L-shaped data, GLM(M) and double constrained correspondence analysis

Presentation summarizing 5 papers on trait-environment relations in ecology, L-shaped data or bipartite graphs. It starts with a description of the data, a simple loglinear (GLM) model and the Rao score test on interaction. This leads to the Legendre's fourth-corner correlation and, for multi-trait, multi-environment problems, double constrained correspondence analysis, which then enables testing and selection of the important trait and environmental variables. A comparison is made with RLQ and double constrained principal component analysis. An alternative title: from fourth-corner correlation to double constrained correspondence analysis and their relation with GLM(M) models. The pptx and pdf files contain the same information.