TITLE: Targeted Machine Learning For Reliable Evidence Generation From Clinical Studies
SPEAKER: Mark van der Laan, UC Berkeley and Susan Gruber, TL Revolution
MODERATOR:Weili He
Abstract:
Targeted Learning (TL) unifies causal inference, machine learning (ML) and statistical theory to provide a framework for evaluating causal effects from clinical studies. Randomized controlled trials (RCTs) and real-world data (RWD) studies require adequate power, unbiased estimation, adequate control of the Type-I error rate, and good confidence interval coverage. TL provides a causal estimation roadmap that guides the design, analysis, and interpretation of clinical studies to address these challenges. Advances in ML play a crucial role in developing actionable evidence to support decision making.
The first half of the tutorial will use case studies to familiarize attendees with the TL causal estimation roadmap and practical application of targeted maximum likelihood estimation (TMLE) and super learning (SL), a general, flexible approach to ML.
The second half of the course covers three recent developments in targeted machine learning. The highly adaptive lasso (HAL) is the first general nonparametric MLE, offering superior rates of convergence and efficiency without traditional constraints of parametric modeling. The Adaptive (A)-TMLE is an estimator that can optimally combine RCT and external data. Unlike other data fusion methods, A-TMLE is guaranteed to be as efficient as an efficient estimator of the RCT data alone. The final innovation concerns deep learning for solving complex estimation problems concerning the effects of static and dynamic treatment strategies over time, adjusting for time dependent confounders.
Instructors’ Biography:
Mark van der Laan is the Jiann-Ping Hsu/Karl E. Peace Professor of Biostatistics and Statistics at the University of California, Berkeley. He has made contributions to survival analysis, semiparametric statistics, multiple testing, and causal inference. He also developed the targeted maximum likelihood methodology and general theory for super-learning. He is a founding editor of the Journal of Causal Inference and International Journal of Biostatistics.
He has authored 4 books on targeted learning, censored data and multiple testing, authored over 300 publications, and graduated 45 Ph.D. students. He received his Ph.D. from Utrecht University in 1993 with a dissertation titled “Efficient and Inefficient Estimation in Semiparametric Models”. He received the COPSS Presidents’ Award in 2005, the Mortimer Spiegelman Award in 2004, and the van Dantzig Award in 2005.
Dr. Susan Gruber, PhD, MPH, MS, is the founder of Putnam Data Sciences, a statistical consulting and data analytics consulting firm. Her work focuses on the development and application of data-adaptive methodologies for improving the quality of evidence generated by observational and randomized health care studies. Prior to forming Putnam Data Sciences, Dr. Gruber was the Director of the Biostatistics Center in the Department of Population Medicine at Harvard Pilgrim Health Care and Harvard Medical School, and former Senior Director of the IMEDS Methods program at the Reagan Udall Foundation for the FDA.