R/ARS.R

In Then-Terence/LiteXploreR: Concise exploratory data analysis

#' Univariate Adjusted R Squared
#'
#' This function computes the Adjusted R Squared equivalent to that obtained
#' from a linear regression.
#'
#' @param Target A vector of the dependent variable.
#' @param Prediction A vector of the predictions.
#' @export
#' @examples ARS(mtcars$mpg, mtcars$cyl)

ARS <- function(Target, Prediction){
  if(length(Target) != length(Prediction)){
    stop("Target and Prediction must be of the same length.")
  } else{
    NA.Index <- unique(c(which(is.na(Target)), which(is.na(Prediction))))
    Target <- Target[!1:length(Target) %in% NA.Index]
    Prediction <- Prediction[!1:length(Prediction) %in% NA.Index]

    if(is.numeric(Prediction) == T){
      RS <- cor(Target, Prediction)^2
      K <- 1
    } else if(is.factor(Prediction) == T | is.character(Prediction) == T){
      DT1 <- data.table(Y = Target, X = Prediction)
      DT1[, AvgTarget := mean(Y), by = X]
      RS <- cor(DT1[, Y], DT1[, AvgTarget])^2
      K <- length(unique(Prediction)) - 1
    }
    Results <- 1 - (1- RS) * (length(Prediction) - 1) / (length(Prediction) - K - 1)
    return(Results)
  }
}