I wanted to check weather the communities in a tree had different structure significantly, so I used the command parsimony, unifrac.unweigted, and unifrac.weigted like this: parsimony(tree=subsample.jclass.0.03.tre,group=sample.design). There were 11 groups (communities) in the group file (i.e., sample.design) and then I got below result for commands parsimony and unifrac.unweigted :
Tree# Groups ParsScore ParsSig
1 merged 26 0.001
The first problem was that I did not know the meaning of “merged” groups.
Then I added the parameter “groups=all” or “groups=BH-HY-JX-JZ-LN-ML-CS-PJ-SY-XM-XT” (BH and others were the name of communities or sampling sites), I could only get the parsimony test results for each pair of communities (such as BH-HY); however, I don’t know how to get the result for all the communities, i.e., the result for BH-HY-JX-JZ-LN-ML-CS-PJ-SY-XM-XT.
Could anyone anwser me? Thanks a lot!
We have this set up like an anova. The first hypothesis being tested is whether there is one group that is significantly different from the others - this is the merged test. If the p value comes back as being significant (e.g. less than 0.05) then you can do groups=all and it will do all of the pairwise comparisons. You then need to compare these pvalues against a corrected value of 0.05. A conservative approach would be to divide 0.05 by the number of comparisons being performed. So if you had 10 comparisons, your pvalues would have to be less than 0.005 to be considered significant.
Hope this helps,
Pat