At first thank you for your enormous work!
Especially we like the idea of adding metastats for determining significant differences.
But we are not very wise in statistics…

Can you please add some information about how to read the output file to the wiki. How to interpret it?

If I have 12 samples in 4 sets and wish to get p-values for the sets… metastats gives p-values for every OTU? How do we get p-values between sets?

We’ll work on the documentation. metastats only compares otus, not samples. You might try using the unifrac.unweighted/weighted commands, parsimony, or libshuff to test for differences between samples. Typically, there is so much power to detect differences that it isn’t too interesting to test for differences between samples.