Comparison of Covariance Matrices Obtained by Different Methods

At present significant efforts in nuclear data evaluation are focused on the generation of evaluated data files which include also uncertainty information in form of covariance matrices. Several methods for the generation of covariance matrices have been worked out, ranging from Monte Carlo variation of model parameters to detailed Bayesian techniques. In this contribution cross section covariance matrices for neutron-induced reactions on Pb-208 obtained by Monte Carlo variations of model parameters are compared with the corresponding matrices obtained by strict Bayesian techniques as implemented in the code package CENEUS. Both evaluations rely on the same nuclear model, i.e., TALYS-1.0, with the same parameters and account for a similar set of experimental data. Thus the comparison directly reflects the features of the different methods and allows to draw conclusions concerning the reliability of the extracted uncertainty information.