We consider the problem of estimating a measurand based on a sequence of measurements. Each measurement has both type-A and type-B errors. The measurements may have been obtained from a single experiment or several separate experiments. We investigate the estimation problem under two measurement models having different sets of assumptions on type-B errors. We review commonly used procedures for obtaining a confidence interval for the measurand and discuss an alternative approach based on the principle of generalized inference. We use several examples to illustrate the proposed approach.
Pub Type: Journals
generalized pivotal quantity, ISO GUM, key comparisons