ABSTRACT
The previous two chapters introduced antedependence models for time-ordered random variables, first in their most general, or unstructured, form, and subsequently in more parsimonious, or structured, forms. Now, we begin to consider statistical inference for such models. In this chapter especially, we describe informal methods for identifying the mean and covariance structures of normal linear antedependence models for longitudinal data. Such methods include the examination of useful summary statistics and graphical diagnostics and are conducted prior to actually fitting any antedependence models. More formal methods of inference, including likelihood-based parameter estimation, hypothesis tests, and model selection criteria, will be considered in later chapters.
