ANOISM

[Chris Pace]

Hello all, I'm new to qiime and bioinformatics, so please excuse my lack of knowledge. 

I recently ran compare_categories.py using the anoism method to determine significant differences between four groups shown on unifrac PCoA plots.   The p-value was significant for the comparisons, but I'm worried that any strong difference in one group from the other three groups may confound the p-value.  I guess what i'm asking is anoism a good test to use on more than 2 groups.  Similar to an ANOVA or kruskal wallis. 

Any help would be appreciated.

Thanks

[Colin Brislawn]

Hello Chris,

Welcome!

that any strong difference in one group from the other three groups may confound the p-value.

Correct. A low p-value shows that this difference is probably not due to chance alone, but it does not tell you which groups the difference is between. 

I guess what i'm asking is anoism a good test to use on more than 2 groups.  Similar to an ANOVA or kruskal wallis. 

Yes, this is a good test for groups, and just like ANOVA or kruskal wallis, you would follow up these tests with a post hoc test that let's you compare specific pairs of groups to find out which ones are causing the low p-value.

Unfortunately, I'm not sure if qiime 1 has any post hoc tests for anosim. One workaround would be to filter your samples to just have two groups at a time, then run anosim on each of these groups. It's not elegant, but it should work.

If qiime does include post hoc tests, I would like to learn more about them.

Colin