基于多变量的Granger因果检验方法

The Granger Causality Test Method based on the Multiple Variables

针对既有的Granger因果检验方法只能处理两个变量间的因果关系问题,指出这种方法会导致间接因果与直接因果的混淆及由于数据同源而产生的伪因果问题,提出一种可以消除以上问题的多变量因果检验方法,该方法立足于原有的Granger因果检验,适用于短时相依的和平稳的时间序列数据,并根据蒙特卡罗方法给出了统计推断的检验量,设计了方法的实施步骤.最后,应用一个仿真的实例具体展示了方法的使用过程和方法的有效性。

The traditional Granger causality test could only deal with the causality relationship between two variables, so it would lead to the problem of confusing indirect cause with direct one and pseudo- cause according to homologous data. The paper proposed a new multi-variable causality test method to erase the above problems. The new method, which is proper for short-time dependent and stationary time series data based on original Granger causality, adopted the Monte Carlo method to give statistical inference and designed the method steps to implement. In conclusion, the method is proved to be effective through a stimulation practice and verification.