Subgroup Identification for Longitudinal Data with Unspecified Random Effects

• Activity Number: 435
• Type: Invited
• Date/Time: Wednesday, August 6, 2014 : 8:30 AM to 10:20 AM
• Sponsor: Korean International Statistical Society
• Author(s): Hyunkeun Cho*+ and Annie Qu and Peng Wang
• Companies: Western Michigan University and University of Illinois at Urbana-Champaign and Bowling Green State University
• Keywords: Classification tree ; Penalized quasilikelihood ; Quadratic inference function ; Personalized treatment ; Random-effects model ; Subgroup identification

Abstract

We develop new modeling and estimation for personalized treatment for individuals with high heterogeneity. Incorporating subject-specific information into treatment subgroup is critical since individuals could react to the same treatment quite differently. We estimate unobserved individual treatment effects through the conditional random-effects modeling, and identify the optimal treatment for individuals based on the random-effects estimation and subgroup analysis. The advantage of our approach is that the random effects estimation does not rely on the normality assumption, and is more efficient than the random-effect estimator which ignores correlation information from longitudinal data. In addition, the classification tree approach identifies subgroups with similar estimated individual treatment effects. We develop consistency and efficiency theory for the proposed random-effects estimator. Our simulation studies and a data example from AIDS clinical trial also confirm that the proposed method is efficient in identifying an effective treatment strategy for subgroups in finite samples.