摘要
由于非平稳的时间序列不具有有限方差,高斯-马尔科夫定理不再成立,用普通最小二乘法得到得参数估计不再是一致的,出现伪回归的现象,从而导致得出错误的因果关系。针对时间序列的非平稳性,指出目前进行Granger因果分析时所存在的问题,指出对非平稳变量进行因果分析的两种正确方法,并对这些方法的适应范围和优劣性进行比较。
Because the nonstationary time series don't have limited variance and it can't accord with Gauss - Markov Theorem, the Ordinary Least - Squares Estimators are inconsistent and then the spurious regression occurs, thus the incorrect causality can be drawn. Aiming at the nonstationarity of time series, the problems existing in handling the Granger Causality Test have been pointed out, the two correct methods about the Granger Causality Test of nonstationary variables have been brought forward, at the same time the adaptable scale and advantage of them have been compared.
