On the estimation of the direction of information flow in networks of dynamical systems

The inference of the interaction structure in networks of dynamical systems promises novel insights into the functioning or malfunctioning of systems in the neurosciences. This may improve the understanding of mechanisms underlying several diseases like tremor disorders and might eventually help to cure patients. Of particular interest is the estimation of the direction of information flow for which different methods have been suggested and have been applied to data from human tremor. Based on a simulated system motivated by the human tremor application we analyze the performance of three methods. The abilities and limitations of the individual techniques are compared and discussed. An application to essential tremor complements this investigation.

Research highlights▶ Comparison of techniques to infer the direction of information flow. ▶ Only partial directed coherence was able to detect all interactions present reliably. ▶ Application to real-world data from patients suffering from essential tremor.