LEARNING OF MULTIVARIATE TIME SERIES GRANGER CAUSALITY BASED ON GRAPHICAL MODEL METHODS
Tian Zheng · Xitong kexue yu shuxue · 2011
Traditional two-variable Granger causality analysis method is prone to inducing spurious causal relationship and cannot portray the immediate causal relationship.This paper explores how to use graphical model methods to analyze the Granger causality graphs among components of multivariate time series.Granger causality graphs of time series is presented and the structural identification problem of Granger causality graph is investigated.A statistic based on local density estimator is proposed,and a bootstrap methods is considered for determining the null distribution of the test statistic.The validity of the proposed method is confirmed by simulations analysis and investigating the Granger causal relationships of the China's stock market.