R/getKminmax.R

In arianeschad/mbqn: Mean/Median-balanced quantile normalization

#' Get the k largest/smallest elements
#'
#' @description Extract the k largest or smallest values and their indices for
#' each column of a matrix.
#' @param x a data matrix or data frame.
#' @param k an integer specifying the number of extreme values. Must be
#' \code{<= nrows(x)}.
#' @param flag use "min" or "max" (default) to select smallest or largest
#' elements.
#' @details Order the values of each column of \code{x} and determine the
#' k smallest (\code{flag = "min"}) or largest (\code{flag = "max"}) values and
#' their indices. NA's in the data are ignored.
#' @return List with elements:
#' \item{\code{ik}}{indices of ordered extreme values}
#' \item{\code{minmax}}{ordered extreme values.}
# #' @concept quantile, quantile normalization, rank invariance
#' @references Brombacher, E., Schad, A., Kreutz, C. (2020). 
#' Tail-Robust Quantile Normalization. BioRxiv.
#' @examples
#' # Create a data matrix
#' x <- matrix(c(5,2,3,NA,4,1,4,2,3,4,6,NA,1,3,1),ncol=3)
#' # Get indices of the 5 largest values in each column
#' getKminmax(x, k = 5, "max") 
#' @author Ariane Schad
#  Aug. 2017
#' @export getKminmax
getKminmax <- function(x,k,flag = "max"){

    if(flag == "min"){
        decreasing <- FALSE
    }else if(flag=="max"){
        decreasing <- TRUE
    }

    ## sort ##
    na.last <- TRUE
    ik <- apply(x,2,order, decreasing = decreasing, na.last = na.last )
    minmax <- apply(x,2,function(y)y[order(y, decreasing = decreasing,
                    na.last = na.last )])
    ## select the first k elements ##
    ik <- ik[seq_len(k),]
    minmax <- minmax[seq_len(k),]

    ## return results
    return(list(ik = ik, minmax = minmax))
}