Transfer Entropy Estimation via Copula
Xue Tian · 2017
Transfer entropy provides a powerful information theoretic measurement of directed information flow between time series variables.Effective and convenient methods of estimation are desirable in practice.This article discusses the formulation of how to estimate transfer entropy via the statistical copula.Furthermore, this article provides theoretical justifications, and two estimation approaches via the Gaussian copula transformation and kernel methods.The experiment demonstrates that the proposed estimation approaches are competitive with the Linear estimator and the Nearest Neighbour estimator.