中文摘要：

核心CPI能更真实地反映宏观经济运行趋势。然而，如何测算核心CPI一直备受关注。本文构建了一组带有异常因子的随机游走模型，利用MCMC法和Gibbs抽样，对时变参数进行估计，不仅从CPI中分离出核心CPI，而且对非核心CPI，通过捕捉到的异常点，发现与中国政策效应相吻合。本文仅利用了中国单一的CPI数据，不需要CPI子类权重测算及其再分配。进一步，通过与Wind核心CPI比较，以及Marques等（2003）核心CPI评价方法的检验，发现本文的核心CPI更具合理性和科学性。

英文摘要：

Core CPI can more realistically reflect the macroeconomic trends. However, how to measure core CPI has been a matter of great concern. This paper constructs a set of random walk model with abnormal factors and estimates time varying parameters by using the MCMC method and the Gibbs sampling in the models. The results show that not only core CPI is isolated from CPI, but also the abnormal points of non-core CPI are captured, which are consistent with the policy effects in China This random walk identification method uses China single CPI data without calcu- lating CPI sub-class weights and their redistribution. Further, compared with Wind core CPI and tested with the core CPI evaluation method from Marques and Neves （2003）, the core CPI proposed in this paper is more reasonable and scientific.