Adonis plugin versus Mantel beta significance plugin


I was experimenting with the new adonis plugin in Qiime2 and was surprised to find that it would say metadata factors that were considered significant by the Mantel testing plugin were not even close to significant with adonis. For example, I have a BMI factor that was considered a pval of 0.003 with Mantel and 0.15 with adonis. Why are these so far from each other? Shouldn’t both these test give the same overall result?


A quick google search yielded this:

adonis is analogous to RDA or CCA, as it directly estimates the variance
in the distance matrix attributable to an independent variables(s), with
the advantage that one may use any distance measure. It parallels r2 in
linear model.
Mantel test simply calculates the correlation between two distance
matrices (and tests it via permutations), thus, one gets the idea about
wthether there is a linear relationship between them at all.


This topic was automatically closed 31 days after the last reply. New replies are no longer allowed.