Recovering the architecture of links in a chain of three unidirectionally coupled systems using the Granger-causality test

We propose a method for reconstructing links in a chain of unidirectionally coupled systems by means of three tests for estimating direct and mediated coupling with the aid of Granger conditional causality in terms of prognostic models with polynomial nonlinearity. It is shown that this approach allows the architecture of links in these systems to be correctly identified in more than 80% of cases.