Covalidation Of Dissimilarly Structured Models

A methodology is presented which allows comparison between models under different modeling paradigms. Consider the following situation: Two models have been constructed to study different aspects of the same system. One model simulates a fleet of aircraft moving a given combination of cargo and passengers from an onload point to an offload point. A second model is an optimization (linear programming) model, which for given cargo and passenger requirements, optimizes aircraft and route selection in order to minimize late- and non-deliveries. The optimization model represents a much more aggregated view of the airlift system than does the simulation. The two models do not have immediately comparable input or output structures, which complicates a comparison of the two models’ outputs. A methodology is developed to structure this comparison. Two models which compare favorably using this methodology are considered covalid models. We define the covalidity of models in the narrow sense as models which perform similarly under approximately the same input conditions. Structurally different models which are covalid (in the narrow sense) may hold the potential to be used in an iterative fashion to improve the input (and thus, the output) of one another. Ultimately it is hoped that we may, through such a series of innovations, effect a convergence to valid representations of the real-world situation. We define such a condition as covalidation in the wide sense. Further, if one of the models has been independently validated (in the traditional meaning), then we may effect a validation by proxy of the other model through this process.