Abstract for: Investigating an Automated Method for the Sensitivity Analysis of Functions
Automated sensitivity analysis approaches in system dynamics focus primarily on model parameters. Although table functions are often subjectively approximated, they do not form the focus of most sensitivity analyses. Recently, a promising approach that allows automation of sensitivity analysis on functions was proposed by Hearne (2010), but the applicability of this method to system dynamics table functions has not been studied, yet. In this study, the new method is applied to a simple system dynamics model. In the light of the observations a number of shortcomings are identified and a set of extensions to address these are proposed and then tested. The results of experiments with the original and the extended method demonstrate that the method can be used easily and efficiently for table functions. The extensions are shown to be valuable in creating a more comprehensive method, but they also raise the research issue of the trade-off between their added value and the cost of dealing with increased complication. Apart from our experimental results, the article also puts forth a set of directions along which the approach can be improved further. Despite the issues requiring further research, the method holds promise for routine implementation.