Assessing Resemblance, Repeatability, and Reproducibility for Quantitative Methods

Of the eight desirable attributes of a standardized method given in KSA-SM-3, three are the focus of this article: resemblance of the untreated control data, repeatability of the response of interest across multiple tests, and reproducibility of the response across multiple laboratories. These attributes can be assessed by considering three random effects on the response: among-carrier differences within each test; among-test differences in each lab (e.g. tests are performed on different days with different reactors and inocula); and among-lab differences (e.g. labs are in different geographic locations and use different equipment). The magnitude of each of these effects on the response is quantified by a variance component, or a standard deviation (SD), which is the square root of a variance component. By estimating the SD associated with each random effect, a multiple-laboratory study can provide an assessment of resemblance, repeatability, and reproducibility. A study in a single lab can provide a limited assessment of only resemblance and repeatability.