Collaborative studies are widely used for evaluating the performance of analytical methods. The ISO standards suggest first checking for the removal of any atypical observations (outliers) before estimating the repeatability and reproducibility of methods. However, a usual rejection of such data might reduce the reliability of evaluations, especially for smaller samples. Robust statistics provide an alternative procedure capable of estimating the method performance without exclusion of data. However, this alternative procedure does not provide information about the alignment of lab results with the estimated global mean, repeatability and reproducibility as requested for their accreditation. To obtain this information, two a posteriori statistical tests were implemented: one for checking the distance between labs’ means and estimated global mean, and the other for comparing labs’ variabilities to the robust intra-laboratory variability.

In this presentation, the application of ISO robust statistical methods to assess data precision (repeatability and reproducibility) and these two a posteriori tests will be explained, with a worked example using a cigar collaborative study. This example shows that the robust method allows reliable estimation of the method precision when less than eight laboratories are involved in a collaborative study.