Tuomas Borman
Tuomas Borman
Weighted and normalized are different things. We have now - weighted and normalized - weighted and unnormalized (can be found from rbiom) - unweighted and normalized (can be found from...
We have this https://www.nature.com/articles/ismej200997 And rbiom has this https://journals.asm.org/doi/10.1128/aem.71.12.8228-8235.2005
At least this PR https://github.com/joey711/phyloseq/issues/956 that you linked before shows that rbiom gives similar results than others (only phyloseq differs). phyloseq uses fast unifrac. I assume that Felix used phyloseq...
1. The reasoning for having `fast` instead of `weighted` and & `normalized` is that they are different algorithms. Also if we decide that `rbiom` is used if `wieghted=TRUE & normalized=FALSE...
1. The only thing is that our implementation gives possibility to choose whether to normalize or not. rbiom does not have `normalized` argument. It always calculates unnormalized weighted unifrac or...
We decided to completely switch using rbiom::unifrac, and remove our own implementation. Our own implementation is based on phyloseq implementation of Fast Unifrac, and phyloseq's implementation has been showed to...
Tests fail because weighted rbiom::unifrac gives different result every time (
It seems that rownames does not atch with tip labels. They should match and can be matched. Check if that is the case by debugging the lines just before rbiom::unifrac...
Also, check what is wrong with the other error (that is not even related to this PR?)