skip to main content

Title: Quality Quantification of Evaluated Cross Section Covariances

Presently, several methods are used to estimate the covariance matrix of evaluated nuclear cross sections. Because the resulting covariance matrices can be different according to the method used and according to the assumptions of the method, we propose a general and objective approach to quantify the quality of the covariance estimation for evaluated cross sections. The first step consists in defining an objective criterion. The second step is computation of the criterion. In this paper the Kullback-Leibler distance is proposed for the quality quantification of a covariance matrix estimation and its inverse. It is based on the distance to the true covariance matrix. A method based on the bootstrap is presented for the estimation of this criterion, which can be applied with most methods for covariance matrix estimation and without the knowledge of the true covariance matrix. The full approach is illustrated on the {sup 85}Rb nucleus evaluations and the results are then used for a discussion on scoring and Monte Carlo approaches for covariance matrix estimation of the cross section evaluations.
Authors:
 [1] ;  [2] ;  [1]
  1. ENS Cachan, 61 Avenue du Président Wilson, 94235 Cachan Cedex (France)
  2. CEA/DAM/DIF, Bruyères-le-Châtel, 91297 Arpajon Cedex (France)
Publication Date:
OSTI Identifier:
22436781
Resource Type:
Journal Article
Resource Relation:
Journal Name: Nuclear Data Sheets; Journal Volume: 123; Conference: International workshop on nuclear data covariances, Santa Fe, NM (United States), 28 Apr - 1 May 2014; Other Information: Copyright (c) 2014 Elsevier Science B.V., Amsterdam, The Netherlands, All rights reserved.; Country of input: International Atomic Energy Agency (IAEA)
Country of Publication:
United States
Language:
English
Subject:
73 NUCLEAR PHYSICS AND RADIATION PHYSICS; CROSS SECTIONS; DATA COVARIANCES; EVALUATION; MONTE CARLO METHOD; NUCLEAR DATA COLLECTIONS; RUBIDIUM 85