Skip to contents

Permutation-based FWER control via step-down max-T (Westfall-Young)

Source: R/multiple_testing.R
permutation_fwer.Rd

Given observed test statistics and a matrix of test statistics under the permutation null, computes step-down max-T adjusted p-values that strongly control the family-wise error rate without requiring independence across tests.

Usage

permutation_fwer(
  test_stats,
  null_stats,
  alternative = c("two_sided", "greater", "less"),
  alpha = 0.05,
  labels = NULL
)

Arguments

test_stats

Numeric vector of length \(m\) of observed test statistics.

null_stats

Numeric matrix with \(n_{perm}\) rows and \(m\) columns containing test statistics computed on permuted data.

alternative

One of "two_sided" (default), "greater", or "less".

alpha

Significance level for rejection (default 0.05).

labels

Optional character vector of test labels.

Value

A morie_rich_result list with original, adjusted, rejected, method, alpha, n_rejected, n_tests.

Examples

set.seed(1)
m <- 10; nperm <- 200
obs <- c(rnorm(m - 2), 4.0, 3.5)
null <- matrix(rnorm(nperm * m), nperm, m)
permutation_fwer(obs, null)
#> Adjusted p-values (permutation_maxT)
#> ====================================
#> Call: method=permutation_maxT, alpha=0.0500, n=10 
#> 
#>   Method          permutation_maxT
#>   alpha           0.05
#>   Tests (n)       10
#>   Rejected        2
#>   Min adjusted p  0
#>   Max adjusted p  0.975
#> 
#> 2 of 10 hypotheses are rejected at alpha=0.0500. Step-down max-T over 200 permutations.