Configuration of site scores from a model-based ordination, applied to the Minchin 1987 dataset. It shows a perfect lattice.
Configuration of site scores from CA, DCA, DCA with log-transformed data, and NMDS, applied to the Minchin 1987 dataset. All figures show artefacts.
In almost every presentation on model-based ordination I attach these two figures. It shows a re-analysis of (one of) the infamous dataset(s) by Minchin (1987) who concluded NMDS to be the best ordination method. But, #GLLVMs do much better, even if the model is misspecified.