Journal of Statistics: Advances in Theory and ApplicationsAuthor's: HANWEN ZHANG
Pages: [73] - [90]
Received Date: November 2, 2012
Submitted by:
In this paper, we consider the problem of the identification and estimation of TAR models with one threshold variable when the noise process follows a t distribution. A Bayesian procedure of three stages based on Gibbs samplings is proposed. A simulated example indicates that the procedure proposed has a good performance.
TAR models, Gibbs sampling, t distribution.
