许王莉 | One Nonparametric Rank Test for Equality of Two Distributions

发布者:钱琳发布时间:2019-06-21浏览次数:38

时间:2019年6月21日(周五)上午10:00-11:00

地点:中山北路校区理科大楼A1514会议室

题目:One Nonparametric Rank Test for Equality of Two Distributions

报告人:许王莉  中国人民大学 教授

摘要:The Wilcoxon rank-sum test is one of important tests for homogeneity of two-sample in univariate case. But it lose power when there is no location shift. In this paper, we first introduce a rank definition from the pairwise distance of data sample, and propose a nonparametric ranked test statistic for testing homogeneity of two-sample. We illustrate that it is feasible to depict the diver-gence between two density functions with it. Its limits behavior under the null and alternative hypothesis can be derived. Specifically, the statistic under null hypothesis is asymptotically a mixture of chi distributions, while it converges to a normal distribution under the alternative hypothesis. The asymptotical property is also investigated under the local alternative hypothesis. In addition, we study its power performance for fixed alternative hypothesis, it is consistent against all alternative hypothesis regardless of the proportion of small sample. For practice application, we introduce a permutation test to calculate the p-value. The numerical efficiency gain and its superior power performance is further confirmed through extensive simulations and real-world applications.

报告人介绍:

许王莉,2006年7月毕业于中国科学院数学与系统科学研究院应用所概率论与数理统计专业。现任中国人民大学统计学教授,医学与生物统计教研室主任,统计学院学术委员会副主任,中国人民大学教学督导专家。目前是中国现场统计研究会生存分析分会副秘书长、国际生物统计学会中国分(IBS-CHINA)青年理事,中国现场统计研究会高维数据统计分会理事和实验设计分会理事,北京应用统计学会理事,2017年入选首批中国人民大学杰出青年学者A岗。2010年先后入选“新世纪优秀人才计划”和“北京市科技新星计划”。近年来一直从事模型拟合优度检验,高维数据分析,随机缺失数据,两阶段抽样数据以及纵向数据分析等方面的统计推断研究。先后主持了国家自然科学面上基金,国家自然科学青年基金,教育部人文社会科学重点研究基地重大项目和教育部人文社科基金等多项科研课题, 在统计学国际一流期刊(包括顶尖期刊)发表论文60余篇,其中SCI论文52篇,并在科学出版社合作出版《非参数蒙特卡洛检验及其应用》和单著《缺失数据的模型检验及其应用》。