Identifying significance for each level of a factor with envfit

[hkaspersen]

Hello!
I am using envfit to see how two factor variables fit to an ordination.
I see in different tutorials that with numerical variables you get one p-value per variable.
For the factor variables you only get one p-value per variable, but does this then mean that all the factor levels are significant?
I'm trying to only plot significant variables, but I'm a bit unsure how to interpret this, as I will get one arrow per factor level.
Any help is greatly appreciated!

[jarioksa]

You shouldn't get arrows per factor levels, but you should get centroids of factor levels. Function envfit has no way of assessing "significances" of single factor levels, but you really should get only a permutation p-value for the whole factor with all its levels. You should define your factor as an R factor for correct analysis.

Actually, I am not quite sure how would you define significance of a single factor level. Somehow it would be a contrast of each single level against all other levels, and I'm not sure what would that mean.

[hkaspersen]

Thanks for the reply!
I am wondering though, as I do get significance for each factor level in the anova.cca on the overall model with capscale. Is this similar, or completely different? Sorry I do not have a statistical background so all this is new to me!

[jarioksa]

The only way you get a p-value in capscale (or adonis2 or any constrained ordination model) is with by = "onedf" and this gives you contrasts to the first level of the factor. This hardly makes sense in envfit and also will omit the first level of the factor.