Is your ‘homogeneity test’ really useful?

Abstract

‘Homogeneity testing’ is a formal requirement in the preparation of reference materials for certification and for proficiency testing. Few scientists, however, seem to be aware of the rather severe limitations that apply to the outcome of any such test of a size that is economically feasible. Typically the tests have low statistical power to detect significance, and the resulting estimates of between-bottle standard deviation have wide confidence limits. Scientists should bear these limitations in mind and avoid being over-prescriptive when drafting standards and guides.