I have a question concerning the alpha diversity boxplots, on the qzv file that I attached, we can see that for the variable "n.family" the kruskal wallis statistical test is significant for all groups with a p-value of 0.02 but when I look at the kruskal wallis test in pairwise, I see that all the q-values (p-value corrected with Benjamini & Hochberg correction) are not below the threshold of 0.05. My question is what to do? Should I take into account the significance of this variable, or only take the uncorrected p-value? And if so, how would you present it in a publication? Thank you in advance
faith-pd-group-significance-w-sex.qzv (365.4 KB)
This situation is quiet common. I think all, that you can report, is that alpha diversity was affected by this factor, but pairwise comparisons revealed no differences between groups.
You have a big number of groups and with p = 0.0298 (0.03) there is a great probability that all p-values in the pairwise test will be corrected to a higher values than 0.05. In addition, the number of samples within each of the groups is very low.
I would try to pool some of the groups (if it possible) to get smaller number of groups with higher number of samples. Or, if measures are repeated, try mixed effects.
Thank you @timanix for your answer, you're right, I'm going to try mixed effects!
This topic was automatically closed 31 days after the last reply. New replies are no longer allowed.