A novel mutual association measure based on eigenvalue-based criterion

Abstract

The study of association has been one of the domain subjects for the research of multivariate system, which is also known as relationship detection, correlation analysis or dependency identification for a number of different disciplines like economics, biology, chemistry, image or signal processing, etc. The significant effects of association study indicate that the development and improvement of association measure can directly influence the accuracy and possibility of identifying possible relationships among random variables in multivariate system. This paper aims at seeking novel association measure by combining eigenvalue-based criterion. We conduct a brief review of the well established association measures to date, also the formulation process of the novel method is proposed. Furthermore, the novel association measure is compared with the empirical methods by both simulations and a case of real data. The robust performance of this novel method is proved with consistently significant outcomes achieve...