Multifactor permdisp analysis?

Hi All,

Is there a tool that will perform a multi-way PERMDISP test? I have a dataset where I was comparing the dispersion for one grouping factor with the PERMDISP test. However, I would like to perform a test that is similar to ADONIS tests which it is my understanding that it is comparing means of a distance matrix. I tried to write an R formula with the beta_group_significance test with --p-method permdisp but it didn't accept it.

The reason why I wanted to do this is because when I look at the dispersion for one factor (e.g., Factor1), the p-value is significant. However, I want to make sure that the dispersion is still significant after adjusting for other covariates (e.g., Factor2). Ideally, it would be great to calculate the dispersion interactions like in adonis, such as Factor1*Factor2. Is there a reason why I should not do this?

If this tool or approach doesn't exist, should I run permdisp with each factor and correct the p-values with a multiple hypothesis correction test?

Any feedback on this approach and idea would greatly be appreciated!

1 Like

This may be helpful:

Note that this uses the Vegan package directly in R. I'm not sure what Qiime2 supports.

1 Like

Thank you for sending this! I did go through the tutorial and tried to set it up in the way that I described above. However, I wasn't able to get that to work. At this point, I am not sure though if it is how I am setting it up or if the betadisper tool is unable to do what I want.

Yeah, I get it.

The part that jumped out to me is
adonis2(dist($distances) ~$Group)

So they are using the adonis test on a distance matrix... of dispersion measurements!
The adonis test supports covariant adjustments, so this should do exactly what you want.

Let me know if you get this working. I'm also a big fan of adonis

1 Like