Single-command 1-ANOVA followed by pairwise tests possible?


I have a shared-file, 12 samples (rows) and 1001 columns (OTUs). The dataset is divided into three treatments (n=4 in each).

I would like to test if the abundances of the OTUs differ among treatments. I would like to do the main test using 1-ANOVA or Kruskall-Wallis and then proceed to test the pairwise differences among groups. Is this all achievable with a single mothur command. I already tried some commands:

  1. Metastats
  • did pairwise tests (t-test?) but did not do the main ANOVA test
  1. Kruskall-Wallis
  • did the main test but did not do the pairwise tests
  1. Lefse
  • did go closest to that I wanted. It did the main test (Kruskal-Wallis) and did also pairwise tests. However, in the “class” - column of the result file, it seems to show only the name of the treatment with the highest abundance. Is it somehow possible to see the results of all the pairwise tests here? I tried changing the multiclass-parameter but that did not change the result.


Antti J Rissanen, Finland

You’ve done everything you can in mothur. We would take the files into R, filter the shared file to keep the OTUs that meet certain abundance thresholds, and then do ANOVA/KW there with correction for multiple comparisons. It’s just too hard to replicate R in mothur :slight_smile:


Ok. Thanks. Yes, I will go on with another program from here.