Notes to Time Reverse Methods for Detecting Causality of Dynamical Systems

Abstract

Inferring causality between two processes is a complex problem with many applications in meteorology, economics, and other fields. Dynamical systems are a useful tool for modeling many real-world processes. The subject of this work is the detection of causality between two dynamical systems. Causality detection is still an open problem and new methods keep emerging. A recently popular approach to causality detection is based on inspecting causality on reverse time series. This paper focuses on one phenomenon that arises when using methods based on reverse time series for nonlinear dynamic systems, which can lead to misleading results.