In a dynamic system with unobserved variables it is desirable, but impossible to know, the state of all variables at all times. However, if values for some of the variables are known, you can make a good estimate of the values of other variables. Kalman filtering combines data measurements and model output to make indirect measurements of the model variables.
Kalman Filtering is turned on by checking the box in the Advanced tab of the Simulation Control. To use filtering, both a payoff file and a filter control file (kalman.prm) are required. The payoff file needs to be set up with the measurement noise variances entered, as discussed earlier in this chapter.