Title: Interpoint-ranking sign covariance to test independence
Authors: Haeun Moon - University of Pittsburgh (United States)
Kehui Chen - University of Pittsburgh (United States) [presenting]
Abstract: An interpoint-ranking sign covariance is introduced, which is defined for general types of random objects with a meaningful similarity measure. We will show that it is zero if and only if the two random objects under consideration are independent. We will then introduce a test of independence based on the new interpoint-ranking sign covariance, and show that the proposed test is consistent against general types of alternatives. We will also present numerical experiments and data analysis to demonstrate the great empirical performance of the proposed method.