Comparing results between laboratories is not about seeing who got the lowest or highest number. What is evaluated is whether the differences between results can be explained by the uncertainties each laboratory stated. If they can, the results are compatible; if not, there is a bias or an underestimated uncertainty to investigate.
Main schemes
- Proficiency testing: an organizer distributes samples or similar items to several participants and evaluates each result against an assigned value.
- Bilateral comparison: two laboratories measure the same item and compare.
- Round-robin or circulating comparison: an item travels from one laboratory to another and is measured at each.
Requirements for the item and the protocol
The circulating item must remain stable throughout the entire route, with no changes due to transport, humidity or temperature. The protocol sets the method, the measurement points and the conditions. If each laboratory uses a different method, the differences may come from the method and not from the laboratory; this must be taken into account when interpreting the results.
How it is evaluated
A widely used indicator is the normalized error: the difference between the laboratory’s result and the reference value, divided by the root sum of squares of their expanded uncertainties. If the absolute value is less than or equal to one, the result is considered satisfactory; greater than one suggests a discrepancy. In proficiency tests with many participants, scores based on the dispersion of the group are also used.
How to interpret the result
A discrepant result opens an investigation: review the uncertainty calculation, the calibration of the standards, the environmental conditions and the procedure. A satisfactory result in a comparison does not prove that all is well at other points or for other quantities, only for what was measured. This is why comparisons are carried out periodically and at different levels of the range.