Models, Uncertainty, and the Sandia V&V Challenge Problem

Abstract

The purpose of this paper is not to present new results; rather, it is to show that the current approach to model validation is not consistent with the accepted mathematics of probability theory. Specifically, we argue that the Sandia V&V Challenge Problem is ill-posed in that the answers sought do not, mathematically, exist. We apply our arguments to show the types of mistakes present in the papers presented in the Journal of Verification, Validation and Uncertainty Quantification, Volume 1,1 along with the challenge problem. Further, we argue that, when the problem is properly posed, both the applicable methodology and the solution techniques are easily drawn from the well-developed mathematics of probability and decision theory. The unfortunate aspect of the challenge problem as currently stated is that it leads to incorrect and inappropriate mathematical approaches that should be avoided and corrected in the current literature.