B0839
Title: Copula particle filters
Authors: Carlos Erwin Rodriguez - IIMAS-UNAM (Mexico) [presenting]
Stephen Walker - University of Texas at Austin (United States)
Abstract: A novel analysis of the state space model is presented. It is shown that by modifying the standard recursive update, it is possible to apply a copula model to eliminate a particular integral, which is typically performed using importance sampling. With Bayesian models, copulas have recently been shown to provide predictive densities directly, avoiding integrals altogether. As in every particle filter algorithm, particles are generated; hence the proposed algorithm is named the Copula Particle Filter (CPF). As a by-product, the likelihood function of the model is obtained and used for parameter inference.
