I have only one fixed_effect, and it 's divided into two groups in the metadata table file. When I run MaAsLin2, there are pval and qval in the result file of all_results.tsv, and only the result with qval < 0.25 will appear in significant_results.tsv.
pval: the nominal significance of this association.
qval: the corrected significance is computed with p.adjust with the correction method (BH, etc.).
Since I have only two groups in my analysis, could I directly use the pval to assess microbial differences between the two groups instead of qval? My concern is that BH correction is typically applied for multiple comparisons, but here I am making only a single comparison.
I hope someone can enlighten me on this. Thank you!
Even with only one fixed effect and two groups, MaAsLin2 is not performing a single statistical test. It is performing one test per feature (taxon, pathway, metabolite, etc.), all for the same group contrast. The multiple testing issue arises from testing many features in parallel, not from the number of groups in the metadata.
The p-value is the nominal, feature-wise statistic, while the q-value corrects for multiple testing across all features. Using raw p-values would only be appropriate if a single, pre-specified feature were tested. In typical microbiome analyses, q-values remain the appropriate measure of significance, even with just two groups.
Thank you for the detailed explanation — it was very helpful. I’d like to clarify one more point: in my case, I’m only comparing the abundance of taxon between the two groups. Would you still recommend using the qval to assess significance in this scenario?
Thank you for your time and patience. I look forward to your guidance.