PERMANOVA on small sample size?

[MarwaTawfik]

Hi team,

Could you please let me know what are the lowest sample size that could be considered acceptable? I did the ordination plot and can see the difference/dissimilarity visually between intestine, feed, and water.

The problem is that I have n=2 or n=3 in water/feed that I suspect the PERMANOVA results.
Please comment and tell me what you think.

Thanks in advance
Marwa

[jwdebelius (Justine Debelius)]

Hi @MarwaTawfik,

My recommendation is to not do analysis with fewer than 5 samples per group.

This is kind of the rule of thumb minimum sample size for kruskal wallis.

With any permutation test, your space is limited based on the number of samples you have to make combinations. If you have 5 total samples, you have a possibility of 5! = 120 permutations.

When you go to do differential abundance, it's not going to play nicely with ANCOM, which is (a) built on a kruskal wallis test, and (b) requires that you get a p-value small enough after FDR to achieve a specific threshhold (which is unlikley to happen unless you violate some assumptions.)

Best,
Justine