An approach to resolving the parameter sensitivity problem in system dynamics methodology (Q581313)

The parameters of most system dynamics models are subject to a great deal of uncertainty. This means that there are many more equally valid solutions to a model than those which are presented. It is therefore pertinent to ask: Are there any solutions which contradict or invalidate the model conclusions? A number of approaches are available to answer this question for dynamic optimization problems. Many system dynamics models, however, do not have an objective function and in these cases, only a partial attempt is usually made to answer the equation. The common approach is to change one parameter at a time to see if this yields any solutions which differ significantly from the standard solution. Little or no attempt is made to consider the effects of a simultaneous combination of changes in the parameter values because, it is argued, the number of combinations is far too large. In this paper an approach to resolving this problem is suggested. It is shown that by constructing an appropriate objective function relating to the aim of the sensitivity analysis and employing standard numerical optimization techniques the most pertinent parameter combinations can be identified.