A new method for estimating consensus values in interlaboratory comparisons

Abstract

This paper presents a new statistical method for estimating consensus values in interlaboratory comparisons or key comparison reference values. The presented method is developed based on the zero sum of the normalized residual (ZSNR) criterion. Thus, it is referred to as the ZSNR estimator. The presented ZSNR estimator accounts for both within- and between-laboratory variability. It does not require the assumption of normality. The presented ZSNR estimator is capable of dealing with inconsistent data or outliers. Three application examples are presented to demonstrate the effectiveness of the presented method in comparison to several existing methods.