A theory of structural model validity in simulation

During the last decennia, the practice of simulation has become increasingly popular among many system analysts, model builders and general scientists for the purpose of studying complex systems that surpass the operability of analytical solution techniques. As a consequence of the pragmatic orientation of simulation, a vital stage for a successful application is the issue of validating a constructed simulation model. Employing the model as an effective instrument for assessing the benefit of structural changes or for predicting future observations makes validation an essential part of any productive simulation study. The diversity of the employment field of simulation however brings about that there exists an irrefutable level of ambiguity concerning the principal subject of this validation process. Further, the literature has come up with a plethora of ad hoc validation techniques that have mostly been inherited from standard statistical analysis. It lies within the aim of this paper to reflect on the issue of validation in simulation and to present the reader with a topological parallelism of the classical philosophical polarity of objectivism versus relativism. First, we will position validation in relation to verification and accreditation and elaborate on the prime actors in validation, i.e. a conceptual model, a formal model and behaviour. Next, we will formally derive a topological interpretation of structural validation for both objectivists and relativists. As will be seen, recent advances in the domain of fuzzy topology allow for a valuable metaphor of a relativistic attitude towards modelling and structural validation. Finally, we will discuss several general types of modelling errors that may occur and examine their repercussion on the natural topological spaces of objectivists and relativists. We end this paper with a formal, topological oriented definition of structural model validity for both objectivists and relativists. The paper is concluded with summarising the most important findings and giving a direction for future research.status: publishe