Trajectory sensitivity of non‐linear system dynamics models to parameter variations

This paper describes the use of trajectory sensitivity to illustrate how it can be used to help differentiate between competing System Dynamics models ascertaining the most significant parameters. A comparison of trajectory simulation and eigenvalue sensitivity is made in the case of simple models, examining the sensitivity of non‐linear systems to changes in parameters. These methods can be used in the validation of models, in designing systems that are robust or in determining the principle cause of possible change in a given system. The two methods are applied to a case study of the simple SIR epidemic transmission model to illustrate the approach. The trajectory simulation method shows that system performance is most affected by the contact rate and infectivity around the period of maximum infection. A subsequent analysis of Flu data for 1978 shows that the errors between the SIR model and the data are not accounted for by possible parameter variations.