Calculation of Expanded Uncertainty in Measurements Using the Kurtosis Method when Implementing a Bayesian Approach

Abstract

Expanded uncertainty is evaluated according to the revised Guide to the Expression of Uncertainty in Measurement. The methodology of estimation is based on procedures that are independent of the probability density function of the measurand. It is shown that in using this methodology, the relative error of estimation of expanded uncertainty can be greater than 100%. A new method is proposed for computing expanded uncertainty with this shortcoming removed. In this method, the kurtosis of the distribution of input variables is calculated when computing the expanded uncertainty. The proposed method is compared with modeling using the Monte Carlo method, and close results are obtained.