A0290
Title: Bayesian model selection among dispersed integer-valued time series models
Authors: Feng Chi Liu - Feng Chia University (Taiwan) [presenting]
Cathy W-S Chen - Feng Chia University (Taiwan)
Hsiao-Han Hsu - Feng Chia University (Taiwan)
Abstract: The focus is on a class of integer-valued time series models with over-dispersion and extends those models to generalized forms. These new models include: (1) dispersed INGARCH models incorporating negative binomial, double Poisson, or generalized Poisson, and (2) double log-form INGARCH model. The latter model avoids over-restrictions in the parameter space. We perform parameter estimations and model selection within the Bayesian framework, employ adaptive Markov chain Monte Carlo (MCMC) sampling schemes, and calculate the deviance information criterion (DIC) for model selection. Simulation studies demonstrate that the proposed method accurately estimates the model parameters with reliable MCMC samples. Taking monthly crime counts in Bankstown, New South Wales, Australia, for an empirical illustration, the findings show the ability to select the promising models among the competing models in terms of DIC.
