Change-Plane Analysis for Subgroup Detection with Continuous Treatment

• Activity Number: 372 - SPEED: SPAAC SESSION IV
• Type: Topic-Contributed
• Date/Time: Thursday, August 12, 2021 : 12:00 PM to 1:50 PM
• Sponsor: Biometrics Section
• Author(s): Peng Jin* and Wenbin Lu and Mengling Liu
• Companies: New York University Grossman School of Medicine and North Carolina State University and New York University Grossman School of Medicine
• Keywords: Continuous treatment; Doubly robust test; Heterogeneous treatment effects; Spline; Subgroup identification

Abstract

We propose a semiparametric change-plane model for testing and identifying a subgroup with heterogeneous treatment effects under the setting of a continuous treatment. The model includes nonparametric modeling of covariate effects and nonparametric heterogeneous treatment effects that vary within subgroups defined by a change plane. A doubly robust test statistic is constructed based on this model for detecting the existence of a subgroup, which is valid under a correct specification of either the baseline model for the covariate effects or the generalized propensity score model for the continuous treatment. When the null hypothesis of no subgroup is rejected, the change-plane parameters that define the subgroup can be estimated at the supremum of the normalized score statistic. The asymptotic distributions of the test statistic under the null and local alternative hypotheses are derived. Moreover, we further extend the proposed approach to handle time-to-event outcome and a more general setting that response follows an exponential family distribution.