I am studying whether environmental parameters affect the community composition of root fungi living in symbiosis with beeches in a forest. Therefore, I am performing a RDA with a presence-absence community dataset and an environmental dataset which contains continuous environmental parameters (pH, soil moisture...) and a categorical parameter "Regeneration", which has 4 classes corresponding to the developmental stage of the beech (SNReg and SReg are seedlings, RegLow and RegHigh are saplings).
When plotting an RDA with all my parameters included, and subsequently after stepforward selection, this categorical parameter is plotted as centroids instead of arrows as it is supposed to do. However, stepforward selection revealed longitude to be strongly driving the variation in community composition. Therefore, I wanted to do a partial RDA to see what the effect is of the environmental parameters, corrected for the effect of longitude. However, now the categorical parameter "Regeneration" is plotted as arrows instead of centroids, while it is recognized as a factor. If I call the summary of the RDA, there are no centroids for factor constraints, and the factor "Regeneration" is only included in the biplot scores for constraining variables. Replacing by dummy variables did not resolve this problem either.
I post images of the RDA plot with centroids, and the partial RDA plot without centroids, where regeneration is plotted as arrows.
