Bayesian Subset Selection for Two-Threshold Variable Autoregressive Models

In this paper, we propose and study an effective Bayesian subset selection method for two-threshold variable autoregressive (TTV-AR) models. The usual complexity of model selection is increased by capturing the uncertainty of the two unknown threshold levels and the two unknown delay lags. By using Markov chain Monte Carlo (MCMC) techniques with driven by a stochastic search, we can identify the best subset model from a large number of possible choices. Simulation experiments show that the proposed method works very well. As applied to the application to the Hang Seng index, we successfully distinguish the best subset TTV-AR model.