# Chapter 11 Summary

This workshop covered a handful of constrained analyses, which allow us to **test hypotheses** about the drivers of patterns in a response matrix, such as a matrix describing the abundance of species sampled across many sites. We can use RDA, partial RDA, and variation partitioning to quantify the importance of different variables (or, groups of variables) on a response matrix. In many cases, this response matrix was a community composition matrix of sites x species, but these techniques are not limited to community ecology.

We also saw two ways of testing hypotheses about site groupings. We can use multivariate regression trees (MRT) to determine which explanatory variables distinguish groups of sites, and describe how our response matrix is organised into these distinct groups. If we already have an *a priori* grouping of sites, we can use linear discriminant analysis (LDA) to verify whether this grouping aligns with environmental data, and predict the grouping of new sites.