期刊
COMPUTATIONAL STATISTICS & DATA ANALYSIS
卷 174, 期 -, 页码 -出版社
ELSEVIER
DOI: 10.1016/j.csda.2022.107476
关键词
Causal effect; Heterogeneity of treatment effect; Machine learning; Kernel methods; Personalized medicine
资金
- National Cancer Institute (NCI) [R01-CA129102]
- NSF [DMS 1914937]
In clinical trials, individuals in different subgroups may have different treatment effects, highlighting the importance of considering individual differences in treatment benefit. This study introduces the concept of predictive directions and validates its utility through simulation studies and real data analysis.
In many clinical trials, individuals in different subgroups may experience differential treatment effects. This leads to the need to consider individualized differences in treatment benefit. The general concept of predictive directions, which are risk scores motivated by potential outcomes considerations, is introduced. These techniques borrow heavily from the literature from sufficient dimension reduction (SDR) and causal inference. Initially directions assuming an idealized complete data structure are formulated. Then a new connection between SDR and kernel machine methodology for detection of treatmentcovariate interactions is developed. Simulation studies and a real data analysis from AIDS Clinical Trials Group (ACTG) 175 data show the utility of the proposed approach.(c) 2022 Elsevier B.V. All rights reserved.
作者
我是这篇论文的作者
点击您的名字以认领此论文并将其添加到您的个人资料中。
推荐
暂无数据