ABSTRACT
The typical GEE or QLS analysis involves data with one level (or source) of correla-
tion within the clusters. For example, in a longitudinal study it is often reasonable to
assume that measurements between subjects are independent, so that the only source
of correlation is due to the similarity between the repeated measurements on a sub-
ject. However, if the longitudinal study simultaneously assessed systolic and diastolic
blood pressure on patients, then we might anticipate two sources of correlation, due
to the fact that measurements within a patient might tend to be more similar if they
represent the same type of blood pressure, or are measured more closely together
in time. Analyses such as this that assess multiple longitudinal outcomes simultane-
ously are also referred to as multivariate longitudinal analyses.