Beta diversity significant but post hoc not significant. (how do i interpret (these) results of beta diversity)

[Adeel_Adeel (Adeel Abbasi)]

Experiment: i performed beta diversity analysis on samples from four different provinces (3 samples each province). analysis was performed on galaxy (which has qiime2 v2022.11.1)
when i look at overall test result, the bray_curtis beta diversity shows p-value 0.005 as highly significant (should i say unique) diversity between groups.
but looking at results of pair wise. is see that no group is significant to each other when looking at p-value or q-value. how do i write it in article?

Pics are attached

link to q2view file: https://view.qiime2.org/visualization/?type=html&src=https%3A%2F%2Fcancer.usegalaxy.org%2Fdisplay_application%2Fc0966d495f623ff1%2Fq2view%2Fmain_q2view%2Ff06bbc38a521118f%2Fdata%2Fgalaxy_c0966d495f623ff1.qzv

[jwdebelius (Justine Debelius)]

Hi @Adeel_Adeel,

I would interpret your results as "the group sizes are too small". You've got n=3 per group, which doesn't even get you to 999 unique permutations.

So, I think interpretation is hard.

Best,
Justine

[Adeel_Adeel (Adeel Abbasi)]

Thank you for the response. yes i have 3 samples in each case. should i set the permutations value to 3 before running the command?

[jwdebelius (Justine Debelius)]

Hi @Adeel_Adeel,

No, you should calculate the number of possible permutations where 12 samples can be allocated into 4 groups and that should be the maximum number of permtuations you can calculate.

You should maybe also consider whether you violate the assumptions of other statistical tests with the sample size. (Kruskal wallis, for example, wants at least 5 samples per group.)

For me, 3 samples per group is too small to do anything other than prove the ability to collect and sequence samples and so I would be very careful with my interpretations.

Best,
Justine