zpkfit

zpkfit uses the nonlinear least-squares fitter (lsqnonlin) to fit an analytic model of any (fixed) number of poles and zeroes to numerical (presumably measured) frequency response data. It takes as arguments a frequency vector, a frequency response vector, and vectors of the initial guesses of pole and zero locations. The number of poles and zeroes given in the initial guesses determines the order of the model that will be fit.

This works extremely well for models that can be well fit by a few poles and zeroes. For more complicated cases, vectfit.m or one of the functions in the System ID toolbox might be more appropriate.