时 间：11月29日 10:00-11:00
地 点： 理科大楼A1514
腾讯会议ID：992-165-109 会议密码：1309
报告人：任好洁 副教授
主持人：刘玉坤 教授
摘 要：
Two-sample multiple testing has a widerange of applications. Most of the literature considers simultaneous tests ofequality of parameters. This talk takes a different perspective andinvestigates the null hypotheses that the two support sets are equal. Thisformulation of the testing problem is motivated by the fact that in manyapplications where the two parameter vectors being compared are both sparse,one might be more concerned about the detection of differential sparsitystructures rather than the difference in parameter magnitudes. Focusing on thistype of problems, we develop a general approach, which adapts the newlyproposed symmetry data aggregation tool combined with a novel doublethresholding (DT) filter. The DT filter first constructs a sequence
of pairs of ranking statistics that fulfillglobal symmetry properties, and then chooses two data-driven thresholds alongthe ranking to simultaneously control the false discovery rate (FDR) andmaximize the number of rejections. Several applications of the methodology aregiven including high-dimensional linear models and Gaussian graphical models.We show that the proposed method is able to asymptotically control the FDRunder certain conditions. Numerical results confirm the effectiveness androbustness of DT in FDR control and detection ability in many settings.
报告人简介：
任好洁是上海交通大学数学科学学院 长聘教轨副教授，2018年博士毕业于南开大学，随后在宾州州立大学从事博士后研究，导师是李润泽教授。研究方向包括统计异常探查、在线学习与监控、高维数据推断等。在JASA，Biometrika等杂志上发表论文10余篇。