When analyzing a sample statistically, the quality of the resulting parameters (measures of location, dispersion, etc.) depends on the reliability or credibility of the data and on the analysis undertaken. The data provided by interlaboratory comparisons, and especially key comparisons, is of sufficient importance that approaches are required which are as defensible and robust as possible. Central to the area of key comparisons is the degree of equivalence of measurement standards, which is the degree to which the standards are consistent with reference values determined from the key comparisons and hence are consistent with one another. This paper indicates some initial thinking at the National Physical Laboratory on methods for determining key-comparison reference values. It is based on the concept of making appropriate assertions relating to the measurement standards provided in a key comparison and devising and utilizing statistical techniques that are consistent with them. The concept is illustrated by applying it to data from a spectral-irradiance key comparison. The paper does not attempt to advocate a specific solution, but to convey what is believed to be an appropriate attitude to data analysis in this area.
展开▼
