演講摘要 : The purpose of this talk is to propose and investigate inference and model comparison for an asymmetric nonlinear double smooth transition heteroskedastic model. Parameter estimation and inference are performed in a Bayesian framework by designing a Markov chain Monte Carlo scheme. We present an informative prior for the smoothing parameter that allows reliable inference and a proper posterior, despite the non-integrability of the likelihood function; plus a mixture prior is suggested that solves the identifiability problem as the smoothing parameter tends to zero. Next, we employ importance sampling to develop a formal Bayes factor procedure to compare the proposed double smooth transition model with two special cases: the double threshold GARCH and symmetric ARX GARCH models. The proposed methods are illustrated using both simulated and international stock market return series. Results suggest that previous studies employing smooth transition models may have been better suited with a sharp threshold model.
