统计与数据科学系系列学术报告之四百八十五期

时间： 2025年9月26日（周五）10:45-11:30

主持人：复旦大学管理学院统计与数据科学系沈娟副教授

地点：史带楼303室

报告人：马慧娟副教授

华东师范大学

题目：Efficient estimation of quantile treatment effects by incorporating external controls

摘要：Quantile treatment effect (QTE) is increasingly recognized as an essential tool in understanding the heterogeneous impact of interventions across different segments of the outcome distribution and providing a richer perspective beyond average treatment effect. Despite its importance, obtaining precise QTE estimates can be challenging in randomized clinical trials with limited sample sizes. External controls have grown in popularity in recent years as a tool to empower the clinical trials. In this paper, we propose a novel approach that combines clinical trial data with external controls to enhance the efficiency of QTE estimation. By leveraging external controls, we first derive the efficient influence function of the QTE estimand and then propose a sieve-based weighted M-estimator that achieves the semiparametric efficiency bound. The proposed estimator is shown to be more efficient than the one solely using trial data. We also extend the proposed framework to other QTE estimands defined in different populations and provide an alternative approach for more efficient estimation of average treatment effects. Extensive simulation studies and application to the job training data demonstrate the practical utility of the proposed methods.

个人简介：马慧娟，华东师范大学统计学院副教授。主要研究方向包括生存分析，分位数回归，因果推断等。在统计学期刊 JASA, Biometrika, Biometrics等期刊发表论文二十余篇。先后主持上海市浦江人才项目，国家自然科学基金青年项目、面上项目和重点项目子项目等。参与国家自然科学基金重点项目、科技部重点研发项目和上海市科委重点项目等。