Year: 2021 Vol.: 70 No.: 2
Authors: Mohammad Bolbolian Ghalibaf
Abstract:
This paper proposes two new tests for tail independence in extreme value models. We use the conditional distribution function (df) of X + Y, given that X + Y > c based approach of Falk and Michel to test for tail independence in extreme value models. We recommend using Cramervon Mises and Anderson-Darling tests for tail independence. Simulations show that the two tests are better than the Kolmogorov-Smirnov test which has good results among the proposed tests by Falk and Michel. Finally, by using two real datasets, we illustrate the application of the two proposed tests as well as the traditional tests of Falk and Michel.
Keywords: extreme value model, tail independence, Copula function, Cramer-von Mises test, Anderson-Darling test, Neyman- Pearson test, Kolmogorov-Smirnov test, Fisher’s ? test, Chi-square goodness-of-fit test