Statistical validation of simulation models: A case study

Rigorous statistical validation requires that the responses of the model and the real system have the same expected values. However, the modeled and actual responses are not comparable if they are obtained under different scenarios (environmental conditions). Moreover, data on the real system may be unavailable; sensitivity analysis can then be applied to find out whether the model inputs have effects on the model outputs that agree with the experts' intuition. Not only the total model, but also its modules may be submitted to such sensitivity analyses. This article illustrates these issues through a case study, namely a simulation model for the use of sonar to search for mines on the sea bottom. The methodology, however, applies to models in general.