On the validation by inter-laboratory study of ‘procedures’ in chemical measurement

Abstract

This study examines formally the proposition that validation of analytical procedures can be best undertaken by interlaboratory study. It approaches the task by considering the fundamental nature of validation and the meaning of ‘procedure’ in relation to uncertainty. The contention is that (a) only procedures as such can be validated, (b) the results of a procedure are empirical, (c) empirical procedures are unbiased, and (d) the interlaboratory reproducibility standard deviation (SD_R) associated with an unbiased procedure is an unbiased estimate of the uncertainty. The definitive interlaboratory study for estimating SD_R is the collaborative trial, but the case is made that, with due caution, data from proficiency tests can also provide useful estimates of SD_R and thence estimates of uncertainty.