PERMANOVA : Significant p-value in the overall test but no significant q-value

[Parix]

Hi,

I have performed PERMANOVA to see if the samples are clustering based on studied subjects. The resulting is attached. So the p-value of the test is significant which I guess means that at least one group should be significantly different from the others, however, the q-values in the Pairwise test are all non-significant and quiet high. I was wondering how to interpret this outcome?
bray-curtis-individual-significance-permanova.qzv (914.1 KB)

[jwdebelius (Justine Debelius)]

Hi @Parix,

I would intrepet the outcome as possibly underpowered for paired analysis. You have 40 samples with 20 groups. You've got multiple problems here. First, permutaiton with four groups isn't likely to give you a result because you're running up against the possibly permutational space (There are only 4!=24 ways to permute the data. Your best case correct p-value is 1/25). You might consider looking at some of the paired sample tests in the q2-longitudinal plugin.

Best,
Justine