Several procedures designed to reduce nonconformity in interlaboratory studies by shrinking data toward a consensus weighted mean are suggested. Some of them are shown to have a smaller quadratic risk than the vector sample mean. Shrinkage toward a weighted means statistics appearing in random effects model and in models which include systematic, Type B errors is also considered. The results are illustrated by two examples of collaborative studies.
Citation: Journal of the Royal Statistical Society Series B-Statistical Methodology
Pub Type: Journals
Birge ratio, DerSimonian-Laird estimator, heteroscedasticity, key comparisons, Mandel-Paule algorithm, meta-analysis, normal mean, reference value, shrinkage estimators, Stein phenomenon, weighted means statistics.