Causal coupling inference from multivariate time series based on ordinal pattern transition networks

Identifying causal relationships is a challenging yet a crucial problem in many fields of science like epidemiology, climatology, ecology, genomics, economics and neuroscience, to mention only a few. Recent studies have demonstrated that ordinal partition transition networks (OPTNs) allow to infer the coupling direction between two dynamical systems. In this work, we generalize this concept to the interaction between multiple dynamical systems and propose a new method to detect causality in multivariate observational data. We demonstrate that our approach can reliably identify the direction of interaction and the corresponding delays with numerical simulations using linear stochastic systems as well as nonlinear dynamical systems such as a network of neural mass models. Finally, we apply our method to real-world observational microelectrode array data from rodent brain slices to study the causal effect networks underlying epileptic activity. Our results from simulations as well as real-world data suggest that OPTNs can provide a complementary approach to reliably infer causal effect networks from multivariate observational data.