The study of species association is of great interest in ecology due to its role in understanding key issues such as patterns of habitat use by animals, species coexistence, biotic interactions, and in general factors affecting community structure and assembly. There are many indices that ecologists commonly use, all based on the observed frequencies of organism occurrences, to evaluate the association between a pair of species. However, few of these indices correspond to proper statistical measures of association, and the inferential aspects of their analysis are often overlooked. In this paper, we propose a Bayesian approach based on a simple multinomial-Dirichlet structure to provide a comprehensive inferential framework for any set of association indices. Our approach provides a full statistical analysis for any association index of interest, free of special requirements on the sample size. We illustrate our procedure with a camera-trapping real-dataset, but the analysis of any other dataset of the same type can be readily produced using the R package basa that accompanies this paper.
