A new approach to evaluating inter-laboratory testing precision

Abstract

A new approach to the evaluation of inter-laboratory testing precision has been developed based upon the analysis of a large inter-laboratory testing program on Mooney viscosity. The new approach (based on demonstrated non-random or ‘biased’ inter-laboratory test differences) consists of : (1) the adoption of newly proposed ‘between-lab’ and ‘within-lab’ statistical parameters, h and k, as developed by ASTM Standard E-691 (a standard practice for evaluating precision) and (2) a number of special graphical analysis techniques that allow for an identification of the good, the intermediate and the poor laboratories on the basis of both their between-lab (precision), reproducibility, R, and their within-lab (precision), repeatability, r. The h parameter indicates how the deviation d, of an individual laboratory test result from a reference value (the overall average measured property), compares to the range of deviations for all laboratories and the k parameter is a ratio of the within-lab variation of any laboratory to the overall level of variation of all laboratories. The deviation values may be used to prepare Deviation Profiles that show how the various laboratories rank; negative to zero to positive ‘biases’ or deviations compared to the reference value. Analysis of the relationships of: (1) h and the standard deviation of h across all the test materials, (2) k and the standard deviation of k across the materials and (3) average h and average k for all materials, provide valuable information for the diagnosis of laboratory testing behavior. The INTERCAL concept, a procedure for using independently obtained calibration curves to correct for the biased between-lab differences, gives a substantial reduction in the between-lab reproducibility, R; provided that: (1) all laboratories have good within-lab precision or small r and (2) that testing proficiency is maintained in a stable condition over a period of time. A global Quality Testing concept has been proposed based on the ability of this new approach to classify laboratories into one of three grades: Grade A, laboratories that have good inherent agreement with each other and to reference values; Grade B, laboratories that may be corrected by INTERCAL to the same level of agreement as Grade A, and Grade C, laboratories that cannot be corrected by INTERCAL because of poor within-lab control and resulting high within-lab variation.