The context of this work lies in limitations associated with a key component of the dynamic modeling process - calibration. Classically, calibration of a simulation model seeks to estimate little-known model parameters by comparing the emergent behavior exhibit of that model with corresponding behavior observed from the external world. For example, we may be seeking to use calibration to estimate the values of parameters concerning contact rates on which we have limited data directly from studies or from surveys that had been conducted. Within this calibration, we will observe how the simulation model as a whole - or large subpieces thereof - behaves in terms of its emergent behavior. With simulation models, the complexity of such behavior is generally such that we can't simply "back-calculate" the values for parameters such that model output will match the empirical data. Instead, we compare the emergent behavior of the model against empirical data on corresponding quantities to which we have recourse to, and try to use an optimization algorithm to estimate the values of model parameters that yield model behavior most closely corresponding to that empirical data.
展开▼