Granger因果关系检验的适用性

研究了Granger因果关系检验在小样本下的适用性。采用Monte Carlo模拟方法,以1阶滞后模型为例,对其可能的27种数据生成过程进行模拟。从变量的平稳性、样本容量的大小、变量依存关系的显著性等方面,对其性质和适用性条件进行了较为全面的探讨。结果表明,序列的不平稳性是造成虚假因果关系最主要的因素之一;对于两个平稳序列,随着样本容量的增大,判断出存在Granger因果关系的概率将显著增大;当二者之间依赖关系增强时,判断出存在因果关系的概率也将显著增大。

Applicability of the Granger causality test

This paper studies the applicability of the Granger causality test for small samples. A Monte Carlo simulation was used for a first-lag model example with 27 possible data generation processes. The simulation was used to study the relations effects of various factors, including the stability of the series, the sample size, and the dependence between samples in the series. The results show that the instability of the series is one of the most important factors resulting in false causality. For two stationary series the probability of finding Granger causality increases with sample size and the significance of the dependence between series.