题目：Feature screening for ultrahigh-dimensional data (a case study of Spearman filter)
报告人：严晓东副教授
时间：2018年6月20日下午15:00
地点：实验楼东区2楼B-215（数字媒体实验室）
欢迎广大师生届时参加！
公共卫生学院、科技处
2018年6月15日
报告人介绍：
严晓东，山东大学经济学院副教授，硕士生导师。香港理工大学研究助理,于2017年获得云南大学博士学位; 2014.10-2014.12在香港中文大学访问；2015.7-2018.4 在香港理工大学做全职研究助理； 研究兴趣主要集中在高维数据的变量选择和特征筛选、缺失数据统计建模、贝叶斯局部影响分析、生存分析以及最近热门的子群分析、融合分析和深度学习。论文发表在诸如经济领域权威期刊《Journal of Econometrics》以及概率统计领域一流期刊 《Computational Statistics & Data Analysis》, 《Science China Mathematics》.
摘要(报告简介)：
In this presentation, we show a comprehensive overview of feature screening methods for dealing with ultrahigh-dimensional data, and highlight a newly proposed screening procedure called Spearman rank correlation method for ultrahigh-dimensional data with complete, censored, missing or categorical response cases, respectively . The proposed method is model-free without specifying any regression form of predictors and a response variable and it is invariant to monotone transformations of a response variable and predictors. The sure screening and rank consistency properties are established under some mild regularity conditions. The simulation studies demonstrate that the new screening method performs well in the presence of the heavy-tailed distribution or strongly dependent predictors or outliers and that it has the superior performance over the existing nonparametric screening procedures. In particular, the new screening method still works well when a response variable is observed under a high censoring or missing rate. And when the response is categorical, it can deal with categorical-adaptive screening procedure. An illustrative example is provided.
