A1010
Title: Variance partitioning-based priors for species distribution models
Authors: Luisa Ferrari - University of Modena & Reggio Emilia (Italy) [presenting]
Massimo Ventrucci - Department of Statistical Sciences, University of Bologna (Italy)
Abstract: Species distribution models for community ecology often require complex models to capture the influence of numerous abiotic factors with complex relationships, as well as spatial and temporal patterns in the data. For this reason, the Bayesian hierarchical framework is a popular choice for analysing species occurrence data. However, traditional prior specification approaches have been found to have several disadvantages. Recently proposed variance partitioning-based priors offer a promising new framework to handle the problem of incorporating prior assumptions, particularly for complex models. The core idea of this prior class consists of designing a more intuitive reparametrization of the original parameters for which specification becomes easier for the user. The applicability is, however, so far limited to models that include a subset of all possible effects. The purpose is to explore how to extend this new method to a wider class of models. In particular, it focuses on incorporating spatial and temporal smoothing effects, which are highly popular within species distribution models. The aim is to expand the usefulness of variance-partitioning-based priors to new fields of application, such as the context of ecology studies, among others.
