Pairwise-differences kruskal wallis

Dear Qiime2 developers,

I am running the qiime longitudinal pairwise-differences test on various alpha diversity metrics between two timepoints for two groups, and wanted to check I am interpreting the output correctly (the second and third tables).

Does the second table (Multiple group tests) compare the differences between the two groups, and the third table (Pairwise group comparison tests) compare the values between the two groups? For example if I was interpreting the attached qzv file, would it be correct to say "there a significantly greater (alpha div) distance between weeks 1 and 4 for MR2 compared to MR1, indicating an increased or more rapid accumulation of diversity in this group".

Also for the second table (Multiple group tests), is a corrected p value not necessary?

shannon_W1_W4.qzv (257.8 KB)

Thank you for your help,

Hi @JenKelly,
Thanks for posting!

When you have only 2 groups, the 2nd and 3rd tables are effectively the same question but with different tests. The 3rd table is really only there for when multiple groups are being tested (in which case there is one pairwise test for each pair of groups).

Yes, that sounds correct based on the Kruskal-Wallis test (if you are considering p < 0.05 significant). The Mann-Whitney U test is giving an insignificant result, however. So I would be a little cautious, especially since you have a small sample size (5 per group) and a P close to 0.05.

When you have 2 groups, it is not necessary. But again, this table is really only there for when multiple groups exist, in which case multiple test correction is warranted.

I hope that helps!

1 Like

Hi @Nicholas_Bokulich,

Thanks for the explanation, it’s much clearer now!


1 Like

This topic was automatically closed 31 days after the last reply. New replies are no longer allowed.