报告题目: New adaptive designs and their statistical inference
报告人: Feifang Hu 教授 美国乔治华盛顿大学
报告时间:2021年1月11日 10:50-11:40
报告地点:Zoom 会议 (Zoom 会议id: 770 311 8512, 密码: 378548)
校内联系人:王培洁 wangpeijie@jlu.edu.cn
报告摘要: Covariate balance is one of the most important concerns for successful comparative studies (about causal effects), such as causal inference, online A/B testing and clinical trials. However, chance imbalance may still exist in traditional randomized experiments, and are substantially increasing in big data. In this talk, we discuss several new adaptive designs and their advantages. The proposed methods show substantial advantage over the traditional methods in terms of covariate balance and computational time. Since the randomization inevitably uses the covariate information when forming balanced treatment assignments, the variability of classical statistical inference following such randomization is often unclear in the literature. Further, we derive the theoretical properties of statistical inference post general covariate-adjusted randomization under the linear model framework. More importantly, we explicitly unveil the relationship between covariate-adjusted designs and inference properties. We apply the proposed general theorem to commonly used procedures and compare their performance analytically. These results open a new door to understand and analyze comparative studies based on covariate-adjusted randomization.
主讲人简介: 胡飞芳,美国乔治华盛顿大学统计系教授。1994年被授予加拿大英属哥伦比亚大学统计学博士学位,并获得加拿大优秀博士论文奖。2004年美国自然科学基金会杰出青年基金的得主。2009年当选为美国统计协会和国际数理统计协会双料fellow. 是自适应设计与方法的世界领先研究专家。他一直致力于统计理论及相关应用的研究。研究内容涉及自适应方法,生物统计,个性化医疗,临床实验设计,大数据分析,线上A/B 检测的实验设计与分析方法等。2000年以来国际顶级统计杂志发表学术论文80余篇。出版自适应设计与方法的英文专著1部,为实际应用提供了牢固的数学基础和理论指导。2007年受美国FDA 邀请撰写白皮书论文2篇。胡飞芳教授曾担任JASA 和 Annals of Statistics 等国际顶级统计杂志的副主编。胡教授主持六项美国国家自然科学基金研究项目,以及多项香港新加坡研究项目。近些年,他在自适应设计与方法的研究成果引起了同行的高度关注,多次受邀在国际学术会议上做主题报告。组织多次国际学术会议并担任会议共同主席。是多个临床试验的数据与安全监察委员会(DSMB)成员。同时受邀为多家世界五百强企业提供统计咨询。