While doing a PERMANOVA (R-vegan function adonis), I get a significant P value, however, the "betadisper" by RVAideMemoire package also gives a significant difference in the "multivariate spread" among the test groups (P=0.0003). I conducted a permutation-based ANOVA to get this result (significant dispersal). My understanding is that the PERMANOVA results could be compromised. Does anybody know what should be my next step? Do I just report the significant dispersal along with the significant PERMANOVA results or is there any way to fix the issue. I tried scaling and transformation, nothing helps!!!
One approach is to report these statistics along with some sort of ordination (NMDS for example). This will help you see if your effect is a location effect, a dispersion effect, or both. It's common to have both a location and a dispersion effect at the same time. Depending on how many groups you are comparing, you may also want to follow up with the individual pairwise comparisons.
I agree with Julian! Please see the following papers to get an idea how to deal with the issue and report the right statistics effectively:
1) Article Drought and increased CO 2 alter floral visual and olfactory...
2) Article Environment and host species shape the skin microbiome of ca...
Good luck!
Aziz, in circumstances like this I generally look to Marti Anderson.
Here's a reference I have found useful and keep returning to
Article PERMANOVA, ANOSIM, and the Mantel test in the face of hetero...
effectively you are looking for a description of the multi-dimensional shape of the data cloud in response to your treatments. It is not unreasonable to state that there are both location (PERMANOVA) and dispersion (betadisper or more correctly PERMDISP) effects.
As Julian suggests above, a 2D- or 3D-NMDS plot may help. To test a specific hypothesis, you may want to consider Constrained Analysis of Principal Coordinates (CAP), another Marti Anderson approach
- Badges
- Science topic
- Similar topics
- Organizational Psychology
- Training