A random sampling procedure is used to construct resonances in the unresolved region. The success of this procedure depends on the ability to determine statistically meaningful reactivity coefficients. To establish an estimate of the statistical dispersion of the Doppler effect for a carbide-fueled fast power reactor, many different resonance ladders were studied for each total angular momentum state of the compound nucleus for each isotope. It is shown that the one-standard-deviation statistical uncertainty in the calculated total Doppler effect for the core is ∼3%, which is quite satisfactory. However, the statistical uncertainty in the 239Pu Doppler effect was determined to be ∼40% in the unresolved region, and ∼35% over all energy. The manner in which the ladders are chosen is investigated, with the conclusion that any ladder, giving the proper distributions of resonance parameters, may be used. Based on these results, it may be advantageous to use the random sampling method of resonance construction in place of the usual statistical averaging procedure. This would eliminate some of the approximations inherent to these statistical averaging procedures, by including all interference and overlap effects.