Improving model understanding using statistical screening

Abstract

Models of dynamic systems are often constructed to improve system performance by identifying and modifying structures and parameters that drive system behavior. Once identified, these can be used to design and test policies for performance improvement. A preliminary step in developing policies is the identification of high leverage parameters and structures, the influential model sections that drive system behavior. The current work describes the use of statistical screening as a tool to improve model understanding, explanation, and development with a six step process. Statistical screening offers system modelers a user-friendly tool that can be used to help explain how model structure drives behavior.