Different results from parsimony and unifrac.unweighted/unifrac.weighted

Dear all,

Hope we have all enjoyed a wonderful Thanksgiving holiday.

Here I have a question about the parsimony and unifrac.unweighted/unifrac.weighted commands. In the Mothur SOP, the results of the commands of parsimony, unifrac.unweighted,
and unifrac.weighted are the same: There was clearly a significant difference between the clustering of the early and late time points.

However, when I ran my own data, I found that:
–The parsimony results indicate that there is no significant difference.
–The unifrac.unweighted/unifrac.weighted results indicate that: The difference is significant.

I understand that the parsimony ignores the branch length and unifrac.unweighted/unifrac.weighted command does. However, I am still not sure how to appropriately interpret the results.



As you’ve noted, the tests test different things. You should only pick one test and then report the result. I would never recommend using unifrac.unweighted since it is highly sensitive to undersampling. I’d probably put more value on having the branch length included.


Thanks a lot for the clarification, Pat. I will then focus on the results from the unifrac.weighted command. Thanks.

This topic was automatically closed 10 days after the last reply. New replies are no longer allowed.