Marginal likelihood calculation for gelfand-dey and Chib Method

One advantage of Bayesian estimation is its solid theoretical ground on model comparison, which relies heavily upon the accurate calculation of marginal likelihood. The Gelfand-Dey (1994) and Chib (1995) methods are two popular means of calculating marginal likelihood. A trade-off exists between these two methods. The Gelfand-Dey method is simpler and faster to conduct, while Chib method is more accurate, yet intricate. In this paper, we compare the two methods by their ability to identify structural breaks in a reduced form volatility model. Using the Markov Chain Monte Carlo method, we demonstrate that the performance of the two methods is fairly close. Since the Chib method is normally more di±cult to implement in many econometric problems, it is safe to choose Gelfand-Dey method when calculating marginal likelihood.