AMOVA significance


I have a question about the significance in AMOVA and the meaning of the star. I assumed the threshold of significance is 0.05 and the star indicates the significance. I did AMOVA for Bray-Curtis and UniFrac for one set of samples. For the UniFrac AMOVA I get p=0.014 and no star.

Does that mean communities marked with Hp and La in the designfile are significantly different for both UniFrac and Bray-Curtis or only for Bray-Curtis?
What is the meaning of the star?

Hp-La Among Within Total
SS 0.645631 2.8314 3.47703
df 1 13 14
MS 0.645631 0.2178

Fs: 2.96433
p-value: 0.008*

Hp-La Among Within Total
SS 0.300402 1.3066 1.607
df 1 13 14
MS 0.300402 0.100508

Fs: 2.98884
p-value: 0.014


I suspect that you actually have more than 2 groups in your overall analysis, right? By default, we apply a Bonferroni correction. If you have 4 groups, you would have 6 unique pairwise comparisons and your critical p-value would be 0.05/6 or 0.0083. So, in the first case it would be significant because it’s below 0.08 and in the second it would not be significant because it is above 0.0083 (assuming you have 4 groups). Note that many consider Bonferroni to be very conservative and you should look into other methods of correcting for multiple comparisons. Note also that many people think that the 0.05 threshold is somewhat arbitrary and meaningless. Regardless, to avoid getting accused of p-hacking you should pick methods a priori and apply them and not change them to get the result you want :wink:

Thank you very much for the fast answer!

I do have 4 groups.