A data depth based nonparametric test of independence between two random vectors

Abstract

A new family of depth-based test statistics is proposed for testing the hypothesis of independence between two random vectors. In the procedure to derive the asymptotic distribution of the tests under the null hypothesis, we do not require any symmetric assumption of the distribution functions. Furthermore, a conditional distribution-free property of the tests is shown. The asymptotic relative efficiency of the tests is discussed under the class of elliptically symmetric distribution. Asymptotic relative efficiencies along with Monte Carlo results suggest that the performance of the proposed class is comparable to the existing ones, and under some circumstances, it has higher power. Finally, we apply the tests to two real data sets and also discuss the robustness of our tests.