Bayesian Inference for Multiple Change Points in Time Series-Demonstration Based on Chinese GDP Series

摘要

In this article, a basic Bayesian analysis is introduced to detect multiple change points in time series firstly. The Bayes factor is used to judge the number of change points. Then, GDP series is analyzed based on this theory and Gibbs sample is carried out by dint of WinBUGS software. There are three change points in GDP series, which happened in 1961, 1976, 1989 respectively. It's general in agreement with our country's economic seedtime. At last, it is familiar that the model with inserting proper change points is much better in forecast.