A Criterion for Comparing Measurement Results and Determining Conformity with Specifications

In this paper a new criterion for comparing measurement results and determining conformity with specifications is proposed, which essentially is a strategy of estimating the empirical relationships of objects. Comparing with traditional methods given in GUM: 2008 and ISO 14253-1, this criterion improves the resolution of comparison by reducing the sizes of the coverage intervals to be compared. Interval order (a binary relation) is used for comparing the coverage intervals of the measurand and represents the empirical relations. The systematic effects of measurement are classified into two types: monotonic and non-monotonic effects, so that, without correcting the monotonic effects, a biased measurand can be specified to represent the empirical relations. Thereby the uncertainty components arising from the monotonic effects can be removed from the combined uncertainty. A strategy is given for determining the relationships among measurement results and specification limits. An example is given to demonstrate the application of the criterion.