PP检验对异方差时间序列的伪检验

Spurious Tests to Heteroscedastic Time Series With PP Test

文章研究了PP检验是否判别方差突变时间序列的平稳性。通过构造均值方差都不变、均值突变、方差突变、方差递增、方差递减五个时间序列,应用时序图法、自相关图法和PP检验研究了这些序列的平稳性。研究结果表明,对于异方差时间序列,PP检验出现了伪检验。因此,对于异方差序列慎用PP检验,建议将方差突变列入结构突变的范畴。

This paper studies whether Phillips-Perron test（PP） identifies the stationarity of time series with variance mutation. By constructing such five time series as mean variance unchanged, mean variance mutation, variance mutation, variance increasing and variance decreasing, the paper uses the method of time series, autocorrelation and PP test to study the stationarity of these sequences. The results show that for time series with heteroscedasticity, the pseudo test appears in PP test. Therefore, PP test should be used with caution for heteroscedasal sequences. It is recommended to include variance mutation in the category of structural mutation.