On testing extreme value conditions (Q2463699)

Two tests are considered for testing the hypothesis that the distribution of an i.i.d. sample belongs to the domain of max-attraction of an extreme value distribution: the Dietrich\,-\,de Haan\,-\,Hüsler (E) test and the Drees\,-\,de Haan\,-\,Li (T) test. The quantiles of the limit distributions of the test statistics are calculated via Monte Carlo. The finite sample behaviour and power of the tests are assessed by simulated samples. The authors' conclusion is that if the extreme value index can be assumed to be positive, then the T-test is preferable. Otherwise the E-test is advisable.