Open AccessCoverage Intervals
Author(s) -
Sara Stoudt,
Adam L. Pintar,
Antonio Possolo
Publication year - 2021
Publication title -
journal of research of the national institute of standards and technology
Language(s) - English
Resource type - Journals
SCImago Journal Rank - 0.202
H-Index - 59
eISSN - 2165-7254
pISSN - 1044-677X
DOI - 10.6028/jres.126.004
Subject(s) - probabilistic logic , statistics , monte carlo method , confidence interval , bayesian probability , tolerance interval , computer science , interpretation (philosophy) , interval (graph theory) , econometrics , mathematics , combinatorics , programming language
Since coverage intervals are widely used expressions of measurement uncertainty, this contribution reviews coverage intervals asdefned in the Guide to the Expression of Uncertainty in Measurement (GUM), and compares them against the principal types ofprobabilistic intervals that are commonly used in applied statistics and in measurement science.Although formally identical to conventional confdence intervals for means, the GUM interprets coverage intervals more as if they wereBayesian credible intervals, or tolerance intervals.We focus, in particular, on a common misunderstanding about the intervals derived from the results of the Monte Carlo method of theGUM Supplement 1 (GUM-S1), and offer a novel interpretation for these intervals that we believe will foster realistic expectationsabout what they can deliver, and how and when they can be useful in practice