Bonferroni type (sequential or strict)

I am trying to figure out which type of Bonferroni test is used by Qiime 2. Is it sequential or strict? Thanks!

Hi Bryan,

Can you share the command you ran? ( I can use that to look into the code and see what kind of test was done.)

Thanks!

1 Like

qiime composition ancombc --i-table collapsed-table-lvl-2.qza --m-metadata-file metadata-cropland.tsv --p-p-adj-method bonferroni --p-reference-levels "species::Ag.placidum" --p-formula species --o-differentials differentials-agpl-2.qza

This qiime2 plugin is just wrapping ancombc() in R. Here are those docs:
https://rdrr.io/github/FrederickHuangLin/ANCOMBC/man/ancombc.html

See ?stats::p.adjust for more details.

Here are the docs for p.adjust()

So as far as I can tell, --p-p-adj-method bonferroni is strict while "holm" is sequential.

While we are here, I wanted to bring this to your attention before reviewer three does :grimacing: :

There seems no reason to use the unmodified Bonferroni correction because it is dominated by Holm's method, which is also valid under arbitrary assumptions.

1 Like