Short report on the results for 2004

Full text: Results of nuclear data evaluation carried out with statistical methods are usually presented as a vector of estimated values and corresponding covariance matrix. As a rule for the same object (cross-section, angular distribution) there is a set of evaluations performed within the framework of various physical models. Covariance matrices from different evaluations . quadratic tables of numbers . can be compared on the basis of integral measure of uncertainty. Besides, a check of validity of the estimated data is of prime interest. Exceeding estimated integral uncertainty over experimental one indicates error in statistical processing. Square root of determinant of a covariance matrix was considered as an integral measure of uncertainty for a random vector. It was emphasized that small uncertainties of the components of a random vector can be .compensated. by large correlations between components of other one at keeping approximately the same value for the determinants of the covariance matrices. An algorithm for comparison of experimental and estimated (after model and 'non-model' fits) covariance matrices was proposed. The algorithm is consisted in 1) separation of all the energy range of interest by intervals number of which is less or equal to the number of parameters in the model fit, 2) folding experimental and estimated covariance information in covariance matrices for uncertainties of averaged (over intervals) experimental and estimated cross-sections, 3) comparison of the determinants of the covariance matrices. The algorithm was applied for comparison of the results of 6Li(n,t) reaction cross-section evaluation carried out on the basis of five sets of measurements. As shown, difference between integral uncertainties of cross-sections estimated in model and 'non-model' fits is essentially smaller the one between pointwise (differential) cross-section uncertainties. So, consideration of alone uncertainties of estimated data leads to misunderstanding evaluation results. (author)