Add worm-plot?

[mattansb]

Might consider adding worm-plots - an upgraded qqplot with CI boudries (from https://doi.org/10.1002/sim.746).

Example code (from https://gist.github.com/mattansb/ba2f1b9dbd9b4479ae31826a6cc32a80) that also works with non-normal:

qq_worm_plot <- function(x, distribution = "norm", ...) {
  d <- match.fun(paste0("d",distribution))
  p <- match.fun(paste0("p",distribution))
  q <- match.fun(paste0("q",distribution))
  dparams <- list(...)

  if (!require(ggplot2)) stop("Need 'ggplot2' to use this function")

  worm_ci_UL <- function(nx){
    1.96 * sqrt(p(nx, ...) * (1 - p(nx, ...)) / length(nx)) / d(nx, ...)
  }

  worm_ci_LL <- function(nx){
    -1.96 * sqrt(p(nx, ...) * (1 - p(nx, ...)) / length(nx)) / d(nx, ...)
  }

  ggplot(mapping = aes(sample = x)) + 
    stat_qq(aes(y = stat(sample - theoretical)), distribution = q, dparams = dparams) + 
    stat_qq(aes(y = stat(worm_ci_UL(theoretical))), geom = "line", distribution = q, dparams = dparams) + 
    stat_qq(aes(y = stat(worm_ci_LL(theoretical))), geom = "line", distribution = q, dparams = dparams) + 
    labs(x = paste0("theoretical ", distribution, " quantiles")) + 
    NULL
}

x <- rt(200, df = 4)
qq_worm_plot(x)
#> Loading required package: ggplot2

qq_worm_plot(x, distribution = "t", df = 4)

^{Created on 2020-01-31 by the reprex package (v0.3.0)}

[goodekat]

Thanks for the suggestion! This is a good idea. We do have confidence interval boundaries available with the current implementation of the qqplot. These make use of the computations available from the qqplotr package. We could even work on adjusting the regular qqplot to allow for non-normal boundaries.

[mattansb]

Missed the qqbands option, thanks!

Would it be possible to add the detrend argument from qqplotr?

[goodekat]

Oh! Good idea. I will add that to the to do list.