Determining dispersion for beta diversity

Hey all,

I have read quite a few other posts on this but I am not sure I understand the solutions fully.

To sum up my experiment, I have conditioned soil with different cover crop treatments for 3 months and I want to see if the fungal communities differ under those cover crop treatments. I understand the PERMANOVA results but is there a test to tell if the dispersion in one group is larger than another (if some treatments cluster tightly vs others being spread out on the PCoA)? You can visually see dispersion on the PCoA but can I obtain p values or exact numbers?

It would be very important to know if one cover crop treatment (clover for example) supported more dissimilar communities than another (wheat let’s say).

Can I determine these things by specifying --p-method permdisp in my script?

Thank you

Hi @Andrew_Richards,

I think your understanding of the hypotheses are a bit mixed.

PERMANOVA is the same as a regular ANOVA: is there a difference in mean and variance between any of the groups?

Permdisp asks if there is a difference in dispersion (variance) between any of the groups.

So, if I have two groups with centroids (red and blue here)
|960px;x278px;

with permanova, I'm asking mainly if the centroid of the red and the blue groups are in the same place (although it ends up also asking about hte variation) and the second asks if the spread of the red and blue are different. Its possible, for example, that red and blue overlap perfectly but red has way more variation than blue which can be its own kind of signal.

So, with a possibly confusing romp through theory, you can specify you want permdisp (variance) with the --p-permdisp flag with beta group significance. Permanova is the default.

Best,
Justine

8 Likes

Thank you @jwdebelius for your quick and thorough reply!

1 Like

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