[en] Although random matrix theory had its initial application to neutron resonances, there is a relative scarcity of suitable nuclear data. The primary reason for this is the sensitivity of the standard measures used to evaluate spectra--the spectra must be essential pure (no state with a different symmetry) and complete (no states missing). Additional measures that are less sensitive to these experimental limitations are of significant value. The standard measure for long range order is the Δ₃ statistic. In the original paper that introduced this statistic, Dyson and Mehta also attempted to evaluate spectra with thermodynamic variables obtained from the circular orthogonal ensemble. We consider the thermodynamic 'internal energy' and evaluate its sensitivity to experimental limitations such as missing and spurious levels. Monte Carlo simulations suggest that the internal energy is less sensitive to mistakes than is Δ₃, and thus the internal energy can serve as a addition to the tool kit for evaluating experimental spectra