A0886
Title: Joint species distribution modeling with mixture models
Authors: Ching-Lung Hsu - Duke University (United States) [presenting]
Tommaso Rigon - University of Milano-Bicocca (Italy)
David Dunson - Duke University (United States)
Abstract: Community ecology seeks to understand the interactions between species and their driving environmental factors. Joint species distribution models are increasingly studied and applied to ecological data for estimating species associations and the predictive power in future samples. A Bayesian mixture model is proposed for co-occurrence probabilities allowing the discovery of unseen species. The results are leveraged from previous work, and the predictive formulas are derived for species discovery in this setting. It is shown that asymptotically the model is a mixture of a two-parameter Indian buffet process. A simple Gibbs sampler is developed for posterior computation. As an implementation, the model is applied to the Guelph Arthropod dataset.
