Bayesian Nets I: An Introduction
Riccardo Rebonato · 2012
This chapter defines Bayesian nets as directed, acyclical graphs where each node or vertex is associated with a two-valued (Boolean) random variable and with a list of numbers called a conditional probability table. The power of Bayesian nets comes from the fact that we can interpret the existence of an arrow between two nodes as representing a causal link between the associated random variables. Both marginal (prior) probabilities are modified by knowledge of occurrence of the other event. However, one does not think that a rainy day causes the season to become autumn. A causal relationship is linked to a model of how the world works. Availability of such a model helps statistician's understanding, explaining what is essential and what is accidental. It alerts them to when the world has changed and when they should abandon the model.