讲座题目：Promote  similarity of sparsity structures in integrative analysis

主讲人：马双鸽

讲座时间：2017年11月8日（周三）下午14:00-15:30

讲座地点：浙江工商大学综合楼601会议室

主讲人简介：1978年生，15岁以优异成绩考入中科大少年班，2004年7月获得美国威斯康辛大学统计博士学位，现任美国耶鲁大学生物统计系副教授。美国统计协会（American  Statistical Association）及国际统计协会（International Statistics  Institute）会士。主要研究领域为生物统计、半参数和非参数检验等。迄今为止，马双鸽博士在国际、国内期刊发表学术论文一百余篇，其中多篇论文发表于Annals  of Statistics、Journal of the American Statistical  Association、Biometrika等国际一流统计学学术期刊。

摘要：  

For multiple  high-dimensional problems, it is desired to conduct the integrative analysis of  multiple independent datasets. Under a few important scenarios, it can be  expected that the estimates of multiple datasets are “similar” in certain  aspects, which may include magnitude, sparsity structure, sign, and others. The  existing approaches do not have a mechanism promoting such similarity. In our  study, we conduct the integrative analysis of multiple independent datasets.  Penalization techniques are developed to explicitly promote similarity. The  consistency properties are rigorously established. Numerical studies, including  simulation and data analysis, show that the proposed approach has significant  advantages over the existing benchmark.