ABSTRACT
This chapter explores some of the uses of a graphical model after its construction, and looks at the problems of exploring the marginal distributions produced by the fusion and propagation algorithm. It focuses on the statistics produced by the typical analysis of a graphical model, and describes the global conflict, and the modifications that must be made to the fusion and propagation algorithm to produce this statistics. The explorations of the model yield insight into the nature of the model and the phenomenon being modelled; they are typically far more valuable than the single numeric summaries. There are a number of ways to expand graphical models. The chapter also looks at the role of uncertainty and imprecision in models for the parameters of the factors of the graphical belief function. This leads naturally to the idea of second-order models—models over the parameters of a graphical belief function.
