A systematic description of the generation of covariance matrices

In this paper, we describe techniques used to determine realistic and appropriate uncertainties and correlations (or, equivalently, covariances) for multigroup cross sections in the resolved-resonance region, starting from fundamental principles. The entire process is described, with emphasis on the propagation of uncertainties through each step of the process. The key steps are data reduction (conversion from measured counts-per-time-channel to experimental cross section as a function of energy), data evaluation (determining appropriate parameter values to fit theory to experiment), generation of point-wise cross sections (reconstruction of theoretical cross sections as a function of energy), collapsing into multigroup cross sections (averaging the point-wise cross sections to give multigroup cross sections), and calculating integral quantities from those multigroup cross sections.