The analysis of shape data including normalization and the impact on prompt fission neutron spectrum measurements
While measurements of physical quantities contain an intrinsic absolute scale based on details of the experiment, sometimes only the value of each measured data point relative to the others, i.e., the shape, is of interest. Additionally, in contrast to absolute measurements of quantities like nuclear cross sections, shape measurements are unique in that the result is contained entirely in the relative value of each measured point with respect to all others. This fundamental attribute of shape data is sometimes discussed by data evaluators, but it is common that shape data are reported by experimenters with covariances directly from the raw data as though an absolute measurement has been performed, thereby misrepresenting the true knowledge of the measured shape. A normalization procedure must be carried out including careful covariance propagation in order to accurately represent the uncertainty in a measured shape. Known consequences of this procedure, such as the exclusion of constant fully-correlated uncertainties, are sometimes applied ad hoc to shape data. However, normalization also produces features of shape data that are otherwise difficult to apply, such as partial exclusion of strongly-correlated uncertainty sources, a redistribution of uncertainties across data points, and consistency of the covariance matrix obtained when combining separate shape measurements. In this study all of the above features are shown in a generalized fashion and exhibited using prompt fission neutron spectrum data from the Chi-Nu experiment with recommendations for analysis of future shape measurements.