power calc or rules of thumb for N samples

At the risk that this is too general a question:

Does anyone have references or rules of thumb for how many samples are generally required / minimally required to detect significant community differences for any of the permutation based methods?

For example, I’m comparing 2 locations, N plots per location, M treatments per plot. Based on preliminary studies, with N=2, M=2 I’m observing ~500 OTUs (fungal ITS), with about 50 of these being most prevalent.

I’m suspecting there is a lower limit of samples required simply based on using permutation to test significance…

Any insight would be appreciated!


from an R class that targets big ecologists who have been grappling with those kinds of questions for a while


Thanks, looks like a useful reference to get started!