It seems that standardized Euclidean metric ( --p-metric : seuclidean) is almost never used for beta diversity and group significance tests. Is it not appropriate metric for amplicons data due to some assumptions?
1 Like
Hi @sbombin,
In general, Euclidean distance makes bad assumptions about microbiome data. Part of this has to do with the way it projects into 3D space and provides a lot of artifacts (see figure 4 from this paper). I'm also going to recommend looking at the DEICODE paper, which also addresses some of the issues with euclidean distance.
I may be missing some of the nuance to this, but in general, yes
Best,
Justine
5 Likes