Granger因果性是衡量系统变量间动态关系的重要依据.本文利用图模型方法研究变量间的Granger因果性,建立了Granger因果图.基于条件互信息和置换检验法建立了Granger因果图结构的辨识方法,利用混沌理论中的关联积分估计相应的检验统计量,给出了统计量的渐进分布,并用所给方法研究国际主要股市间的Granger因果关系.
The Granger causality is an important criterion for measuring the dynamic relationship among system variables. In this paper, we apply the graphical model method to explore the Granger causal relations among variables. The Granger causality graph is established and its structural identification is investigated based on the conditional mutual information and permutation test. The test statistics is estimated using the correlation integral of chaos theory and its limiting distribution is proved. Finally, the Granger causality among main international stock markets is investigated using the proposed method.