Generalized multivariate models for longitudinal data

Abstract

A general class of multivariate models is proposed for unbalanced and incomplete longitudinal data. The proposed model is an extension of the seemingly unrelated regression model (Zellner, 1962). The advantage of this model is discussed regarding its applicability to a larger class of problems and the ease of estimation. The application of the model includes the model for the time varying covariates proposed by Patel (1988) and growth curve models. Two estimation methods are considered; one method is the generalized least squares method based on Zellner's noniterative two-stage estimation and the other is the iterative maximum likelihood estimation method using the EM algorithm (Dempster,Laird,and Rubin,1977). Simulation studies are conducted to compare the small sample properties of the two estimators.

Keywords:

• Longitudinal data
• Missing data
• Repeated measures analysis
• Seemingly unrelated regressions