朱利平教授学术报告

发布时间:2016年01月04日 作者:   消息来源:    阅读次数:[]

报告题目: A post-screening diagnostic study in sufficient dimension reduction for ultrahigh dimensional data 

报告人:人民大学统计与大数据研究院 朱利平教授 

时间:2016年元月7日下午4:00-5:00 

地点:数理楼145报告厅 

摘要:In this talk I will introduce a consistent lack-of-fit test to examine whether or not replacing the original ultrahigh dimensional covariates with a given number of linear combinations will result in loss of regression information. To attenuate spurious correlations which are often seen in ultrahigh dimensional covariates and may substantially inflate type-I error rates, we suggest to randomly split the observations into two halves. In the first halve of observations we screen out as many irrelevant covariates as possible. This helps us reduce the ultrahigh dimensionality to a moderate scale. In the second halve we perform a lack-of-fit test for conditional independence within the context of sufficient dimension reduction. This data-splitting strategy helps us retain the type-I error rate pretty well. We propose a new statistic to test conditional independence, and show that our propose test procedure is n-consistent under the null and root-n-consistent under the alternative hypothesis. Our proposed test procedure is consistent in the sense that it has nontrivial power against all feasible alternatives. In addition, we suggest a bootstrap procedure to decide critical values and show that our bootstrap procedure is consistent. We demonstrate the effectiveness of our test procedure through comprehensive simulations and an application to the rats red-eye data set. 

朱利平教授: 

华东师范大学取得博士学位,入选教育部新世纪优秀人才计划、中组部青年拔尖人才计划等,并获得国家自然科学基金委员会优秀青年基金项目资助,现任中国人民大学统计与大数据研究院教授。朱利平博士一直从事统计理论、方法与应用研究,研究兴趣涉及半参数建模、高维数据分析、充分降维以及变量选择等领域。在统计学重要学术期刊发表论文近60篇,其中有10篇以上的论文发表在国际统计学科四大顶级学术期刊上。曾为Annals of Statistics副主编,现在是国际统计著名杂志Statistica Sinica的副主编。 



下一篇:学术报告

打印】【收藏】 【关闭