Correlation of test results and influence of a mass balance constraint on risks in conformity assessment of a substance or material

• Published in: Measurement : journal of the International Measurement Confederation Vol. 163; p. 107947
• Main Authors: Pennecchi, Francesca R., Di Rocco, Aglaia, Kuselman, Ilya, Hibbert, D. Brynn, Sega, Michela
• Format: Journal Article
• Language: English
• Published: London Elsevier Ltd 15.10.2020; Elsevier Science Ltd
• Subjects: Bayesian analysis; Bayesian multivariate modelling; Conformity assessment; Correlation; Evaluation; Mass balance; Mass balance constraint; Monte Carlo method; Monte Carlo simulation; Programming environments; Risk of false decisions; Uncertainty; Monte Carlo method; Conformity assessment; Bayesian multivariate modelling; Risk of false decisions; Mass balance constraint
• ISSN: 0263-2241; 1873-412X
• DOI: 10.1016/j.measurement.2020.107947

[Display omitted] •Three kinds of correlation of component contents involved in a mass balance are discussed.•Three Bayesian multivariate models describing compositional data are proposed.•A Monte Carlo method for conformity assessment at a mass balance constraint is developed. When components of a substance or material are subject to a mass balance constraint, test results of the components’ contents are intrinsically correlated because of the constraint. This so-called ‘spurious’ correlation is observed in addition to possible metrologically-related correlation of test results, and natural and/or technological correlation of the components’ contents. Such correlations may influence understanding of test results and evaluation of risks of false decisions, due to measurement uncertainty, in conformity assessment of the substance or material. The objective of the present paper is the development of a technique for appropriate evaluation of the risks. A Bayesian multivariate approach to evaluate the conformance probability of materials or objects and relevant risks is discussed for different scenarios of the data modelling, taking into account all observed correlations. A Monte Carlo method, including the mass balance constraint, written in the R programming environment, is provided for the necessary calculations.