Reduced-Rank Generalised Dissimilarity Modelling (RRGDM): Model-Based Ordination of Dissimilarity Data

Reduced-Rank Generalised Dissimilarity Modelling (RRGDM) reframes dissimilarity regression as model-based ordination, enabling interpretable inference on environmental gradients driving beta diversity.

• Why it matters: Existing approaches link site-level predictors to pairwise dissimilarities by using predictor distances, which obscures how variables act on individual samples and complicates random effects and dispersion modeling.
• What they did: RRGDM represents dissimilarities via sample positions in a shared low-dimensional latent space, then models positions as linear functions of predictors with error terms plus a dispersion sub-model.
• The result: In simulations and two empirical examples, RRGDM recovers known environmental effects and disentangles influences on composition versus dispersion, supporting inference on relative gradient importance.