Dear all,

I am a new member here, working mainly in Ecology. I hope this is the right section to ask my question.

I am dealing with unbalanced designs (I am comparing groups with different sample sizes) in PERMANOVA (Permutational Multivariate Analysis of Variance). The PERMANOVA runs fine, and results are consistent with patterns I can observe in a NMDS or PCoA plot. But I would like to rule out any possible effect of the difference in size between my samples on the significant result I obtain. This is because one of my sample is half the size of the others. So my question is: How could I do that ?

I was thinking about a resampling method (bootstrapping) prior to the analysis, although I am not familiar with these. But then I realized there was an option in the software I am using (Primer Software) to calculate p-value using monte carlo approximation instead of permutations. I know that this option is meant to deal with small samples that don’t allow a sufficient number of permutations to calculate a p-value. In which case, monte carlo approximation is more suitable. But – please correct me if I am wrong - if this « monte carlo p-value » is based on a resampling method shouldn’t it already rule out the possible sample sizes effects in the result I obtain ? In which case I could just use the monte carlo approximation in the test instead of having to go through bootstraping.

I am really not familiar with resampling methods, so I might be completely wrong and any help would be much appreciated,

Thank you in advance


Yser