ABSTRACT
In this chapter, the authors need to include variables that are numeric. Such variables could be discrete or continuous; they generally require an infinite number of states. The authors provide a useful introduction into some of the specialist terminology used by Bayesian statisticians, who mainly use numerical variables in their inference models. The standard inference algorithms for Bayesian network (BN) propagation only work in the case where every node of the BN has a finite set of discrete states. The authors present three examples to show the power and usefulness of dynamic discretization. All of the problems of static discretization of numeric variables can be avoided that an efficient algorithm using dynamic discretization has been implemented in AgenaRisk. The authors describe the important problem of risk aggregation, which is easily solved using the AgenaRisk compound sum analysis tool and provides an elegant method for calculating loss distributions by taking account of both frequency and severity.
