All measured values are wrong, but some are useful.
One score and 13 years ago the Bureau International des Poids et Mesures (BIPM), in conjunction with the IEC, IFCC, ISO, IUPAC, OIML, published the "Guide to the Expression of Uncertainty in Measurement" (GUM). This document defined a reportable measurement as Y ± ku, where Y is a plausible estimate of the unknowable true value of the quantity intended to be measured (the measurand), u is a quantitative estimate of the uncertainty of the measurement process, and k specifies the level of plausibility associated with the estimated uncertainty, where k = 1 (plausible), k = 2 (highly plausible), k = 3 (very highly plausible). This concept has not been embraced by the general public, journalists, and many technical personnel involved in the production and regulation of pharmaceutical products, all of whom still blissfully assume that the reportable value for a measurement is some single numerical "best estimate" and who frequently misunderstand or are ignorant of the fact that such a "best estimate" is at best a fuzzy number that could plausibly be somewhat different because this "best estimate" is the product of a fallible measurement process that always produces results with some associated uncertainty-and that they are duty-bound to also simultaneously report their "best estimate" of this uncertainty! Over ten years ago I and coauthors presented a history of the development of the concept of measurement uncertainty for an audience of statisticians ["A Historical Perspective on Analytical Measurement Uncertainty: From Cotes, Laplace and Gauss to GUM and Implications for Current Practice." Presented at the Midwest Biopharmaceutical Statistics Workshop, Ball State University, Muncie, IN, May 18–20, 2015] and later I wrote a "sermon" on this topic [William R. Porter, "Are You Uncertain About Uncertainty?" J GXP Compliance 21(3) (May 30, 2017)], both attached below. My question to the community is how successful have you been in educating others about the concept of measurement uncertainty and the need to report an estimate of uncertainty in conjunction with each reported "best estimate" of the "measurand"? What works?
------------------------------
William Porter Ph.D.
Retired
Peak Process Performance Partners LLC
Vernon Hills IL
[email protected]Disclaimer: Opinions expressed are solely my own and do not express the views or opinions of my employer.
------------------------------