A1030
Title: Covariate dependent Beta-GOS process
Authors: Weixuan Zhu - Xiamen University (China) [presenting]
Abstract: Covariate-dependent processes have been widely used in Bayesian nonparametric statistics thanks to their flexibility in incorporating covariate information and allowing for correlation among process realizations. Unlike most of the existing work that focuses on extensions of exchangeable species sampling processes such as the Dirichlet process, we propose a new class of covariate-dependent nonexchangeable priors by considering the generalization of the Beta-GOS model. We show that the proposed prior has an equivalent formulation under a continuous kernel mixture and its latent variable representation, which leads to a natural nonexchangeable parallel with the classical dependent Dirichlet process formulation. We further apply the proposed prior for regression and autoregressive models and show that its posterior sampling algorithm enjoys the same computational complexity as that of the Beta-GOS. We demonstrate the excellent numerical performance of our method via simulation and two real data examples.