ABSTRACT
This chapter provides a short introduction to some of the most important concepts and definitions of Graphical Modelling. Graphical Modelling can be seen as a type of multivariate analysis that is of particular usefulness in very complex multivariate systems with complicated structures of dependency. Graphical Models are, as well as Structural Equation Models, an extension of the Path Models introduced by Wright, but in contrast to SEMs, they can also be used to model categorical variables. Additionally, Graphical Models get input from other analysis techniques, including Log-linear Models and Covariance Models, as Whittaker points out. Furthermore, the principles of independence and conditional independence are important contributors. Cox and Wermuth propose a selection strategy that bypasses the multivariate problem by fitting the Chain Graph approximately with a series of univariate conditional regressions, based on the factorization of the joint density. Several selection strategies to fit Graphical Models are available, for instance the Edward-Havranek procedure.
