MSc. Stats Project Presentation: Methods for Clustered Longitudinal Data

CANDIDATE:   GUYAN YANG

ABSTRACT:

Clustered longitudinal data arise when longitudinal responses are collected from subjects within different groups. There are two sources of correlation that should be considered when fitting statistical models for such data: the correlation between subjects within each cluster and the correlation between repeated observations of the same subject.  In this project, three methods including linear mixed model (LMM), generalized estimating equations (GEEs) and cluster-weighted generalized estimating equations (CWGEEs) will be introduced and applied to a simulated data and to a real data set. It is concluded that the three methods provide comparable estimates for both data sets. It is also noted that for all three methods, the estimates for standard error are biased when the sample size is small.