p_kruskal.test: p-value from Kruskal-Wallis Rank Sum Test simulation

In Spower: Power Analyses using Monte Carlo Simulations

p-value from Kruskal-Wallis Rank Sum Test simulation

Description

Simulates data given two or more parent distributions and returns a p-value using kruskal.test. Default generates data from Gaussian distributions, however this can be modified.

Usage

p_kruskal.test(
  n,
  k,
  means,
  n.ratios = rep(1, k),
  gen_fun = gen_kruskal.test,
  ...
)

gen_kruskal.test(n, k, n.ratios, means, ...)

Arguments

n	sample size per group
k	number of groups
means	vector of means to control location parameters
n.ratios	allocation ratios reflecting the sample size ratios. Default of 1 sets the groups to be the same size (n * n.ratio)
gen_fun	function used to generate the required data. Object returned must be a list of length k, where each element contains the sample data in each group. Default uses gen_kruskal.test. User defined version of this function must include the argument ...
...	additional arguments to pass to gen_fun

Value

a single p-value

Author(s)

Phil Chalmers [e-mail]

Examples

# three group test where data generate from Gaussian distributions
p_kruskal.test(n=30, k=3, means=c(0, .5, .6))

# generate data from chi-squared distributions with different variances
gen_chisq <- function(n, k, n.ratios, means, dfs, ...){
  dat <- vector('list', k)
  ns <- n * n.ratios
  for(g in 1:k)
 	dat[[g]] <- rchisq(ns[g], df=dfs[g]) - dfs[g] + means[g]
  dat
}

p_kruskal.test(n=30, k=3, means=c(0, 1, 2),
   gen_fun=gen_chisq, dfs=c(10, 15, 20))


  # empirical power estimate
  p_kruskal.test(n=30, k=3, means=c(0, .5, .6)) |> Spower()
  p_kruskal.test(n=30, k=3, means=c(0, 1, 2), gen_fun=gen_chisq,
         dfs = c(10, 15, 20)) |> Spower()

Spower documentation built on April 4, 2025, 5:11 a.m.