R/TDAGroupMedian.R

In mrdwab/TDAtools: A collection of functions written for students at the Tata-Dhan Academy

#'Estimate the sample median of grouped data
#'
#'Often, we are provided with summary frequency tables from which we need to perform some calculations. This function estimates the \code{\link{median}} of grouped data, entered as a \code{data.frame}.
#'
#'@param inDF The input \code{data.frame}. This \code{data.frame} \emph{must}
#'include the following as the first three columns (see \emph{Examples}):
#'\itemize{ \item The lower boundaries of the class \item The upper boundaries
#'of the class \item The frequency for that class }
#'@note It is also possible to call this function directly on the stored output
#'of the \code{\link{TDAFreq}} function, though it would make more sense to
#'calculate the \code{\link{median}} from the raw data if that
#'is available
#'@author Ananda Mahto
#'@seealso \code{\link{median}}, \code{\link{TDAGroupMean}}
#'@examples
#'
#'## Your input data.frame should look like this
#'
#'MyGroupedDF <- data.frame(
#' ClassLow = c(0, 50, 100, 150, 200, 250, 300, 350, 400, 450),
#' ClassHigh = c(50, 100, 150, 200, 250, 300, 350, 400, 450, 500),
#' Freq = c(78, 123, 187, 82, 51, 47, 13, 9, 6, 4))
#'MyGroupedDF
#'TDAGroupMedian(MyGroupedDF)
#'
#'set.seed(1)
#'x <- sample(300, 100, replace = TRUE)
#'median(x)
#'temp <- TDAFreq(x, 10)
#'
#'## Now, imagine that we have deleted x, but still want the median
#'##   Not exactly the same, but close...
#'TDAGroupMedian(temp)
#'
#'\dontshow{rm(x, temp, MyGroupedDF)}
#'
TDAGroupMedian <- function(inDF) {
  n <- sum(inDF$Freq)
  T1 <- cumsum(inDF$Freq) - (n/2)
  T1 <- sum(T1 <= 0)
  FSum <- sum(inDF$Freq[1:T1])
  MedianClass <- inDF$Freq[T1+1]
  
  if (isTRUE("TDAFreq" %in% class(inDF))) {
    MyClasses <- gsub("\\[|\\]|\\(|\\)", "", inDF$Class)
    temp <- sapply(strsplit(MyClasses, ","), as.numeric)
    X1 <- temp[1, T1+1]
    X2 <- temp[2, T1+1]
  } else {
    X1 <- inDF[T1+1, 1]
    X2 <- inDF[T1+1, 2]
  }
  
  ClassWidth <- X2 - X1
  AA <- ((n + 1)/2 - (FSum + 1))/MedianClass
  AB <- AA * ClassWidth
  AC <- AB + X1
  AC
}