Tukey’s test v.s. Benjamini-Hochberg. The latter is better.
When people have multiple groups and have no prior hypothesis about which specific groups might differ, people usually use Tukey’s test as a post hoc test that test all possible group comparisons.
x <- factor(rep(LETTERS[1:6],rep(4,6)))
p_matrix = matrix(NA, 1000, 15) for(i in 1:1000){ ps = as.numeric(pairwise.t.test(rnorm(24),x,p.adjust.method=‘none’)$p.value) ps = ps[!is.na(ps)] p_matrix[i,] = ps }
For attribution, please cite this work as
Fan (2018, Dec. 9). Metabox-Blog: Just use Benjamini-Hochberg!. Retrieved from https://hushuli.github.io/Metabox-Blog.github.io/posts/2018-12-09-just-use-benjamini-hochberg/
BibTeX citation
@misc{fan2018just,
author = {Fan, Sili},
title = {Metabox-Blog: Just use Benjamini-Hochberg!},
url = {https://hushuli.github.io/Metabox-Blog.github.io/posts/2018-12-09-just-use-benjamini-hochberg/},
year = {2018}
}