I have a pair of groups that are not significantly different in PERMANOVA (p > 0.05), yet in SIMPER the average dissimilarity between these two groups are very high, even higher than other pairs of groups that are actually significantly different in PERMANOVA (p < 0.05). How can that happen? Shouldn't the dissimilarity in SIMPER reflect the results in PERMANOVA?
Simper measures dissimilarity and PERMANOVA measures equivalence. Just decide which one you like best and use and trust that one. No-one is going to ask why you didn't compare one with the other. Always use the best method and ignore the others!
Actually, in the Simper results the within-group resemblances are similarities, and the between-group resemblances are dissimilarities, at least they are in Primer (can't speak for Vegan or whatever). The other point is that Simper is not a test, so will give a breakdown of resemblances regardless of whether the differences among groups are significant under some other hypothesis test. Permanova (and ANOSIM) take into account the variability in resemblances, as Andrew says.
- Badges
- Science method