摘要
本文发现Perron于1989年的研究在趋势突变情形下的结论 "统计量T(^α-1)的极限分布会随着突变点位置参数的变化收敛在0~0.5"值得商榷。其原因在于模型设定中出现了错误,导致在结构突变的趋势平稳过程的数据生成过程中,统计量T(^α-1)的极限分布在截距突变的情况下发散,而在斜率突变的情况下退化。本文对其进行修正并补充推导了三种含结构突变的趋势平稳过程的单位根检验统计量的分布,并给出能够证实和证伪的蒙特卡洛模拟结果。
We find that the conclusion of Perron (1989), which said that the statistics T(a - 1) ' s limit distribution converge to an interval between 0 and 0.5 for different position of the parameter of structural break under trend break, is debatable. The reason is that a mistake in the model setting which leads to the limit distribution of the statistics T(a--1) volatilization when intercept break or degeneration when trend break under the data generating process of trend stationary process with structural break. So we correct this mistake and supplement this conclusion with unit root test on three kinds of the trend stationary process with structural break and Monte Carlo simulations which support our findings.
出处
《数量经济技术经济研究》
CSSCI
北大核心
2013年第1期117-134,共18页
Journal of Quantitative & Technological Economics
基金
教育部项目"时间序列非平稳检验理论与应用研究"(09YJA790111)
国家社会科学基金项目(09CJY014)
教育部人文社科项目(08JC790058)的资助
关键词
结构突变
趋势平稳过程
单位根检验
Structural Break
Trend Stationary Process
Unit Root Test