题目：Estimating Heterogeneous Treatment Effect at High Quantiles for Heavy-Tailed Distributions

汇报人：唐炎林

会议时间：2025年3月21日(周五)  14：30-16：00

地点： 综合楼644会议室

报告人简介：唐炎林，华东师范大学统计学院教授，博士生导师，统计学系主任；国家高层次人才计划入选者、上海市浦江人才计划。2012年1月博士毕业于复旦大学统计系，同年5月加入同济大学，2019年1月加入华东师范大学。主要研究方向为分位数回归、高维统计推断、不完全数据统计建模，主持多项国家自然科学基金、上海市自然科学基金，担任SCI期刊《Statistica Sinica》、《Journal of the Korean Statistical Society》的编委。在《Biometrika》、《JRSSB》、《PNAS》、《Biometrics》等发表论文40余篇。

摘要：Extreme events, such as heavy rainfall, low infant birth weight, and large financial loss, are rare but have significant consequences. In this talk, we will introduce the estimation of the heterogeneous treatment effect on these extreme events in observational studies through extreme quantile regression. Assuming that the response variable is heavy-tailed and the conditional quantiles are linear in the covariates at tail quantiles, we first estimate the conditional intermediate quantiles of both the treatment group and control group in a conventional potential outcome framework, and then extrapolate these estimators to the high tails. Based on the tail characteristics of the conditional and marginal distributions of potential outcomes, we develop two Hill-type estimators for the extreme value index (EVI). We present a comprehensive study of the theoretical properties, including the consistency and asymptotic normality of the proposed estimators of the conditional extreme quantile treatment effect and the EVI. We also provide a distributed algorithm for the extreme quantile treatment effect in large scale data, based on convolution smoothing.