Identification of structural vector autoregrssive model based on time series chain graphs

In this paper, structural vector autoregressive model is expressed as time series chain graphs. We show the equivalence between the edge of the chain graph and the corresponding coefficient of the structural VAR, and discuss the links between partial covariance among residuals and partial covariance among contemporaneous components in a VAR model. A statistic based on local density estimator is proposed to test the conditional independence and a bootstrap approach is used to determine the null distribution of the test statistic. Simulations confirm the validity of the proposed method.