Covariances for neutron cross sections calculated using a regional model based on local-model fits to experimental data
Abstract
We suggest a procedure for estimating uncertainties in neutron cross sections calculated with a nuclear model descriptive of a specific mass region. It applies standard error propagation techniques, using a model-parameter covariance matrix. Generally, available codes do not generate covariance information in conjunction with their fitting algorithms. Therefore, we resort to estimating a relative covariance matrix a posteriori from a statistical examination of the scatter of elemental parameter values about the regional representation. We numerically demonstrate our method by considering an optical-statistical model analysis of a body of total and elastic scattering data for the light fission-fragment mass region. In this example, strong uncertainty correlations emerge and they conspire to reduce estimated errors to some 50% of those obtained from a naive uncorrelated summation in quadrature. 37 references.
Related Papers
- → Finding maximum likelihood estimates of patterned covariance matrices by the EM algorithm(1982)56 cited
- → Covariance Matrix Estimation From Linearly-Correlated Gaussian Samples(2019)6 cited
- → Correlation matrix estimation and order selection for spectrum estimation(2003)2 cited
- → Improved covariance matrix for target detection in hyperspectral imaging(2016)1 cited
- Covariance Matrix Estimation from Correlated Samples.(2018)