报告时间：2023年12月22日(周五)16:00-17:00

报告地点：中国人民大学明德主楼1016

报告嘉宾：俞章盛

报告主题：Deep Neural Network for Partially Linear Competing Risk Models

报告摘要

Deep Neural Network for Partially Linear Competing Risk Models

Deep neural networks have demonstrated excellent performance in the field of nonparametric estimation in statistics. However, in subdistribution hazard model for competing risk data, there have been few attempts to extend the linear form of covariates to nonparametric or semiparametric forms. Therefore, this paper proposes the deep partially linear Fine-Gray model (DPLFGM), which utilizes deep neural networks (DNN) to estimate the nonlinear component to circumvent curse of dimensionality issue. To evaluate the predictive performance of the model in more detail, we further develop a time-dependent AUC method specifically tailored for competing risk data and establishes its relationship with the C-index. Theoretical results demonstrate the asymptotic normality of the parameter component at a rate of root n and provide the convergence rate of the nonparametric component, which achieves the minimal limit convergence rate (multiplicative logarithmic factors). Subsequently, the paper validates the excellent performance of DPLFGM in estimation and prediction through numerical simulations and real-world datasets.

作者简介

俞章盛教授目前为上海交通大学生物统计学教授，交大医学院临床研究中心主任。研究领域为生物统计学，在生存分析理论方法研究及应用方面获得一系列创新成果，在包括Biometrika, Nature Communications 等期刊上发表 SCI 论文100 余篇。担任包括 Q1 区杂志 Statistics in Medicine 等四个期刊副主编/编委，上海市优秀学术带头人，上海市海外高层次人才。作为统计学家受聘交大医学院直属部门临床研究中心主任，开展医学交叉研究，在医学四大刊 JAMA 等发表论文；作为首席方法学专家与眼科专家合作制定眼病诊疗指南。