基于广义交叉谱密度方法的线性和非线性格兰杰因果关系检验

A Test for Linear and Nonlinear Granger Causality based on Generalized Cross Spectral Density

鉴于现行的均值格兰杰因果关系检验或者无法检验非线性的格兰杰因果关系,或者存在"维数灾难"问题,我们根据Chung和Hong(2007)的广义交叉谱方法提出了一个能统一检验线性和非线性均值格兰杰因果关系的检验统计量。本文的广义交叉谱检验统计量渐近服从一个标准正态分布,它不但能考虑所有滞后阶的信息,而且避免了"维数灾难"问题。蒙特卡罗试验结果表明广义交叉谱检验具有良好的有限样本表现。

The current tests for Granger causality in nonlinear Granger causality, or have the problem of curse mean can either miss the of dimensionality. Therefore, based on the generalized cross spectral method by Chung and Hong （2007）, we propose a unified test for both linear and nonlinear Granger causality in mean. The generalized cross spectral test statistic follows a standard normal distribution asymptotically, and it not only uses the information of all lag orders, but also avoids the problem of curse of dimensionality. The Monte Carlo simulation results show that the generalized cross spectral test has good finite sample performance.