An Overview on Nonparametric and Semiparametric Techniques for Longitudinal Data

Abstract

Longitudinal data are often highly unbalanced because data were collected at irregular and possibly subject-specific time points. It is difficult to directly apply traditional multivariate regression techniques for analyzing such highly unbalanced collected data. This has led biostatisticians and statisticians to develop various modeling procedures for longitudinal data. Parametric regression models have been extended to longitudinal data analysis (Diggle, et al. 2002). They are very useful for analyzing longitudinal data and for providing a parsimonious description of the relationship between the response variable and its covariates. However, the parametric assumption likely introduce modeling biases. To relax the assumptions on parametric forms, various nonparametric models have been proposed for longitudinal data analysis. Earlier works on nonparametric regression analysis for longitudinal data were summarized in Muller (1988). Kernel regression was applied to repeated measurements data with continuous re-