期刊名称：GREQAM Documents de Travail / Groupement de Recherche en Economie Quantitative d'Aix-Marseille
出版年度：1998
摘要：This paper considers the Bayesian analysis of threshold regression models. It shows that this analysis can be conducted with simple deterministic numericla integration rules of low dimension. The shape of the posterior density is greatly determined by the type of threshold and of transition function considered. Unequal variances between the regimes usually adds one dimension to the integration problem, except in some cases where a simplification occurs. Various particular models with smooth and step transition functions are detailed with empirical illustrations concerning the French consumption function and asymmetries in the US business cycle. Comparison with classical methods are underlined.
关键词：Bayesian inference; nonlinear time series models; structural change.
