报告题目：Renewable regression inference for streaming data with heterogeneous and periodic effects

报告人：陈雪蓉

邀请人：杨炜明

报告时间：11月12日，下午15:00

报告地点：启智楼80602

参加人员：学院师生代表

报告摘要：Sequentially collected business and economic data often carry stable, periodic, batch-specific covariate effects, yet most renewable estimation methods assume homogeneity, which breaks down under seasonal variation. We propose a renewable framework for generalized linear models with homogeneous and periodic effects. Given the period length, it integrates unknown coefficient-structure detection and renewable estimation using current-period data and historical summaries. Under standard regularity conditions, the homogeneous estimator is consistent, asymptotically normal and achieves full oracle efficiency; for periodic effects, we identify two convergence regimes driven by cross-period information and batch errors. Simulations verify accuracy and efficiency, and a U.S. airline study uncovers stable covariate links together with seasonal dependence between departure and arrival delays.

报告人简介：无