Estimation and Model Selection in a Class of Semiparametric Models for Longitudinal Data

• Activity Number: 610
• Type: Topic Contributed
• Date/Time: Thursday, August 5, 2010 : 8:30 AM to 10:20 AM
• Sponsor: WNAR
• Author(s): Jialiang Li*+
• Companies: National University of Singapore
• Keywords: semivarying-coefficient models ; local linear modelling ; cluster level variable ; cluster effect ; cross-validation

Abstract

Stimulated by the data set from Bangladesh about the first birth interval, we propose a semivarying-coefficient model for cluster data analysis which takes both stochastic and deterministic effects into account for the difference across clusters. The model can be viewed as an extension of many familiar statistical models for cluster data analysis, such as growth curve models and multi-level models. Our functional formulation of the fixed and random effects allows a flexible parameterization for the regression components. We investigate the estimation procedure for the proposed model, and establish the asymptotic properties. Model selection is also examined by using cross-validation. The proposed estimation and model selection methods are used to analyse the data set mentioned above to explore how individual-level and cluster-level factors affect the first birth interval in Bangladesh.