General alpha diversity interpretation and pairwise comparison

Hi again :slight_smile:
I need someone help me clarifying the follow issue:

I have 4 treatment with three replicates (16S rRNA V3 and V4 regions data). I have the alpha diversity analysis (Shannon´s Index) between groups with a p-value: 0,033 means it´s significative but in the pairwise comparation taking in to account the q-value there are´n any difference between pairs. What should I take for the interpretation? : S
taxa-bar-plots-tiempo-2-grouped.qzv (1.0 MB)

Shoul I take the p-value of the all groups (significative) or should I take all pairwise p-values (No significative)?

Hi @Jibda,
Looks like you attached the wrong QZV — but that is okay, your description is clear enough.

p=0.033 is fairly borderline, so it makes sense that pairwise comparisons would be insignificant. I think the result speaks for itself: you could report that there is a significant difference between all groups, but the pairwise comparison renders that finding meaningless. You can certainly report the results of the test across all groups — but make sure to also report the pairwise results for full transparency.

Opps, sorry.
I attached the wrong file :see_no_evil:
shannon_group_significance.qzv (300.0 KB)

I don´t know if it´s ok just say that there is not any difference between groups showing up only the pairwise comparison without showing the difference between groups. For me it´s like the same only that we have in the pairwise comparison the security that there aren´t differences.

Thanks for your fast response :slight_smile:

Thanks for attaching the QZV.

I think it is okay to report, as long as you report both results. Looking at the boxplot, it looks like there probably is a difference between groups, but your sample sizes in each group are really low (N=3), so you just don’t have enough power and the pairwise tests are insignificant.

The ideal solution would be to gather more data, though you could also collapse your groups to increase sample sizes (e.g., mulch and mulch MO look quite similar and are both elevated relative to the other groups; collapsing these into a single mulch group may increase power in pairwise comparisons)

Other alpha diversity metrics may also yield more pronounced results, so don’t give up hope!

Thanks!!! :slight_smile:

1 Like

This topic was automatically closed 31 days after the last reply. New replies are no longer allowed.