Hi. I have performed a PERMANOVA analysis on the feeding habits of a fish species. I found in the literature that the results of a PERMANOVA analysis should be confirmed with the PERMDISP test. My questions are: What is the purpose of the PERMDISP test? Is it necessary to do this test to confirm the results of a PERMANOVA analysis?
In general no! PERMDISP is another essential analysis to test null hypothesis. It is used for calculating dispersion effects. For significant test just PERMANOVA or ANOSIM is okay. Thanks
Hi, PERMDISP is a multivariate analog of Levene's test for homogeneity of multivariate variances. Distances are handled by reducing the original distances to principal coordinates. PERMDISP calculates an F-statistic to assess whether the dispersions between groups are significant.
Your argument for doing PERMDISP is slightly circular because it tests for homogeneity, or equality, of multivariate variances. But this is already an assumption underlying PERMANOVA, which tests something different i.e. whether distance differs between groups So if anything, maybe PERMDISP could be done first :)
As with ANOVA, where there are assumptions concerning the nature of the data to be tested (normally distributed variables, homogeneity of variances) and you would do something like a Levene's test to check before running the analysis, , there are also assumptions with PERMANOVA, including homogeneity of multivariate dispersions. Although PERMANOVA is fairly robust to differences in dispersion (Article Some solutions to the multivariate Behrens-Fisher problem fo...
) you should really do a PERMDISP test (which is a test for significant differences in dispersion) before you do PERMANOVA. Alternatively you could use ANOSIM which doesn't make any assumptions, being fully non-parametric.
- Badges
- Science method