Unified Monte Carlo and mixed probability functions

The possibility of using the recently suggested Unified Monte Carlo (UMC) technique for the evaluation of nuclear reaction data other than those that are normally distributed is discussed in this paper. The alternative probability distributions examined here are: 1) lognormal distribution; 2) uniform interval distribution; and 3) exponential distribution. These distributions, together with the normal distribution, are capable of addressing most situations that are likely to be encountered in evaluating nuclear reaction data. The mathematical formalism is described and numerical examples are provided showing instances when the UMC method can lead to results which are quite different from those obtained using the Generalized Least Squares (GLS) method. We conclude that: 1) the GLS method gives reasonable results comparable to UMC for probability functions closely resembling symmetric normal distributions with small to modest uncertainties, but the UMC results are more reliable for broad or strongly asymmetric distributions; 2) the lognormal distribution can be used to good advantage in certain situations involving large uncertainties; 3) use of the uniform and exponential distributions by evaluators should be discouraged.