Assessment of uncertainty in analytical results

For a measurement result to be useful, it is essential that its value be accompanied by its associated uncertainty. Measurement uncertainty is, in fact, a quantification of the degree of doubt (or confidence) we have in the measurement result. It is therefore a decision-making criterion when comparing results—either with one another or against a specification.

Estimating measurement uncertainty is therefore mandatory for the analyst. To do so, he or she has access to various tools and approaches, either within the laboratory or through the laboratory’s participation in collaborative studies (interlaboratory comparisons).

The Guide to the Expression of Measurement Uncertainty (GUM) has evolved significantly and continues to do so: it no longer corresponds solely to the uncertainty propagation approach (JCGM 100), but also covers other approaches (the Monte Carlo simulation approach (JCGM 101), the Bayesian approach, and the interlaboratory approach). The GUM therefore now consists of a set of reference documents.

In this article, the authors aim to provide an overview of the various approaches to assessing measurement uncertainty in chemistry and to demonstrate that the approaches presented in the GUM can also utilize data from measurement method validation. The key concepts underlying the process of estimating measurement uncertainty are therefore reviewed, along with the requirements of the NF EN ISO/IEC 17025 standard. Following a brief overview of the GUM, particularly its recent developments, the various approaches to estimating measurement uncertainty—using either intralaboratory data or out-of-laboratory data (obtained through a collaborative study)—are reviewed. These approaches are illustrated through an example that serves as a common thread throughout this document.