Many solutions to the problem of estimating the consensus mean from the results of multiple methods or laboratories have been proposed. In a Bayesian analysis, the consensus mean is specified through probabilistic dependency as either a ¿parent¿ or a ¿child¿ of the method means. In this paper, we propose a unified approach to some of these Bayes solutions by expressing the consensus mean as a measurable function of the method means and some ancillary variable. This measurement Equation Approach is the standard approach used the ISO Guide to the Expression of Uncertainty in Measurement (ISO GUM). When the measurement equation is linear in the ancillary variable, the uncertainty of our Bayes estimator has a decomposition that is ISO GUM compliant.
Proceedings Title: Proceedings of the American Statistical Association
Pub Type: Conferences
ancillary variable, Bayes, ISO Gum, uncertainty