R/make.distribution.R

In qmethod: Analysis of Subjective Perspectives Using Q Methodology

make.distribution <- function(nstat, max.bin = 5) {

  # Input validation
  if (!is.vector(nstat)) {
    stop("nstat is not a vector.")
  }
  if (!is.vector(max.bin)) {
    stop("max.bin is not a vector.")
  }
  if (!(mode(nstat) == "numeric" && nstat > 0 && nstat %% 1 == 0)) {
    stop("The nstat specified is not a positive integer.")
  }
  if (!(mode(max.bin) == "numeric" && max.bin > 0 && max.bin %% 1 == 0)) {
    stop("The max.bin specified is not a positive integer.")
  }
  if ((2 * max.bin + 1) > nstat) {
    stop("The nstat specified is too small for the chosen max.bin.")
  }

  nbins <- (2 * max.bin) + 1  # make sure that nbins is uneven
  # this +1 ensures that there will always be a 0 bin.
  range <- qnorm(1 / nstat)
  # what's the cutoff for 1/n?
  # This the farthest point under the normal distribution where you'd still expect a card, given nstat
  distribution <- dnorm(seq((-range), range, length = nbins)) / sum(dnorm(seq(-range, range, length = nbins))) * nstat
  distribution <- round(distribution)
  if (sum(distribution) != nstat) {
    warning(
      paste(
        "Could not fit",
        nstat,
        "items neatly under a standard normal distribution.",
        "You need",
        sum(distribution),
        "items for good fit instead.",
        "Accept that or try again with a different max.bin."
      )
    )
  }
  return(distribution)
}