Constrained ordination analysis with flexible response functions

Canonical correspondence analysis (CCA) is perhaps the most popular multivariate technique used by environmental ecologists for constrained ordination; it is an approximation to the maximum likelihood solution of the Gaussian response model. In this article, we look at the constrained ordination problem from a slightly different point of view and argue that it is this particular point of view that CCA implicitly adopts. This gives us additional insights into the nature of CCA. We then exploit the new perspective to generalize the Gaussian response model to incorporate more flexible response functions. A real example is presented to illustrate the use of the more flexible model.