厚尾金融时间序列的贝叶斯单位根检验  

Bayesian Unit Root Testing for Economical Time Series with Heavy Distribution

作　　者：李勇[1] 孙瑞博[2] 王贵银[1] 

机构地区：[1]中山大学管理学院,广东广州510275 [2]南京财经大学经济学院,江苏南京210046

出　　处：《数理统计与管理》2012年第1期184-190,共7页Journal of Applied Statistics and Management

基　　金：国家自然科学基金项目(70901077);教育部社科基金青年项目(094YJC370266)的资助

摘　　要：在金融时间序列分析中,单位根检验是一个相当重要的研究问题。对于数据生成过程为自回归的厚尾金融时序数据,古典的ADF单位根检验统计量应用非常困难。本文则在贝叶斯框架下,发展了检验带有未知自由度厚尾t分布的自回归金融时间序列单位根的贝叶斯方法。蒙特卡罗模拟结果显示本文发展的方法能够取得好的检验功效。最后,用万科地产月度历史收益数据来演示了本文发展的方法。In times analysis, unit root test is a very important research topic. As to time series data which is generated by autoregressive process, classical unit root testing statistics such as ADF can not be applied easily to test unit root for the economical time series with heavy tails. This paper aims to develop a new Bayesian unit root testing approach for these time series with heavy distribution and unknown degree of freedom. Monte Carlo simulation studies show that this new test statistic proposed can obtain good power. At last, the developed approach is illustrated with a month-return series of Wanke real estate Company.

关 键 词：单位根 贝叶斯因子 厚尾金融时间序列 贝叶斯检验 路径抽样 

分 类 号：F222[经济管理—国民经济] O212[理学—概率论与数理统计]