ABSTRACT
Suppose that an antedependent covariance structure has been selected for a set of longitudinal data, using either the informal methods of Chapter 4 or the hypothesis testing procedures or penalized likelihood criteria of Chapter 6. To the extent that the order of antedependence of the selected covariance model is low to moderate and is structured rather than unstructured, inferences on the mean structure can be more efficient if this structure is exploited than if the general multivariate dependence structure is adopted. This chapter describes hypothesis tests for the mean structure of normal linear antedependence models which exploit the antedependence.
