R/qq_line_coeffs.R

In GegznaV/BioStat: Routines for Basic (Bio)Statistics

#' Compute a slope and an intercept for a reference line in a qq-plot
#'
#' Compute a slope and an intercept for a reference line ("qq-line")
#' in a quantile comparison plot given a vector of data an a theoretical
#' distribution.
#'
#' @details
#' The code is based on code of `qqline()` in package \pkg{stats}.
#' But instead of plotting a reference line, it returns its coefficients.
#'
#' @param y (numeric) A numeric vector.
#'
#' @inheritParams stats::qqline
#' @inheritParams stats::quantile
#' @param ... Further arguments for `distribution`.
#'
#' @return A vector with a `slope` and an `intercept` for a qq-line.
#' @export
#' @keywords internal
#' @examples
#'
#' set.seed(254)
#' x <- rnorm(50)
#'
#' qq_line_coeffs(x)
qq_line_coeffs <- function(y,
                           datax = FALSE,
                           distribution = qnorm,
                           probs = c(0.25, 0.75),
                           qtype = 7,
                           na.rm = TRUE,
                           ...) {
  checkmate::assert_numeric(probs, lower = 0, upper = 1, len = 2)
  checkmate::assert_function(distribution)

  y <- stats::quantile(y,
    probs,
    names = FALSE,
    type = qtype,
    na.rm = na.rm
  )

  x <- distribution(probs, ...)

  if (datax) {
    slope <- diff(x) / diff(y)
    intercept <- x[1L] - slope * y[1L]
  } else {
    slope <- diff(y) / diff(x)
    intercept <- y[1L] - slope * x[1L]
  }
  # Output
  c(intercept = intercept, slope = slope)
}
# =============================================================================