A propensity-based framework for measurement uncertainty analysis

Abstract

We regard the measurand (or influence quantity) as a random variable characterized by central tendency and dispersion. We define ‘propensity’ as the sum of the central tendency and dispersion. We argue that the state of propensity can be described by a probability density function (PDF) that we name ‘propensity spectrum’. This paper proposes a propensity-based framework for measurement uncertainty analysis of the measurand that is related to influence quantities through a measurement model. The proposed framework encodes the state of propensity of the measurand in the form of a PDF (propensity spectrum) based on all available information about influence quantities. This can be achieved by an “exact” method based on the principle of propagation of distributions, or a first-order approximation method (the law of propagation of uncertainty). A case study is presented to demonstrate the methods provided by the proposed framework, compared to the GUM method and Bayesian methods.