|
Abstract:
|
Precision medicine, the paradigm of improving clinical care through data driven approaches to tailoring treatment to the individual, is an important area of statistical and biomedical research. Individualized treatment rules (ITR's) formalize precision medicine as mappings from the space of patient covariates to the set of available treatments; ITR's can be used to improve patient outcomes by utilizing biomarkers to target treatment. The robustness of direct estimators of the optimal ITR has led to their increasing popularity in recent years. Outcome weighted learning, or O-learning, is one of the first and most widely used direct estimators. We propose a unifying framework wherein O-learning can be viewed as a semiparametric maximum likelihood estimator. We use this framework to study the theoretical properties of different variants of O-learning and develop novel estimation methods, variable selection techniques, and exploratory data analysis techniques for ITR's.
|
