NUCLEAR-DATA EVALUATION BASED ON DIRECT AND INDIRECT MEASUREMENTS WITH GENERAL CORRELATIONS
Los Alamos National Laboratory
Los Alamos, New Mexico 87545, U.S.A.
Optimum procedures for the statistical improvement, or updating, of an existing nuclear-data evaluation are reviewed and redeveloped from first principles, consistently employing a minimum-variance viewpoint. A set of equations is derived which provides improved values of the data and their covariances, taking into account information from supplementary measurements and allowing for general correlations among all measurements. The minimum-variance solutions thus obtained, which we call the method of "partitioned least squares," are found to be equivalent to a method suggested by Yu.V. Linnik and applied by a number of authors to the analysis of fission-reactor integral experiments; however, up to now, the partitioned-least-squares formulae have not found widespread use in the field of basic data evaluation. This approach is shown to give the same results as the more commonly applied Normal equations, but with reduced matrix inversion requirements. Examples are provided to indicate potential areas of application.
KEYWORDS: data evaluation, updating, adjustment, minimum variance, Gauss-Markov theorem, normal equations, least squares, correlation, uncertainty analysis