I am running calibrations in Vensim by changing four different parameters. I wonder whether the solutions that Vensim finds are not local minima, but actually the global minima? Also whether I run many different iterations, and choose different starting points, this will prevent from falling into a local minima or not? What other suggestions do you have?
As a not, I do change the weights for each iteration of calibration, and set it as the 1/standard deviation of the error (i.e. difference between simulated and data values) from the previous iteration, so that the maximum likelihood solution is guaranteed.
Thanks
Nasim