ABSTRACT
This chapter introduces antedependence models in their most general form, which are called unstructured antedependence models because no additional structure beyond that needed for antedependence is imposed upon the model. We begin with definitions of antedependence and partial antecorrelation, which are equivalent in the important case of normal random variables. We then establish a number of equivalent characterizations of the covariance structure of partially antecorrelated variables, as well as some properties of determinants and traces involving such a covariance structure. These results will, in later chapters, be very important for deriving inferential procedures for antedependence models. The results are specialized to the important first-order case, and then generalized to the so-called variable-order case. Finally, relationships between these models and some other, more well known conditional independence models are described.
