摘要
在贝叶斯框架下检验含有多个均值与趋势双突变点的时间序列的平稳性。引入标号随机变量表示观测值所处的位置,运用贝叶斯因子模型选择的方法判断结构变点数目,对待检参数的先验设定为混合分布,采用可信区间检验序列是否存在单位根,并用蒙特卡罗模拟验证该方法的有效性,且以中国居民消费价格指数为对象进行实证研究,进一步印证了贝叶斯单位根检验在结构突变时间序列单位根检验中的优越性。研究发现:判断序列是否存在单位根时,不能忽视结构突变问题,否则会产生误判;先验设定对单位根检验影响较大,混合先验优于单一先验;中国居民价格消费指数序列在不考虑结构突变时是不平稳的,若考虑结构突变则该序列平稳;贝叶斯单位根检验克服了经典方法在有限样本单位根检验时存在有偏的问题,提高了单位根检验的功效。
Test the stationarity of the time series with multiple structural breaks in Bayesian framework.The labelled random variable is introduced to indicate the position of the observation value,the Bayesian factor model selection method is used to determine the number of structural change points,the mixed prior is prposed for,confidence interval is used to test whether the sequence has a unit root.The performance of the method is assessed by Monte Carlo simulations.Finally,an empirical application is conducted with the aim to detect the structural breaks and test the stationarity in the Chinese consume price index.Research Findings:Ignoring structural breaks are responsible for the unit root test;The prior setting has a greater impact on the unit root test,and the mixed prior is better than the single prior;The Chinese CPI series is not stable when structural breaks are ignored,and if the structural breaks are considered,it is stable;Bayesian unit roots test overcomes the bias problem of the classical method in unit root testing of limited samples,and improves the test power.
作者
史代敏
施晓燕
SHI Dai-min;SHI Xiao-yan(School of Statistics,Southwestern University of Finance and Economics,Chengdu 611130,China;College of Science,Gansu Agricultural University,Lanzhou 730100,China)
出处
《数理统计与管理》
北大核心
2023年第3期427-438,共12页
Journal of Applied Statistics and Management
基金
国家社科基金重点项目(19AZD010)
甘肃农业大学盛彤笙科技创新基金(GSAU-STS-1713)。
关键词
结构突变
模型选择
贝叶斯因子
可信区间
单位根检验
structural breaks
model selection
Bayes factor
credible interval
unit root test
