STAR模型下退势单位根检验统计量的比较  

Comparative Study on Performance of De-trending Unit Root Test in STAR Models

作　　者：欧阳敏华 

机构地区：[1]华南师范大学经济与管理学院,广州510006

出　　处：《统计与决策》2017年第23期19-22,共4页Statistics & Decision

基　　金：中国博士后科学基金资助项目(2014M562245)

摘　　要：文章将Chang(2002)^([1])和Taylor(2002)^([2])提出的递归退势方法拓展到STAR模型下的退势单位根检验中,并采用Monte Carlo数值模拟的方式比较了递归退势、OLS退势和GLS退势下KSS单位根检验统计量的有限样本表现。递归退势KSS单位根检验统计量都有较好的检验势,且不受初始条件的影响。不考虑初始条件,GLS退势KSS单位根检验统计量的检验势要高于OLS退势,而随着初始条件的增大,GLS退势KSS单位根检验统计量的检验势却下降得非常厉害。This paper extends the recursive de-trending procedures proposed by Chang （2002） and Taylor （2002） to de-trending unit root test in STAR model, and then uses Monte Carlo simulation to compare the finite sample performance of the KSS unit root test under recursive, OLS, GLS and recursive de-trending procedures. Generally, the KSS test using recursive de-trending method is more powerful, and not affected by the initial condition. The power of GLS-hased KSS test is higher than that of OLS for the case without consideration of the initial condition, but decreases drastically as the initial deviation gets large.

关 键 词：STAR模型 递归退势 单位根检验 KSS统计量 

分 类 号：F224[经济管理—国民经济]