Inference on average treatment effect with high dimensional covariates
报 告 人:: 何旭铭
报告地点:: 数学与统计学院403室
报告时间:: 2017年10月18日星期三16:00-17:00

Classical inference in statistics is typically model-based, but modern data analysis often uses model selection as part of the protocol. We discuss the challenges in statistical inference on the average treatment effect when a regression model has to be selected in the presence of many possible covariates. We discuss the sources of bias and review some of the recent work in post-selection inference. More specifically, we study the properties of repeated data splitting as a valid inferential tool.

发 布 人:科研助理 发布时间: 2017-10-17
教授,博士生导师,教育部“长江学者奖励计划”讲座教授,2016年入选国家“千人计划”创新人才短期项目,受聘为应用统计教育部重点实验室学术委员会主任。毕业于伊利诺伊大学厄巴尼-尚佩恩分校 (UIUC),美国统计学会院士,也是国际数理统计学会、美国科学进步学会以及国际统计学会当选会员。