Partenaires

CNRS
Logo tutelle
Logo tutelle
Logo tutelle


Rechercher

Sur ce site

Sur le Web du CNRS


Accueil du site > Séminaires > Probabilités Statistiques et réseaux de neurones > Spatio-temporal modelling of epidemiological processes

Vendredi 1er décembre 2006

Spatio-temporal modelling of epidemiological processes

Carlo Gaetan (Université de Venizia, Italie), gaetan@unive.it

Résumé : The aim of this talk is to discuss the potential of the varying coefficient modelling approach for tackling the modelling tasks typically encountered in epidemiology. Epidemiological processes exhibit complicated behavior over an extensive range of spatial and temporal scales of variability, giving rise to complex dynamics. The usual statistical approaches try to simplify complexity of such systems by setting up models which either ignore the multivariate interaction, or assume spatial/temporal stationarity, linearity, and Gaussianity. However, it is increasingly the case that the scientific questions of interest are becoming sufficiently complex that one can no longer justify such assumptions.To address some of the issues raised by such problems, we shall describe the use of regression models (not necessarily Gaussian) where where the regression coefficients are allowed to change in space or in time. We will discuss an parameter-driven approach through latent Gaussian Markov random fields. In the last stage the hierarchical model is completed by specifying a prior distribution for the hyperparameters. Bayesian inference is approximated by drawing samples from the posterior distribution by means of an MCMC algorithm. The resulting models are very powerful in that relatively simple spatial and temporal dependence assigned to subprocesses and parameters can lead to very complicated joint spatio-temporal dependence. Results are illustrated with a real dataset.

Dans la même rubrique :