Adjusts p-values for multiple testing using common methods. Essential for subgroup analyses in GBA/AMNOG dossiers where multiple comparisons require appropriate correction.
Usage
adjust_pvalues(
p,
method = c("holm", "hochberg", "hommel", "bonferroni", "BH", "fdr", "BY", "none"),
alpha = 0.05
)Arguments
- p
Numeric vector of p-values
- method
Character. Adjustment method:
"holm" (default): Holm-Bonferroni step-down (controls FWER)
"hochberg": Hochberg step-up (controls FWER)
"hommel": Hommel method (controls FWER)
"bonferroni": Bonferroni correction (controls FWER, conservative)
"BH" or "fdr": Benjamini-Hochberg (controls FDR)
"BY": Benjamini-Yekutieli (controls FDR under dependency)
"none": No adjustment
- alpha
Numeric. Significance level (default: 0.05)
Value
A data frame with columns:
original_p: Original p-values
adjusted_p: Adjusted p-values
significant_original: Logical for original significance
significant_adjusted: Logical for adjusted significance
method: Adjustment method used
Details
FWER methods control the probability of making any false rejection. FDR methods control the expected proportion of false rejections.
For GBA submissions, Holm or Hochberg methods are commonly recommended as they control FWER while being less conservative than Bonferroni.
Examples
# Adjust p-values from multiple subgroup comparisons
p_values <- c(0.01, 0.04, 0.03, 0.15, 0.008)
adjust_pvalues(p_values, method = "holm")
#> original_p adjusted_p significant_original significant_adjusted method
#> 1 0.010 0.04 TRUE TRUE holm
#> 2 0.040 0.09 TRUE FALSE holm
#> 3 0.030 0.09 TRUE FALSE holm
#> 4 0.150 0.15 FALSE FALSE holm
#> 5 0.008 0.04 TRUE TRUE holm
# Compare different adjustment methods
adjust_pvalues(p_values, method = "bonferroni")
#> original_p adjusted_p significant_original significant_adjusted method
#> 1 0.010 0.05 TRUE FALSE bonferroni
#> 2 0.040 0.20 TRUE FALSE bonferroni
#> 3 0.030 0.15 TRUE FALSE bonferroni
#> 4 0.150 0.75 FALSE FALSE bonferroni
#> 5 0.008 0.04 TRUE TRUE bonferroni
adjust_pvalues(p_values, method = "BH")
#> original_p adjusted_p significant_original significant_adjusted method
#> 1 0.010 0.025 TRUE TRUE BH
#> 2 0.040 0.050 TRUE FALSE BH
#> 3 0.030 0.050 TRUE FALSE BH
#> 4 0.150 0.150 FALSE FALSE BH
#> 5 0.008 0.025 TRUE TRUE BH
# With custom significance level
adjust_pvalues(p_values, method = "holm", alpha = 0.10)
#> original_p adjusted_p significant_original significant_adjusted method
#> 1 0.010 0.04 TRUE TRUE holm
#> 2 0.040 0.09 TRUE TRUE holm
#> 3 0.030 0.09 TRUE TRUE holm
#> 4 0.150 0.15 FALSE FALSE holm
#> 5 0.008 0.04 TRUE TRUE holm