Independence test based on empirical copula does not perform well in the presence of weak dependency or when dependency occurs only in the tails. The copula density, which is estimated by a local likelihood pr obit transformation method, is used to detect the independence. In this paper, three non-parametric tests of independence based on α-divergence and copula density are introduced. These tests are capable of considering weak dependency. The asymptotic consistency of the copula-based α-divergence estimator is also derived. In addition, the empirical powers of the proposed tests are computed through extensive simulations. The results show that the new tests outperform in small sample sizes or weak dependencies. Finally, an application in uranium exploration is presented to illustrate the applicability of the proposed tests.

Keywords