objects, extracts the variable importance estimates, and compiles them in a single dataframe. For \code{\link[party]{RandomForest-class}} models, function uses the \code{\link[permimp]{permimp}} function (Debeer & Strobl 2020) with the default threshold (0.95) and \code{AUC= TRUE} (Janitza et al. 2013).
#' For forests fit with \code{\link[ranger]{ranger}} or \code{\link[randomForest]{randomForest}}, the \code{importance} argument must be specified.
#' #'
#' @return A dataframe
#'
#' @references
#' Debeer, Dries & Carolin Strobl. 2020. Conditional permutation importance revisited. *BMC Bioinformatics* 21(1). 307. https://doi.org/10.1186/s12859-020-03622-2.
#'
#' Janitza, Silke, Carolin Strobl & Anne-Laure Boulesteix. 2013. An AUC-based permutation variable importance measure for random forests. *BMC Bioinformatics* 14(1). 119. https://doi.org/10.1186/1471-2105-14-119.
#'
#' @export
#'
#' @examples
#' \dontrun{
#' fmla <- Type ~ PossrAnimacyBin + PossrWordC + PossmWordC + FinalSibilant +
#'   TypeTokenRatio + ProtoSemanticRelation + PossrExpType
#'
#' rf_fnc <- function(x) ranger::ranger(fmla, data = x, importance = "permutation")
#'
#' rf_list <- lapply(data_list, rf_func)
#'
#' create_rank_table(rf_list, path = FALSE)
#' }
create_rank_table <- function(mod_list, conditional = TRUE) {
  # identify the class of models
  type <- class(mod_list[[1]])
  if (type[1] == "ranger"){
    varimp_tab <- as.data.frame(do.call("cbind", lapply(mod_list, FUN = function(m) ranger::importance(m))))
  } else if (type[1] == "RandomForest"){
    if (conditional){
      message(paste("Computing conditional variable importance for", length(mod_list), "models. This may take some time...\nIf it takes too long, consider setting `conditional = FALSE`."))
    } else message(paste("Computing variable importance for", length(mod_list), "models. This may take some time...\n"))
    varimp_list <- lapply(
      mod_list,
      FUN = function(m) {
        permimp::permimp(m, conditional = conditional, AUC = TRUE,
                         progressBar = FALSE)
        })
    varimp_tab <- as.data.frame(
      do.call("cbind", lapply(varimp_list, function(x) x$values))
    )
  } else if (type == "randomForest"){
    varimp_list <- lapply(mod_list, FUN = function(m) randomForest::importance(m))
    varimp_tab <- data.frame(sapply(varimp_list, c))
    names(varimp_tab) <- names(mod_list)
  } else stop("I don't recognize this class of random forest model...")
  return (varimp_tab)
}