Graphical portrayals of the static and dynamic features of belief analysis and revision are basic to the understanding of large stochastic systems. An important graphical tool consists of the Bayes linear influence diagram. These diagrams show a number of nodes connected by arcs. The nodes represent meaningful collections of random quantities. An arc represents a potential for a source node to influence a destination node. Some nodes are or can be shaded to summarise the features of belief revision for the corresponding collection of random quantities. The arcs can be labelled to summarise information flow across the diagram, thus quantifying the degree of influence.
Bayes linear influence diagrams are introduced in [3], and further developed and considered in relation to a case study concerning stock flow and prediction in the brewing industry, in [7]. A more detailed exposition, using related brewing data, is given in [9]. Bayes linear influence diagrams are applied to the problem of revising beliefs about variance-covariance matrices in [14]. A useful starting reference for influence diagrams in general is given by [12].