Model Analysis and Calibration

This chapter introduces methods that support policy analysis for system dynamics models. First, a mathematical method for calculating loop polarity is presented, and this formal approach can be used to detect shifts in loop dominance, for example, when two feedback loops compete to influence a stock’s value. Second, statistical screening is summarized, and this allows for an exploratory analysis of a system dynamics model in terms of analyzing which of the many uncertain parameters stand out as most influential. Third, model calibration is explored, which is a valuable technique based on optimization methods. This approach can be used to fit model parameters to historical data. In turn, this can improve client confidence, and also provide good parameter estimates that can form the basis of policy design and analysis.