About the treatment of systematic effects in metrology

Abstract

A comparison of the text of VIM recent III Edition with that of the GUM and of its contemporary VIM II Edition alights significant differences in the definition of basic measurement terms in the two documents, and with respect to the basic written standards in the field of testing, ISO 5725 and ISO 3534. The paper intends to introduce author’s interpretation of these – and companion – texts, concerning specifically the terminology and the statistical treatment of the influence quantities and of the effects of their variability (in time and standard-to-standard), either related to replicated measurements performed on a single standard (standard ‘reproducibility’) or to the comparisons of different standards, thus involving the concept of ‘accuracy’ and its estimate, and consequently directly relevant to traceability. Another question that arose a few years ago was whether different types of measurand could be the consequence of the different intrinsic nature of different types of standards. It prompted an analysis that resulted in the proposal of considering two distinct ‘classes’ of standards. These classes require different answers to the issue of the treatment of systematic effects. The distinction is relevant, in particular, to the statistical treatment of comparison data, which form the basis of the traceability assessment. The paper is presenting a discussion on the implications of the above distinction, concentrating on cases where systematic effects are dominating the experimental results, a common case in several metrology fields, and on ways to tackle the problem of the correction required by the GUM for standards of class 2 (standards whose values are accurate measures of a common measurand) – a class often not recognised in the general literature.