COVARIANCE STRUCTURE MODELS AND INFLUENCE DIAGRAMS

Statisticians use covariance structure modeling as a versatile tool fo r modeling and testing theory. The models that result provide explicit and detailed descriptions of stochastic systems. We show how covarian ce structure models are related-mathematically, conceptually, philosop hically and practically-to Gaussian influence diagrams as described by Shachter and Kenley (1989). This relationship suggests ways in which covariance structure modeling can be used to advantage in the prescrip tive domain of decision analysis. The paper includes an example concer ning the management of hazardous materials, in which a covariance stru cture model is converted to an influence diagram for use in a prescrip tive analysis.