p-value in PERMDISP

Dear all,

I have carried out PERMANOVA test to assess the group dissimilarity and cross-checked with PERMDISP to see whether the dissimilarities were due to large within-group dispersions. Please refer to the attached case for my query:-




The first barplot (distances to F1) shows F1 distance is similar to F2 and S1, with permanova p-value for F1 vs F2: q=0.0132; F1 vs S1: q=0.1510. However, permdisp pairwise test shows F1 vs F2:q=0.915; F1 vs S1: q=0.915. Can I propose a statement - the composition of F1 is different to F2 (p<0.05) and the difference is not ascribed to variances within the group?
Therefore, p<0.05 in permdisp indicate there is significant differences in variance for any group?

Benedict

Hi @Benedict,

Yes, I think that’s a good way to describe the differences. I would be careful to specify your metric - remember that different metrics can show different things and what may be significant (or highly dispersed) in one metric might not appear the same way in the other.) My one suggestion is to provide the actual p-values, and of course, the number of permutations.

Best,
Justine