As suggested by Brissenden, it is possible to analyze the reaction rate in the unresolved resonance region by generating sets of random resonance parameters that have the correct statistical properties. Since each set of parameters is itself a random variable, an estimate of the probable error in an average-group cross section or reaction rate can be made by averaging over many random sets. This we have done for a mixture representative of fast breeder reactors and for the energy range 700 to 900 eV. This region is a typical one for studying the Doppler effect. If we make the assumption (a great oversimplification) that the response in this small energy band is typical, not only for the mean but also for the variance, then we would conclude that, if all fine groups (of width 200 eV) have the same weight, the probable error in the fissile component of the Doppler coefficient is about equal to its mean value. For the fine group itself, the probable error in the difference in the relative changes of the fission and the absorption rates is about ten times the mean value.