Bonferroni FWER correction

Thin wrapper over stats::p.adjust(method = "bonferroni").

Usage

bonferroni(p_values, alpha = 0.05, labels = NULL)

Arguments

p_values

Numeric vector of raw p-values.

alpha

Significance level.

labels

Optional character vector of test labels.

Value

A morie_rich_result list (see morie_multiple_testing).

Examples

set.seed(1)
# 60 tests: 50 null (uniform p) + 10 strong signals near zero.
p <- c(runif(50), runif(10, 0, 0.005))

res <- bonferroni(p, alpha = 0.05)
res                       # rich print: method, alpha, tests, rejected
#> Adjusted p-values (bonferroni)
#> ==============================
#> Call: method=bonferroni, alpha=0.0500, n=60 
#> 
#>   Method          bonferroni
#>   alpha           0.05
#>   Tests (n)       60
#>   Rejected        2
#>   Min adjusted p  0.021204
#>   Max adjusted p  1
#> 
#> 2 of 60 hypotheses are rejected at alpha=0.0500. 

# The result carries the full adjusted-p and rejection vectors.
head(res$adjusted)        # each raw p multiplied by n (capped at 1)
#> [1] 1 1 1 1 1 1
res$n_rejected            # how many survive alpha after correction
#> [1] 2
which(res$rejected)       # indices declared significant
#> [1] 55 56

# `alpha` sets the rejection threshold; stricter alpha rejects fewer.
bonferroni(p, alpha = 0.01)$n_rejected
#> [1] 0

# `labels` names each test so the output is self-documenting.
bonferroni(c(0.001, 0.02, 0.3),
           labels = c("geneA", "geneB", "geneC"))$labels
#> [1] "geneA" "geneB" "geneC"

# Bonferroni is the most conservative FWER method -- compare to Holm/BH.
c(bonferroni = bonferroni(p)$n_rejected,
  holm       = holm(p)$n_rejected,
  BH         = benjamini_hochberg(p)$n_rejected)
#> bonferroni       holm         BH 
#>          2          2         10