The evaluation of neutron cross sections in the low energy range (electron volt, mega-electron-volt) is based on formal nuclear models having different types of parameters. Some of them may be fitted to reproduce experimental datasets giving rise to an adjusted covariance matrix. In this paper, a Monte Carlo method is presented to properly consider the influence of the remaining parameters, having a priori uncertainties, on the fitted parameters covariances. This method is based on an exact mathematical description using conditional probabilities. To explain the key points of the methodology, an academic example of average parameters evaluation in the unresolved resonance range is presented using Hauser-Feshbach model calculations.