摘要
在多个结构突变的检测方法中,常用的序贯检测不能同时检测到所有突变点,在实时监测中存在缺陷。文章通过潜在状态变量将对多个结构突变的检测转换为多个潜在状态的检测,基于不可逆的隐马尔可夫链方法对多个未知结构突变进行估计,同时得到多结构突变点的个数和位置估计。利用该方法对中国五省市的实证表明:五省市通胀率在历史时期上表现出五个明显的变化特征,存在三个突变点,分别发生在2006Q4、2008Q1和2013Q3。引起突变的原因在于外部经济环境的变化和内部宏观经济调控手段的影响。
The commonly used sequential detection method can not detect all the breaks simultaneously in multiple breaks detection, which is a shortcoming in real-time monitoring. This paper introduces the latent state variables to convert the detection of multiple breaks into that of multiple latent states, and depends on the nonreversible hidden Markov chain model to estimate the multiple unknown structural breaks, and at the same time obtain the number of breaks and location estimation as well. Finally the paper applies the proposed model to an empirical test, which shows that the inflation rate of the five provinces in China assumes five obvious changing features and three structural breaks, which occurred in 2006Q4, 2008Q1 and 2013Q3 respectively. The change of external economic environment and the internal macroeconomic regulations are the main causes for each break of infla- tion rate.
出处
《统计与决策》
CSSCI
北大核心
2018年第3期14-19,共6页
Statistics & Decision
基金
国家社会科学基金资助项目(17BTJ033)
关键词
潜变量
结构突变
同时检测
通胀率
latent state variable
structural break
simultaneous detection
inflation rate