ABSTRACT
The G M A N O V A model of Chapter 5 extends the A N O V A and
M A N O V A models for repeated measurements by including linear regression-
type arguments into the overall model. It is therefore ideally suited for studies
where we wish to model and compare some underlying response curve using,
say, polynomial growth curves. The difficulty wi th the G M A N O V A model is
that it is only designed to handle balanced and complete data. As was
discussed in Chapter 5, there are methods available for handling missing or
incomplete data provided such data are missing at random. However, these
methods do not address the problem of unbalanced data as found in
longitudinal studies where observations are taken at irregularly spaced
intervals. Nor do they address the issue of missing data when the data are not
missing at random. In this chapter, we focus on a class of linear mixed-effects
models which find numerous applications for studies wi th unbalanced repeated
measurements. W e start first wi th a class of random-coefficient growth curve
models which represent a natural extension of the G M A N O V A model.