Reliability of Calibrations and Comparisons of Reference Standards with an Application to a Group of Line Standards

Abstract

The problem discussed is that of combining interrelated data which refer to the calibration and intercomparison of a group of standards, to give best estimates for the values of the standards together with their reliabilities. A resume is given of least squares in matrix notation, which is particularly suited to computations using automatic computers, and this technique is applied to results obtained over a period of 15 years at the Bureau International des Poid set Mesures and the National Standards Laboratory, in the calibration and intercomparison of five I m reference line standards of length. One of these scales was a platinumiridium prototype (No. 20), while the remainder were in nickel or 58% nickel-steel. The analysis yields time-dependent values, where appropriate, for the lengths of the scales together with their variances, and an estimate of the observational variance. The estimate of the observational variance, (0·077 /Lm)", is based in part on "external consistency", as the data used in the analysis were obtained in more than one laboratory.