Updated documentation. Especially new recommendations for K in ppc_pi…

…t_ecdf*() and mcmc_rank_ecdf().

R/helpers-ppc.R

@@ -416,12 +416,12 @@ adjust_gamma_simulate <- function(N, L, K, prob, M) {
 #' simultaneous coverage of the ECDF traces.
 #' @noRd
 interpolate_gamma <- function(N, K, prob, L) {
-  # Find the precomputed values ueful for the interpolation task.
+  # Find the precomputed values useful for the interpolation task.
   vals <- get_interpolation_values(N, K, L, prob)
   # Largest lower bound and smalles upper bound for N among precomputed values.
   N_lb <- max(vals[vals$N <= N, ]$N)
   N_ub <- min(vals[vals$N >= N, ]$N)
-  # Approximate largest lower bound and smalles upper bound for gamma.
+  # Approximate largest lower bound and smallest upper bound for gamma.
   log_gamma_lb <- approx(
     x = log(vals[vals$N == N_lb, ]$K),
     y = log(vals[vals$N == N_lb, ]$val),

man-roxygen/args-pit-ecdf.R

@@ -1,7 +1,9 @@
 #' @param K An optional integer defining the number of equally spaced evaluation
-#'   points for the ECDF. If the submitted PIT values are known to be discrete,
-#'   this should be the number of the discrete cases. Defaults to the smaller of
-#'   `length(y)` and `ncol(yrep)` when applicable.
+#'   points for the ECDF. Reducing K when using `interpolate_adj = FALSE` makes
+#' computing the confidence bands faster. For `ppc_pit_ecdf` and
+#' `ppc_pit_ecdf_grouped`, defaults to `ncol(yrep) + 1`, or `length(pit)` if PIT
+#'  values are supplied. For `mcmc_rank_ecdf` defaults to the number of
+#'  iterations per chain in `x`.
 #' @param prob The desired simultaneous coverage level of the bands around the
 #'   ECDF. A value in (0,1).
 #' @param plot_diff A boolean defining whether to plot the difference between