We propose new data driven score rank tests for univariate symmetry about an unknown center. We construct test statistics, state assumptions and define estimators of nuisance parameters. We prove that the test statistics are asymptotically distribution-free under the null hypothesis. Using simulations, we verify these asymptotic results for finite samples and show that, under the assumptions and when they are somewhat violated, the size of the test is stable when changing the null distribution. We also compare the empirical behaviour of the new tests with those proposed in the literature. We show that for families of distributions commonly applied to model asymmetry the new tests overcome their competitors on average and for most individual alternatives.
2
Dostęp do pełnego tekstu na zewnętrznej witrynie WWW
We propose a modification of the data driven score rank tests studied recently in Inglot et al. (2012) by an appropriate choice of the orthonormal system. The simulation study confirms much better performance of the new tests for alternatives with dominating asymmetry in the tails and comparable sensitivity for other types of alternatives. In effect we obtain omnibus tests for symmetry which are equal to the best existing procedures for typical alternatives and overtake them significantly for atypical ones.
JavaScript jest wyłączony w Twojej przeglądarce internetowej. Włącz go, a następnie odśwież stronę, aby móc w pełni z niej korzystać.