学术报告：Covariate-adaptive design: An overview and recent advances

报告时间：11月1日（星期五）上午10:00-11:30

报告地点：沙河校区，二教107

报告人：马维，中国人民大学统计与大数据研究院，长聘副教授

报告摘要：Covariate-adaptive designs are a class of experimental design methods that dynamically adjust treatment allocation probabilities to achieve balanced covariates across treatment groups. Because of their strengths in enhancing treatment group comparability, increasing the precision of treatment effect estimation, and producing more convincing experimental results, these designs are extensively employed in randomized controlled settings, including clinical trials, economic field experiments, and online A/B testing. This talk first provides a methodological review of various covariate-adaptive design approaches and then discusses a recent advancement in the field, which proposes a novel and unified framework for covariate-adaptive designs. The challenges and solutions in analyzing data collected from covariate-adaptive designs will also be addressed. This talk is primarily based on the works of Ma, Ye, Tu, and Hu (2023), Ma, Li, Zhang, and Hu (2024), Ma, Tu, and Liu (2022), and Liu, Tu, and Ma (2023).

报告人简介：马维，中国人民大学统计与大数据研究院长聘副教授、博士生导师，国家级青年人才计划入选者，国际统计学会推选会员(ISI Elected Member)。2009年本科毕业于浙江大学数学系，2014年博士毕业于美国弗吉尼亚大学统计系。在Journal of the American Statistical Association、Biometrika、Biometrics等期刊发表多篇学术论文。担任Statistica Sinica副主编，中国现场统计研究会试验设计分会理事、因果推断分会理事。

撰稿人：刘洁

审稿人：邓露