OVERVIEW

A model of any kind (physical, analytical, simulation, etc.) is a simplified representation of some system under study, and as such, a model would rarely capture precisely system characteristics. Consequently, a model cannot always be expected to predict accurately system performance. Therefore, the cardinal goal of modeling is merely to construct an acceptably "correct" model. The notion of model correctness (often referred to as model goodness) is necessarily relative rather than absolute, and its precise meaning will vary from case to case. Suffice it to say that model goodness should be interpreted as an operational notion that requires the model to produce "sufficiently accurate" performance predictions for the system under study. Again, the notion of sufficient accuracy of model predictions depends on the individual cases under consideration, and the task of defining accuracy criteria is often left to the analyst. In many cases, model accuracy cannot be exactly gauged (e.g., when the system under study does not exist), so that any assessment of model goodness becomes merely an educated guess. Since the determination of model goodness is rarely a crisp or algorithmic procedure, the enterprises of modeling and goodness assessment are necessarily "best effort" activities.

This chapter focuses on goodness assessment, and specifically on its two components: verification and validation. Verification assesses the correctness of the formal representation of the intended model (in our case, a computer simulation program), by inspecting computer code and test runs, and performing consistency checks on their statistics. Validation...