Proficiency tests with uncertainty information: Detection of an unknown random effect. (April 2016)

• Record Type: Journal Article
• Authors: Shirono, Katsuhiro; Shiro, Masanori; Tanaka, Hideyuki; Ehara, Kensei
• Is Part Of: Measurement. Volume 83(2016:Apr.)
• Page Start: 144
• Page End: 152
• Publication Date: 2016-04
• Subjects: Model selection -- Empirical Bayes -- Marginal likelihood -- Interlaboratory comparison -- ISO 13528; Weights and measures -- Periodicals; Measurement -- Periodicals; Measurement; Weights and measures; Periodicals; 530.8
• DOI: 10.1016/j.measurement.2016.01.002
• Languages: English
• ISSNs: 0263-2241

Abstract

Highlights: What causes the inconsistency in a proficiency test (PT) is focused. When an unknown random effect exists, the performance evaluation cannot be implemented. The likelihood of the existence of the random effect is investigated in this study. Abstract: In proficiency tests (PTs), the possible existence of an unknown random effect (such as instability or inhomogeneity of the measured items, or vagueness in the definition of the measurand) that could affect the reported values is a matter of concern, and may influence the performance evaluations in an unfair manner. If an unknown random effect is the dominant source of uncertainty, it is not appropriate to conduct performance evaluations without correcting for that random effect (increasing the uncertainties, correcting the biases or the both). This study presents a statistical method to detect an unknown random effect before the performance evaluation in a PT with uncertainty information. The method is validated through simulations using various types of virtual but possible data sets. Through the application of this method, the applicability of the PT data to the performance evaluation can be checked.