A methodology for cost-risk analysis in the statistical validation of simulation models

A methodology is presented for constructing the relationships among model user's risk, model builder's risk, acceptable validity range, sample sizes, and cost of data collection when statistical hypothesis testing is used for validating a simulation model of a real, observable system. The use of the methodology is illustrated for the use of Hotelling's two-sample T 2 test in testing the validity of a multivariate stationary response stimulation model.