Speaker: Gang Li (UCLA)

Time: June 22, 2024  10:00-12:00

Abstract:

My talk will start with an introduction to UCLA’s Biostatistics program, followed by a technical presentation on prediction performance measures for survival data. Evaluating the performance of a prediction function is a fundamental task in statistics and machine learning. However, despite the availability of numerous prediction performance measures, there is no consensus on the best one to use, especially when dealing with censored time-to-event data. In this talk, I will illustrate that many common prediction performance measures may fail to distinguish the prediction performance between different models using a well-known clinical trial data. I will then introduce a new pseudo R-squared statistic, which extends the classical R-squared statistic to any nonlinear prediction function and to right-censored time-to-event data. The pseudo Rsquared statistic is obtained from a pair of orthogonal prediction performance measures based on a variance decomposition and a prediction error decomposition. Its effectiveness will be demonstrated using simulations and various real world data examples. Extension to time-dependent performance measures and competing risks data will also be discussed. 

Biography:

Dr. Gang Li is a Professor of Biostatistics and Computational Medicine at the University of California, Los Angeles (UCLA). He also serves as Director of the UCLA Health Jonsson Comprehensive Cancer Center Biostatistics Shared Resource. Dr. Li is an Elected Fellow of the Institute of Mathematical Statistics, the American Statistical Association, and the Royal Statistical Society, as well as an Elected Member of the International Statistical Institute. Among the many significant roles he has held in the statistical profession, Dr. Li serves as the co-Editor-in-Chief (2022-2024) of the Electronic Journal of Statistics, published by the Institute of Mathematical Statistics and the Bernoulli Society. Additionally, he has served as the President (2022-2024) of the International Chinese Statistical Association. Dr. Li’s research encompasses a broad range of areas, including survival analysis, longitudinal data analysis, high-dimensional data analysis, clinical trials, statistical learning, causal inference, and high-performance statistical computing for large-scale electronic health records (EHR) and biobank data. He has made significant contributions to these fields, co-authored/edited three research monographs, and published over 150 peer-reviewed papers, many of which are featured in renowned journals such as the Annals of Statistics, the Journal of the American Statistical Association, and the Journal of the Royal Statistical Society-B. In addition to his methodological research, Dr. Li actively engages in collaborative research in basic science, translational science, and clinical trials. He has served as the Principal Investigator for numerous studies funded by the National Institutes of Health (NIH) and the National Science Foundation (NSF