You can use optimization to look for model errors, adjust model parameters based on data, test parametric sensitivity, and choose the best policy levers. Optimization can be used with or without the filtering capability (itself a form of optimization), which is described later in this chapter. To optimize is, quite simply, to achieve the best. Optimization requires that you define a payoff function that summarizes how good a simulation with a single number (as described above).
Your definition of the payoff function will determine whether optimization is used to calibrate a model to data or choose a best policy. In addition to defining the payoff function you will need to choose the Constants in the model that you wish to optimize over. Once you have chosen these Constants, the optimizer will try to find the values for those parameters that make the payoff as big as possible. If you have a calibration payoff, that means making the model fit the data as closely as possible. If you have a policy payoff, that means maximizing the weighted sum of performance measures.
At the end of an optimization, the results are written to a file called runname.out (where runname is the name you have specified for the simulation). Depending on what options you have chosen, other files might also be created. The file runname.out contains the values of the Constants you specified that optimize the payoff you specified.
NOTE Optimization can be interrupted by clicking on the Stop Button or pressing the Esc key and will shut down cleanly for later resumption.
