Abstract
I have analyzed reduced neutron widths ($\Gamma
_{n}^{0}$) for the subset of 1245 resonances in the nuclear data ensemble (NDE)
for which they have been reported. Random matrix theory (RMT) predicts
for the Gaussian orthogonal ensemble (GOE) that these
widths should follow a $\chi ^{2}$ distribution having one degree of freedom
($\nu =1$) - the Porter Thomas (PT) distribution. Using the
maximum-likelihood (ML) technique, I have determined that the $\Gamma
_{n}^{0}$ values in the NDE are best described by a $\chi ^{2}$ distribution
having $\nu =0.801\pm 0.052$, which is 3.8 standard deviations smaller than
predicted by RMT. I show that this striking disagreement is most likely due
to the inclusion of significant \textit{p}-wave contamination to the
supposedly pure \textit{s}-wave NDE. Furthermore, when an energy-dependent
threshold is used to remove the \textit{p}-wave contamination, ML analysis
yields $\nu =1.217\pm 0.092$ for the remaining data, still in poor agreement
with the RMT prediction for the GOE. These results cast very serious doubt
on claims that the NDE represents a striking confirmation of RMT.