统计研究 ›› 2009, Vol. 26 ›› Issue (9): 55-63.

• 论文 • 上一篇    下一篇

基于Bayesian Gibbs Sampler的状态空间模型估计方法研究及其在中国潜在产出估计上的应用

赵昕东 耿鹏   

  • 收稿日期:1900-01-01 修回日期:1900-01-01 出版日期:2009-09-15 发布日期:2009-09-15

Estimation Method of State Space Model Using Bayesian Gibbs Sampler and Its Application on Estimating China’s Potential Growth Rate

Zhao Xindong & Geng Peng   

  • Received:1900-01-01 Revised:1900-01-01 Online:2009-09-15 Published:2009-09-15

摘要: 本文将贝叶斯吉伯斯样本生成(Bayesian Gibbs Sampling,BGS)方法应用到状态空间模型的估计。首先介绍了BGS方法的基本内容和计算步骤,然后给定参数生成满足状态空间模型的模拟数据,并对模拟数据应用BGS方法估计。结果表明参数与状态向量的估计值与参数值与状态向量的真实值相当接近,明显优于基于Kalman滤波的最大似然估计结果。最后,本文将BGS算法应用于中国1980年至2008年的潜在增长率与增长率缺口的估计。

关键词: 贝叶斯估计, 吉伯斯样本生成, 状态空间模型, 潜在增长率

Abstract: In this paper we apply the Bayesian Gibbs Sampling (BGS) method to the estimate of state space model. First, we introduce the basic contents and computing algorithm, then giving the parameters we generated the data that satisfy a state space model and estimate the model using BGS method. The results show that the estimated parameters and state vector are very close to the real ones and are clearly prior to those of the maximum likelihood estimation based on Kalman filter. Finally, we estimate China’s potential GDP growth rate and growth rate gap from 1980 to 2008. 
 

Key words: Bayesian Estimation, Gibbs Sampler, State Space Model, Potential Growth Rate