BAYESIAN HIERARCHICAL MODELS: MODELING THE SPATIOTEMPORAL DISTRIBUTION OF THE RISKS OF PHENOMENA MULTIPLE STRUCTURES

Computer Engineering

A good understanding of the spatiotemporal distribution of complex phenomena and multiple structures (such as the occurrence of fires in the environmental field, epidemic disease in the field of health ...) is an important element for the risk management of these phenomena and the development of strategies. However, the data are subject to complexities caused by heterogeneities among host classes and spatiotemporal processes. This article seeks to suggest or propose a Bayesian spatiotemporal model to modeling and mapping the relative risks of these phenomena in space and time. In this paper, we used spatiotemporal Bayesian hierarchical models to study the relative risk patterns of these phenomena. The R-INLA method with R packages was used for simulations and parameter estimation. The most suitable model is selected using multiple validation criteria (DIC, WAIC, CPO, ...). Among the spatiotemporal models used, the Knorr-Held model with a type III and IV space-time interaction fits well with the data, but type IV seems better than type III. We begin with the introduction to explain the theoretical and problematic context of our work. Then, the spatiotemporal statistical models in the literature review, followed by the proposed methodology which is the Bayesian hierarchical models, followed by an application on the occurrence of fires in the Ankarafantsika National Park (ANP) and ends by conclusion and perspective.

This work is licensed under a Creative Commons Attribution-ShareAlike 4.0 International License.