Multivariate analyses of interaction networks

This project started during my PhD. It aims at better understanding rules structuring interactions networks, in particular trait matching. For this, we use methods from the correspondence analysis family.

n a first step, we to used simple correspondence analysis (CA) with reciprocal scaling. These unconstrained ordination methods allow to order species along latent gradients, which can be interpreted as their unmeasured traits involved in their interactions. This provides a method to investigate trait matching between species, even when traits are not measured! These results have been published here.

Then, we investigated constrained ordination methods, Constrained CA (CCA) and double-constrained CA (dc-CA). They allow to incorporate trait into the analysis, to get a measure of matching, partition the variation and project species in a constrained space. This has been discussed in my PhD thesis, and an article is also in preparation.