A0165
Title: Change points detection in VAR models
Authors: Siddhartha Chib - Washington University in Saint Louis (United States) [presenting]
Abstract: A Bayesian method for conducting inference on possible single or multiple change points in VAR models is presented. Under a conjugate prior to the parameters of the VAR model, it is shown that under regularity conditions, the posterior distribution of the change-point location (in a model with a single change-point) concentrates on the true change point as the sample sizes become large. By a novel method of proof, this result is extended to a family of non-conjugate priors on the VAR parameters. VAR models are considered with multiple change points where results on posterior consistency of change-point determination are conditioned on knowledge of other change points. Establishing joint posterior consistency of multiple change points remains an open problem. The change-point detection methodology is applied to several problems and shows that change points are common in typical VAR applications.
