Simultaneous Variable Selection and Estimation for Analysis of Longitudinal Data Arising in Clusters Under Generalized Linear Mixed Pairwise Models

• Activity Number: 326
• Type: Topic Contributed
• Date/Time: Tuesday, August 5, 2014 : 10:30 AM to 12:20 PM
• Sponsor: Biometrics Section
• Author(s): Haocheng Li*+ and Grace Yi
• Companies: Texas A&M and University of Waterloo
• Keywords: Model misspecification ; Model selection ; Penalized pairwise likelihood

Abstract

Longitudinal data analysis is often challenged by complex data structures and large dimensionality of covariates. To provide the flexibility of modeling while retaining the feasibility of computation, we propose a class of models, generalized linear mixed pairwise models, to facilitate longitudinal data arising in clusters. To handle the high dimensionality of covariates for which only some of them are important, we propose a method to conduct simultaneous model selection and parameter estimation. Asymptotic properties of the proposed method are established, and the influence of model misspecification is explored. The method is applied to the Waterloo Smoking Prevention Project data, and is evaluated empirically by simulation studies.