This report describes the initial development of the Step-Through variant of Monte Carlo simulation, a new procedure for implementing decision analysis or for training decision makers. Like regular Monte Carlo simulation, it involves sampling possible aftermaths of an initial action, and generating a distribution of outcome values for it. However, the detailed structure and/or assessments of the decision model are elicited as called for in the execution of each trial. It, therefore, permits substantial economy of elicitation if there are few trials. The Step-Through procedure also offers economy of elicitation and calculation over a traditional extensive tree decision-analytic model without requiring simplifications or aggregations in the model's conceptualization. In addition to describing this procedure, this paper presents the results of a preliminary test and evaluation of its viability.
