Arthur, William B. with Robert L. Eberlein, "Sensitivity Simulation", 1996
ua435
Systems dynamics modelers normally develop models and implement findings from a deterministic perspective. This approach has great merit. It focuses attention on system structure and behavior as well as ways to change them. Once developed, however, a good system dynamics model is an excellent tool for analyzing systems behavior under a wide variety of parametric assumptions. Though such sensitivity analysis can (and should) be done manually through repeated simulation, automated tools allow more complete exploration. Moreover, they can provide information on the distribution of outcomes that strongly effects decision-making. In this paper we present the basic ideas behind doing multivariate sensitivity simulations (MVSS) and describe how these have been implemented in Vensim. Then, we present a case study that uses MVSS in the pharmaceutical industry.