SMU SOE Seminar (July 24, 2019): Optimal Dynamic Treatment Regimes with Imperfect Compliance

Please click here if you are unable to view this page.

TOPIC:

OPTIMAL DYNAMIC TREATMENT REGIMES WITH IMPERFECT COMPLIANCE

ABSTRACT Dynamic treatment regimes are treatment assignments tailored to heterogeneous individuals’ needs. In a dynamic regime, each period’s assignment rule maps past outcomes and treatments onto a current allocation decision. The optimal dynamic treatment regime is a sequence of assignment rules that maximizes average welfares. This paper considers identiﬁcation of optimal dynamic regimes when observed outcomes and treatments are generated from multi-period experiments in the presence of non-compliance, or more generally from observational studies. The biostatistics literature that studies optimal treatment regimes relies on the sequential randomization assumption, although non-compliance is prevalent especially in multi-period settings. This paper relaxes sequential randomization and proposes a simple nonparametric structural models, in which we can partially learn optimal dynamic regimes via exclusion restrictions. We ﬁrst characterize the identiﬁed set of optimal regimes as a subset of all possible regimes, and calculate resulting bounds on the optimal welfares and regrets. We then show how additional variation in exogenous variables helps shrink the identiﬁed set of the objects of interest. In the paper, we also discuss the cardinality reduction of the set of possible regimes for computational, institutional, and practical reasons. Keywords: Optimal dynamic treatment regimes, endogenous treatments, non-compliance, average treatment effect, instrumental variable. JEL Codes: C14, C32, C33, C36. Click here to view the paper. Click here to view the CV.
		PRESENTER Sukjin Han University of Texas at Austin
		RESEARCH FIELDS Econometric Theory Applied Econometrics Machine Learning Visual Data
		DATE: 24 July 2019 (Wednesday)
		TIME: 4pm - 5.30pm
		VENUE: Meeting Room 5.1, Level 5 School of Economics Singapore Management University 90 Stamford Road Singapore 178903

© Copyright 2024 by Singapore Management University. All Rights Reserved.