Abstract

In this article, we propose a flexible weighted test of independence based on Cramér-von Mises statistic of the Bernstein empirical copula process. We investigate the asymptotic properties of the new weighted test. A Monte Carlo study is conducted to measure the power performance of the new test under various dependence structures and also to compare the performance with some other alternative tests. The effects of the weights and the selection of optimal Bernstein polynomial degree are significantly influencing the power of the test according to the particular type of dependence. Finally, the procedure is also applied to a dataset on diabetes for illustrative purpose.

摘要

在本文中，我们基于伯恩斯坦经验联结过程的 Cramér-von Mises 统计提出了一种灵活的独立性加权检验。我们研究了新加权检验的渐近特性。进行蒙特卡罗研究以测量新测试在各种依赖结构下的功率性能，并将性能与其他一些替代测试进行比较。根据特定类型的相关性，权重的影响和最佳 Bernstein 多项式次数的选择显着影响检验的功效。最后，出于说明目的，该程序还应用于糖尿病数据集。