Validation of Simulation Models via Simultaneous Confidence Intervals

Abstract

The purpose of this paper is to present state-of-the-art and new research results on the use of simultaneous confidence intervals (and joint confidence regions) for determining the operational validity of a multivariate response simulation model of an observable system. A methodology is presented which allows the use of different types of statistical procedures and provides for a tradeoff analysis among sample sizes, confidence levels, sizes of confidence intervals (or regions), and, if desired, cost of data collection. The methodology is illustrated for the validation of mean behavior.