Granger Causal Networks and Indirect Feedback

Vector Autoregressive (VAR) models struggle to differentiate between direct, indirect, and aggregate feedback among variables in a system. This analysis introduces causality network graphs as a solution to visualize and understand these complex structural relationships. By using rules like pairwise Granger causality to define connections, these graphs can map out both direct causal links and intermediate feedback paths. An example using financial data demonstrates how this technique uncovers structural causality that simple correlation analysis misses, providing a more intuitive framework for decomposing aggregate effects in multivariate systems.