Abstract
Variability introduced into numerical measurements by a measuring instrument is referred to as measurement instrument repeatability. Measurement instrument repeatability estimation methods exist when the measurements are repeatable. However, when measurements are destructive and repeated measurements are not possible, estimating measuring instrument repeatability is difficult since repeatability is confounded with, and usually cannot be separated from item variability. In this paper an estimator for destructive measuring instrument repeatability is obtained from measurements of two different item types, under the assumptions of constant item type measurement coefficient of variation, constant repeatability, and independent and normally distributed measurements. The estimator’s variance is also derived, from which a confidence interval for repeatability can be computed. The results add to the body of existing methods that exploit some assumed pattern of item variability, or alternatively make other assumptions needed to estimate measurement instrument repeatability when repeated measurements are unavailable.