I try to compare the output from permanova (run in PrimerE) and adonis (in R) on the same dataset. SS, MS, and F are identical. However, I am getting more and more confused about the terms "estimated variance components" (ECV) and "explained variance" (R-square ?). adonis outputs R-square as SSfactor/SStotal (which is sometimes also called eta-square (η2) in a regular anova). As this statistic is somewhat biased, omega-square (ω2) is sometimes calculated, but this is another story. However, looking at the permanova output and the explanations in the manual, the story seems to be much more complex. permanova estimates the components of variation in the model using the mean squares. If I understand correctly, the square-root of these ECV are then directly related to the explained variance and percent explained variance is derived by getting the proportions of the total (=sum). There has been a post on researchgate on this (https://www.researchgate.net/post/Can_you_report_PERMANOVA_estimated_components_of_variation_ECV_as_a_proportion_of_variation_explained_by_factors_terms), but it didn't really solve that issue, did it?

Can someone help to explain the differences between ECV and explained variance and the way permanova and adonis is handlig these differently? Any guidance is greatly appreciated.

More Martin Hartmann's questions See All
Similar questions and discussions