Files
Abstract
This article describes the process of Bayesian specification analysis using state of the art simulation methods. It distinguishes between predictive specification analysis (comparison of model predictions with actual outcomes) and post-predictive specification analysis (comparison of predictions of a replication of an experiment with actual outcomes). The methods are illustrated using a specific example, evolving volatility in financial returns. Partial financial support from National Science Foundation grants SBR-9731037 and SES-9996332 is gratefully acknowledged. George Chang provided help with the computations and useful discussions. The usual disclaimers apply