摘要
通过推导Dickey-Fuller检验功效函数,研究表明:即使中小型的傅里叶型结构突变,都会严重影响Dickey-Fuller检验的功效,从而使得含傅里叶型平滑结构突变的平稳过程被误判为单位根过程。使用3、6、9个月期和一年期Shibor日度数据发现:传统的ADF、PP、DF-GLS和KPSS几乎都指出Shibor是单位根过程;考虑平滑结构突变的单位根检验则在1%的显著性水平下拒绝了单位根的原假设,这表明Shibor是含结构突变的平稳过程。因此,预测Shibor和理解其动态行为必须考虑其结构突变特征。
By deriving the power function of Dickey-Fuller test, we show that even a small or medium break can dramatically deteriorate the power of DF test, hence lead to the conclusion that a stationary process with a smooth break is mistaken as a unit root process. Based on the daily data of Shibor with the periods of 3, 6, 9 and 12 months, the empirical results show that the traditional tests such as ADF, PP, DF-GLS and KPSS all tend to indicate the nonstationarity of Shibor, but the Fourier-augmented Dickey- Fuller test can reject the unit root null at the 1% significant level. This shows that Shibor is stationary with breaks. Therefore, it is necessary to concern the structural breaks in forecasting and understanding the dynamics of Shibor.
出处
《统计与信息论坛》
CSSCI
北大核心
2016年第5期21-26,共6页
Journal of Statistics and Information
基金
中央高校基本科研业务费专项资金项目<考虑结构突变的计量预测方法及其应用>(31920150119)