University of Alberta Longitudinal Data Analysis in Social

This project discusses the Generalized Estimating Equation (GEE) model and its application for longitudinal data. For many longitudinal studies involving a discrete (count or categorical) response variable, and for some studies involving a continuous response variable, it is difficult to justify a fully parametric model for the data. In these circumstances, the method of GEE provides consistent estimates for the mean parameter when a model for the correlation may not be reliably specified. In the second chapter, we begin by discussing Generalized Least Squares (GLS) and then give the theory of the GEE and the methodology of fitting GEE model. GEEs belong to a class of semiparametric regression techniques as they rely on specification of only the first two moments. Under mild regularity conditions, parameter estimates from GEEs are consistent. They are a popular alternative to the likelihood based generalized linear mixed model which is more sensitive to variance structure specification. They are frequently applied in multi-site cohort studies as they can handle many types of unmeasured dependence. The third chapter discusses project background and the questions of interest, how the dataset is simulated for analysis is presented in Appendix A written by John Simpson. Finally, this project discusses the application and comparison of GLS and GEE on the Social Science Data, and conclusion is given in the last section.