R/ntester.R

In PASWR2: Probability and Statistics with R, Second Edition

#' @title Normality Tester
#' 
#' @description Q-Q plots of randomly generated normal data of the same sample size as the tested data are generated and plotted on the perimeter of the graph while a Q-Q plot of the actual data is depicted in the center of the graph.
#' 
#' @details Q-Q plots of randomly generated normal data of the same size as the tested data are generated and plotted on the perimeter of the graph sheet while a Q-Q plot of the actual data is depicted in the center of the graph. The p-values are calculated based on the Shapiro-Wilk W-statistic. Function will only work on numeric vectors containing less than or equal to 5000 observations.  Best used for moderate sized samples (n < 50).
#' 
#' @param actual.data is a numeric vector. Missing and infinite values are allowed, but are ignored in the calculation. The length of \code{actual.data} must be less than 5000 after dropping nonfinite values. 
#' 
#' @author Alan T. Arnholt <arnholtat@@appstate.edu> 
#' 
#' @export
#' 
#' @references 
#' Shapiro, S.S. and Wilk, M.B. 1965. \emph{An analysis of variance test for normality (complete samples)}. Biometrika \bold{52}: 591-611.
#' 
#' @examples
#' ntester(actual.data = rexp(40, 1)) 
#' # Q-Q plot of random exponential data in center plot 
#' # surrounded by 8 Q-Q plots of randomly generated  
#' # standard normal data of size 40.
#' 
#' @keywords hplot
##############################################################
ntester <- function(actual.data)
{
  opar <- par(no.readonly = TRUE)
  Ared <- "#C00000"
  Ablue <- "#0080FF"
  par(mfrow = c(3, 3))
  par(oma = c(1, 0, 2, 0))
  par(mar = c(0, 0, 2, 0))
  par(pty = "s")
  for(i in 1:4) {
    SimData <- rnorm(length(actual.data))
    s <- shapiro.test(SimData)
    qqnorm(SimData, xlab = "", ylab = "", axes = FALSE, col = Ablue, pch = 19, cex = .5, main=paste("SimNorm p-val = ", round(s$p.value, 3)), col.main = Ablue)
    box()
    qqline(SimData, col = Ared)
  }
  sx <- shapiro.test(actual.data)
  qqnorm(actual.data, xlab = "", ylab = "", axes =FALSE, col = Ared, pch = 19, cex = 0.5, main = paste("YourData p-val = ", round(sx$p.value, 3)), col.main = Ared)
  box()
  qqline(actual.data, col = Ablue)
  for(i in 6:9) {
    SimData <- rnorm(length(actual.data))
    s <- shapiro.test(SimData)
    qqnorm(SimData, xlab = "", ylab = "", axes = FALSE, col = Ablue, pch = 19, cex = 0.5, main= paste("SimNorm p-val = ", round(s$p.value, 3)), col.main = Ablue)
    box()
    qqline(SimData, col = Ared)
  }
  mtext("Simulated Normal Data on Perimeter - Actual Data in Center",
        side = 3, outer = TRUE, cex = 1.5, col = Ared)
  par(oma = c(0, 0, 0, 0))
  on.exit(par(opar))
}