Generalized Semiparametric Varying-Coefficient Regression Models for Longitudinal Data with Applications to Treatment Switching

• Activity Number: 431
• Type: Contributed
• Date/Time: Tuesday, August 11, 2015 : 2:00 PM to 3:50 PM
• Sponsor: Biometrics Section
• Author(s): Li Qi* and Yanqing Sun and Peter B. Gilbert
• Companies: and The University of North Carolina at Charlotte and Fred Hutchinson Cancer Research Center
• Keywords: Longitudinal biomarkers ; AIDS clinical trials ; Local linear smoothing ; exposure variable ; Treatment effects ; GLM

Abstract

This paper investigates a generalized semiparametric varying-coefficients model for longitudinal data that can flexibly model three types of covariate effects: constant effects, time-varying effects, and covariate-varying effects, i.e., the covariate effects that depend on other possibly time-dependent exposure variables. The proposed model assumes that time-varying effects are unspecified functions of time and the covariate-varying effects are a parametric function of an exposure variable specified up to a finite number of unknown parameters. Local linear smoothing and weighted least squares estimation techniques are used to estimate these effects. The asymptotic distributions of the proposed estimators are established and extensive simulation studies are studied. A working formula for bandwidth selection is discussed and examined through simulations. The proposed method is applied to the ACTG 244 AIDS clinical trial to examine the effects of treatment switching before and after developing 215-mutation. Our analysis shows the benefit of treatment switching before developing 215-mutation.