rationalfit vs. invfreqs

Hello,

I have a set of s-parameters from a VNA measurement. For each frequency point a get a 2x2 complex S-Matrix. I want to incorporate the measurement data into a Simulink/SimPowerSystems time domain simulation using a transfer function for each s-parameter. I assume the transfer function should look like H(s)=B(s)/A(s). Therefore I need the coefficients for the numerator and denomiator polynomial. 1. The rationalfit function fits my measurement data very well, and I get the resiudes and poles for each fit. If I calculate the coefficients for my polynomials with [b,a] = residue(r,p,k) I get partly complex coefficients. Due to that I can't use them later in Simulink because complex coeff. are not allowed. 2. The invfreqs function fits my measurement data as well, and I directly get coefficients directly get the b,a coefficients. But this coefficients are not complex.

So my questions is, where is the difference between the rationalfit and invfreqs function. When do I use rationalfit und when invfreqs. What's the right way to fit and get the polynomials coefficients from a set of measured s-parameters data?

It would be of great help to me to get an answer onto that questions.

Looking forward to hearing from you.

Thanks, Matthias