统计研究

• 论文 • 上一篇    下一篇

面板数据单位根似然比检验研究

张华节 黎实   

  • 出版日期:2015-04-15 发布日期:2015-05-21

Research on Likelihood Ratio Type Tests for Unit Roots in Panels

Zhang Huajie& Li Shi   

  • Online:2015-04-15 Published:2015-05-21

摘要: 本文采用似然比类检验统计量进行面板单位根检验(简称为LR检验)研究,在局部备择假设成立的条件下,推导了其在无确定项、仅含截距项以及存在线性时间趋势项三种模型下所对应的渐近分布与局部渐近势函数。Monte Carlo模拟结果显示,当面板数据中含确定项(截距项或时间趋势项)时,LR检验水平比LLC和IPS检验水平更接近于给定的显著性检验水平;此外,当面板数据中包含发散个体时,经水平修正后的LR检验势要远远高于经水平修正后的LLC与IPS检验势,其中,经水平修正后的LLC与IPS检验势接近于零。

关键词: 似然比类检验, 局部备择假设, 渐近分布

Abstract: This paper proposes Likelihood Ratio (LR hereafter) type test statistics for testing unit roots in panels. Under the local alternative hypothesis, it derives the asymptotic distributions and power functions of the LR type test statistics based on the three models, i.e. without deterministic terms, with intercept and linear time trend. Monte Carlo simulations show that the empirical size of LR test is closer to the given significance level than that of LLC and IPS tests, when there are intercepts or time trends in models. In addition, when some units may be explosive in panel data, the size-adjusted power of LR test is much larger than that for LLC and IPS tests, whose size-adjusted power is close to zero in the cases.

Key words: Likelihood Ratio Type Test, Local Alternative Hypothesis, Asymptotic distribution