A UNIFIED STRATEGY FOR ESTIMATING AND CONTROLLING SPATIAL, TEMPORAL AND PHYLOGENETIC AUTOCORRELATION IN ECOLOGICAL MODELS

Abstract

The goals of this paper are to expose ecologists to the problem related to statistical inference when testing the association between data sets that are autocorrelated and to introduce a relatively new method for controlling the bias introduced by autocorrelation that can be easily incorporated in any statistical approach. In addition, I show the flexibility of this class of methods to the types of data that ecologists are currently most interested, namely temporal, spatial and phylogenetic data. In this contribution, I also stress the point that is not all variation due to autocorrelation that affects statistical inference and is important to control only the component that biases inference. Thus, statistical frameworks should attempt to separate the autocorrelation component that biases inference from the one that may prove interesting for understanding important ecological processes, such as contagious processes, driving spatial patterns in species distributions.