In this paper, a method is presented to quantify the uncertainty on an estimated transfer function model due to disturbing process noise and model errors. The model errors are modeled with a stochastic model from which parameters are estimated from the residuals. The identification is performed in the frequency domain.