To date, there is no widely-accepted procedure for comparison (statistical treatment) of qualitative and related property values of a substance or material obtained in different laboratories, which can lead to misunderstanding and illogical interpretation of these values. The objective of the present project is harmonization of the approaches to interlaboratory comparisons of qualitative and related property values of a substance or material for further applications of the interlaboratory comparisons for proficiency testing the laboratories dealing with qualitative analysis and other metrological purposes.
Qualitative property values of a substance or material are nominal values. A nominal property value is a word or alphanumerical code given for identification reasons, where the property (variable) has no magnitude (JCGM 200:2012 – VIM clause 1.30). Nominal variables are coded by exhaustive and disjoint classes or categories with no natural ordering. For example, imperfections of welds are coded by the following five classes/categories: cracks, cavities, inclusions, lack of fusion/penetration, and geometrical shape errors.
If a quantity has a clear ordering, then it is an ‘ordinal quantity’ (VIM clause 1.26). For example, the ordinal Mohs scale of mineral hardness is based on the ability of one mineral natural sample to scratch another: corundum (Al2O3) is harder than topaz (Al2SiO4(OH−,F−)2), but diamond (C) is harder than corundum, etc. Such ordinal property values are categorical, as nominal data, but are not entirely qualitative. They are not quantitative also and could be named as semi-quantitative values. Moreover, some products are characterized by combined quantitative and qualitative property values. For example, the property values of dry sausages include mass fractions of fat, protein, moisture and salt (quantitative values), as well as color of the sausage cross section, smell and taste (qualitative values). A comprehensive (universal) scale of property values is necessary for comparison of complete test results of the sausage in different laboratories.
The problem of analysis of qualitative data is recognized by international groups, such as ISO TC 334 (former ISO/REMCO), ISO TC69 SC6 and Eurachem/CITAC WG, working on development of guidelines for different applications of metrological concepts to qualitative data. There are also publications on intensive development of statistical methods CATANOVA for nominal values and ORDANOVA for ordinal values, e.g., papers of Tamar Gadrich et al. https://doi.org/10.1007/s42452-020-03907-4 and https://doi.org/10.1016/j.jspi.2021.04.005, respectively. These methods are analogous to ANOVA (analysis of variance) for quantitative values.
IUPAC has contributed terminology efforts in this field, starting from Rene Dybkaer’s “Ontology on Property” (https://ontology.iupac.org/ or https://doi.org/10.1351/978-87-990010-1-9), the IUPAC Silver Book (https://iupac.org/what-we-do/books/silverbook/) and Nominal Properties 2017 Recommendations (https://doi.org/10.1515/pac-2011-0613), and as a member organization of JCGM developing the forthcoming International Vocabulary of Metrology (VIM4).
Harmonization of approaches to interlaboratory comparisons of nominal and ordinal values, alone and together with quantitative values, will be helpful for characterization of examination methods and reference materials with qualitative and related properties, for proficiency testing of chemical laboratories involved in qualitative analysis, and will contribute to world-wide conformity assessment.
Page last updated 31 Aug 2021