有限样本递归均值调整单位根检验与Bootstrap研究

针对理论研究和模拟研究参数设置不一致问题,推导零均值递归均值调整单位根检验功效公式,结果表明其分布在大样本下与非零均值结果相同.为纠正临界值检验的不足,提出了3种Bootstrap样本构造方法,证明了Bootstrap方法既可以用于研究检验水平也可以分析检验功效.蒙特卡洛模拟表明,Bootstrap方法不但具有完美的检验水平和较低的水平扭曲,而且也具有功效优势; 实证研究结论也表明Bootstrap方法可以用于实际序列的单位根检验.研究结论既丰富了单位根递归均值调整检验理论,也为实证研究提供一种新检验方法.

The Recursive Mean Adjustment Unit Root Test and Bootstrap Research for Finite Sample Size

To resolve the inconsistency on parameter setting between the theory research and simulation,the power formula for recursive mean adjustment unit root test with zero mean is derived,concluding that the formulae are the same to those with nonzero mean under large samples.To correct the deficiency of critic value test,three kinds of Bootstrap sample construction method are proposed and the theoretical proof shows that the Bootstrap method can both be used to study the test level and also to analyze the power.The results from Monte Carlo simulation indicate that Bootstrap method not only has perfect test level and lower level distortions,but also has advantages on power.Empirical research shows the Bootstrap method can be used to perform unit root test for actual series.This study not only enriches the recursive mean adjustment unit root test theory,but also provides a new method for empirical research.