Setting alpha cutoffs for KW and Wilcoxon tests

[lina....@gmail.com]

Hi all,

I was under the impression that the KW test is the first one to be performed by LEfSe. I.e. if I set -a equal to 1, I should get all features as significant.

However, this was not the case for my dataset. My input dataset contains 767 features but only 461 features that are non-zero in at least one sample.

When I set both -a and -w to 0.99, LEfSe found 370 features.

When I set both -a and -w to 2, LEfSe found 412 features.

Is there another step performed that removes features early on? Which step am I misinterpreting?

Thanks for any input!
~Lina

[Jonathan Greenberg]

Interesting question, I am curious if this was ever answered.