Measurement error models for interlaboratory comparison measurement data

Test laboratories with ISO/IEC 17025:2005 accreditation are obliged to calculate measurement uncertainty and declare the calculated value. Furthermore, they have to ensure the quality of the test results, and their participation in interlaboratory comparisons is mandatory for the accreditation. To this end, a standard procedure is available and the laboratory's performance is also assessed by comparing its results with the reference value. While several studies consider the problem of analysing interlaboratory comparison data, the problem still remains of how to include all the measurements (containing uncertainties and outliers) and all the dispersion effects arising during the test activity, in the analysis. This paper aims to improve the analysis of interlaboratory comparison data by focusing on an error measurement model which considers the declared measured values and the corresponding uncertainties, and by also accounting for other dispersion effects involved in the interlaboratory activity. The problems of the small sample size and the presence of outliers are taken into account through the calculation of confidence intervals, by also evaluating the contribution of the variances estimated for the uncertainties, e.g. by the signal-to-noise and reliability ratios. Moreover, the laboratory's performance is assessed by discriminating for the presence of outliers related to the reference value and/or to the uncertainty. The results are satisfactory in view of the issues addressed in this study, especially if we consider the specific kind of data.