Thread Subject: How to reconstruct a model with the RLS identified parameters?

Hi, I have an RLS algorithm for identifying model parameters for an ARX model like below:

y = -ay(k-1) -by(k-2) + cu(k-1) + du(k-2)

The identified parameters will be a,b,c and d. My RLS algorithm is written in s-function. I want the a,b,c,d identified from the RLS algorithm to be used to construct a new transfer function having the following form:

G(z) = [cz^-1 + dz^-2]/[1 + az^-1 + bz^-2]

At every time instant, I want the a,b,c,d that is being identified from the RLS algorithm to be fed to the transfer function, so that I can simulate the new model and compare the results with the original process. Which means I need a way to channel the a,b,c,d to the transfer function. I cannot decompose the transfer function and use multipliers to drive a constant set because I cannot decompose the denominator of the transfer function. Does anyone know how can I solve my problem?