Methods of comparing means of measurement

Autor: A. I. Vereskov, V. V. Novikov, B. M. Belyaev, V. I. Belotserkovskii
Rok vydání: 1992
Předmět:
Zdroj: Measurement Techniques. 35:134-139
ISSN: 1573-8906
0543-1972
DOI: 10.1007/bf00978859
Popis: In practice, one cannot always readily obtain a complete comparison graph, but the error estimates may be quite satisfactory for the incomplete graph. We have developed a method of processing the results and a program to implement it on the basis of pair comparisons, which envisages defining the error estimates by least squares. One determines the error estimates for the means of measurement participating in the comparisons from the minimum in the sum of the squares of the deviations and from the variances and the covariance matrix for an appropriate regression model. There is then no need to use working formulas explicitly as in [I], as the latter can be extremely laborious and unreliable for a general comparison graph. However, it may be possible in a radial scheme to obtain comparisons between individual adjacent vertices, or if such information already exists, it can be used in the processing to improve the error estimators. This cannot be done with the formulas in [i]. The program used in our method envisages comparisons for means of measurement in various laboratories, which may be widely separated. In such cases, the comparisons are irregular in type, and it is almost impossible to obtain a complete graph. The program does not plan the comparisons but enables one to carry them out where and when it is practically convenient and desirable, with the data processed after an adequate volume has been accumulated. The method also does not assume the comparison only of equally accurate means of measurement. The comparisons may involve means of measurement differing in accuracy, including substandard ones, or may be based on data from checks on the means of measurement. This considerably improves the reliability in the error estimators. Each pair of comparisons is assigned weight in accordance with the accuracy classes of the instruments when these differ; the weight is the higher the higher the accuracies of the instruments.
Databáze: OpenAIRE
Externí odkaz: