I have approximately 200 samples with outcome labels (success or failure) and various covariates, collected across multiple time points. My analysis plan includes using PERMANOVA to identify significant covariates, followed by differential abundance analysis using DESeq2, adjusting for the important covariates identified. If I subset the data to focus on a specific time point or compare selected time points to assess differential expression between outcomes, should I re-evaluate and include covariates specific to that subset? Additionally, is it necessary to re-run PERMANOVA on the subset?
Hi @Naorem,
I maybe have bad news in that PERMANOVA (at least how we have it implemented in QIIME 2) isn't well suited to repeated-measures. Specifically, it will permute without regard for subject.
If you were to slice down to specific time points, then this issue disappears, so in a way, it may make the best sense to simply do that if it's already well suited to your hypotheses.
You may also wish to use adonis2 in R directly as that lets you specify permutations via how() which lets you separate out your repeated measures (I will be honest, the logic of how() still eludes me, so I'm not even sure how you would specifically construct it, but I know that's the mechanism to use).
I hope the above kind of indirectly answers some of the later questions in your post.