R/showdata.R

In CorReg: Linear Regression Based on Linear Structure Between Variables

#' To show the missing values of a dataset
#' 
#' @description Plot the dataset with marks where there are missing value. It allows to have a quick idea of the structure of missing values (Missing at Random or not for example).
#' @param X the matrix to analyse (matrix with missing values or correlations matrix)
#' @param what indicates what to plot. If \code{what="correl"} and \code{X} is a correlation matrix then the plot is a correlation plot. Else it shows the missing values positions in the dataset.
#' @param pch for missing, symbol to plot (can set pch="." for large datasets)
#'
#' @examples
#' data <- mtcars
#' datamiss = Terminator(target = data, wrath = 0.05) # 5% of missing values
#' showdata(datamiss) # plot positions of the missing values
#' 
#' # missing values with a structure
#' datamiss = Terminator(target = data, diag = 1) # diag of missing values
#' showdata(datamiss) # plot positions of the missing values (no full individuals, no full variable)
#' 
#' opar = par(no.readonly = TRUE)
#' showdata(X = cor(data), what = "correl")
#' par(opar)
#'
#' @export
showdata <- function(X = X, what = c("miss", "correl"), pch = 7){
   what = match.arg(what)
   
   if(what == "miss"){
      M = which(is.na(X), arr.ind = TRUE)
      if(nrow(M) > 1){
         plot(M[, c(2, 1)], pch = pch)
         title("Missing values in the dataset")  
      }else{
         message("No missing values")
      }
   }else if (what == "correl"){
      corrplot(corr = X, addrect = NULL, is.corr = TRUE, method = "color", tl.pos = "n", diag = FALSE, outline = FALSE)
   }else{
      correl = cor(X[, !is.na(colSums(X)) & apply(X, 2, sd)!= 0])
      corrplot(corr = correl, addrect = NULL, is.corr = TRUE, method = "color", tl.pos = "n", diag = FALSE, outline = FALSE)
   }
}