摘要
首先对单位根检验的两类常见的数据生成系统进行比较,然后利用蒙特卡洛实验研究了时间序列单位根检验式的设定问题。研究发现在利用DF检验和DF-GLS检验进行时间序列的单位根检验时,检验式设定错误直接影响着检验结果,尤其在推断时间序列是趋势平稳过程还是有时间趋势项的随机游走过程或有二阶时间趋势多项式的随机游走过程时,检验式的错误设定很容易将趋势平稳过程误判为非平稳过程。
Firstly, two kinds of data generating process are compared; then, specification of unit root test on time series is studied by Monte Carlo simulations. It is founded that the misspecification of the test model have direct impact on behavior of DF and DF - GLS unit root test. Especially, to judge whether time series is a trend stationary process or a random walk process with time trend term or with two order time trend polynomials, the misspecification of the test model induce that easily let a trend stationary process is mistaking for a non-stationary process.
出处
《统计与信息论坛》
CSSCI
2008年第4期9-13,共5页
Journal of Statistics and Information
基金
国家自然科学基金项目<基于Bayesian方法的单位根检验和协整检验方法研究>(70771072)
关键词
单位根检验
检验式设定
小样本性质
unit root test
specification of test model
finite sample property
