Analysing the allometry of multiple interacting traits

Since form and function are tightly integrated in plants, and since plant attributes often scale allometrically, it follows that plant allometry is inherently multivariate. Unfortunately, traditional statistical methods for studying allometric relationships are very restrictive and do not allow one to model multivariate allometric patterns that follow realistic biological hypotheses. In this paper I describe a new statistical test ('d-sep test') that allows one to test, and potentially falsify, alternative multivarlate orderings of cause-and-effect in the context of allometry.