主题： Online High-Dimensional Quantile Regression
主讲人：林楠 美国圣路易斯华盛顿大学
主持人：王国长 暨南大学
时间：2026年6月25日（周四）上午10:30-11:30
地点：暨南大学石牌校区经济学院（中惠楼）306室
摘要
Online high-dimensional quantile regression is important for streaming data with heavy-tailed or heterogeneous outcomes, but it is also challenging due to the nonsmooth loss, high dimensionality and sequential data structure. Our proposed method avoids repeatedly solving quantile regression problems at each data block by using the asymmetric Laplace representation to transform the local update into a least-squares problem. A ridge penalty is further incorporated to accommodate high-dimensional covariates, leading to a scalable and computationally efficient online algorithm. Under mild conditions, our proposed estimator attains the statistical efficiency of the pooled estimator while avoiding repeated optimization over all historical data. We further extend our framework to sequential quantile treatment effect (QTE) estimation in the presence of high-dimensional confounders. We use an orthogonal estimating equation formulation to reduce the impact of nuisance estimation errors on the QTE estimator and enables valid sequential estimation of QTE in genuinely high-dimensional online settings.
主讲人简介
林楠，圣易路丝华盛顿大学统计与数据科学系教授，博士生导师。博士毕业于 University of Ill inois at Urbabna-Champaign ,在耶鲁大学从事博士后研究。主要研究兴趣包括大数据的统计计算方法、分位数回归和因果推断等，现任国际统计权威期刊JCGS和 Interational STatistical Review的副编辑。
校对 |王国长
责编 | 彭毅
初审 | 姜云卢
终审发布 | 何凌云
（来源：暨南大学经济学院微信公众号）
