报告题目：Estimating the causal treatment effect in multi-site current status data with the additive hazards model and neural networks

报 告 人：李树威 教授 广州大学

报告时间：2026年1月6日 10：00-11：00

报告地点：腾讯会议737928483

校内联系人：王培洁 [email protected]

报告摘要：Inverse probability weighting offers a valuable tool to eliminate the impact of endogenous treatment selection and attain unbiased causal treatment effect estimation in observational studies. In practice, to improve the estimation efficiency, researchers are often advocated to conduct sensible integrative analysis with multi-site studies. In such setting, how to avoid utilizing individual-level data directly is a key concern due to privacy concerns, regulatory constraints or other reasons. This work concerns multi-site current status data and provides an inverse probability weighted estimator of causal treatment effect with the additive hazards model. In particular, we develop a two-stage distributed estimation approach involving artificial neural networks and a combined estimating equation that mainly leverages summary statistics provided by each site. Asymptotic properties of the proposed estimator, including root-n consistency and asymptotic normality, are established. Extensive simulation studies demonstrate that the proposed method can reasonably adjust the endogenous treatment selection and is comparable to the causal method based on pooled individual-level data regarding estimation accuracy and efficiency. Moreover, the proposed estimator is more efficient than that of a single data-contributing site, manifesting the practical utility of using integrative analysis. An application to a real world data set is also provided.

报告人简介：李树威, 统计学博士, 现任广州大学经济与统计学院教授、统计系主任。2017年6月博士毕业于吉林大学统计系。主要研究方向为生物统计、大数据处理及机器学习，在Biometrika、 Biometrics、Statistics in Medicine、Statistica Sinica、JCGS等期刊上发表论文30余篇。主持国家自然科学基金面上项目、国家自然科学基金青年项目等。