Obtaining a Coherent Solution II: Bayesian Nets
Riccardo Rebonato · 2012
This chapter discusses how to obtain a coherent solution for Bayesian nets based on certain assumptions such as the structure of the Bayesian net and the associated conditional probability tables. Once the risk manager has at her disposal the required probability tables, the solution of the problems becomes mechanical and is directly obtained via the Bayesian-nets technology, using the properties of Bayesian nets and of conditional and joint probabilities. One possible way to ensure that risk manager is able to guess correctly is to restrict the set of possible Bayesian nets underlying a given problem. This ultimately means simplifying the underlying causal structure among the variables at hand, and should not therefore be undertaken lightly. When this ‘structural pruning’ can be done in a sensible way, however, dramatic simplifications can be derived and the full joint probability structure can be determined with relative ease.