This report presents two methods for generating conditional probability tables (CPTs) for Bayesian network models.
The methods are based on the notion of departures from a normal state of affairs. The elicitation methods ask experts either for their assessment of the influence of parent nodes on a child node, or for their expectations for the influence of parent nodes on the departure of the child node from a normal state.
Download working paper (pdf, 163kb)
Design and development by Soapbox.