Evaluation of measurement uncertainty components, based on judgment, (a Type B evaluation) is considered. The problem amounts to choosing, on the basis of the information available, an appropriate prior probability distribution of the variable at issue within the specified bounds. Three typical situations are highlighted that lead to simple model distributions: normal, triangular, and uniform (rectangular) distributions. It is demonstrated, with examples of uncertainty estimation, taken from literature, including the Guide to the Expression of Uncertainty in Measurement (GUM) and publications on chemical analysis, that the uniform distribution model is often used in those cases where its application is unjustified. This practice is borne out by the advisability of a conservative approach resulting in a "pessimistic" evaluation. It is stressed that the unjustified choice of the uniform distribution model lowers the value of a scientific judgment and the role of subjective probability in the measurement uncertainty evaluation.
measurement uncertainty, Type B evaluation, prior probability distribution, uniform distribution, triangular distribution