Causal Inference by Identification of Vector Autoregressive Processes with Hidden Components

A widely applied approach to causal inference from a time series X, often referred to as “(linear) Granger causal analysis”, is to simply regress present on past and interpret the regression matrix B^