Model-based approach for evaluation of pooled measurement data

In metrology, often there arises a task of joint data processing of groups of measurement results. The most striking example is an analysis of data obtained in interlaboratory comparisons. The interlaboratory studies refer to the complex and expensive researches in metrology, which makes the stages of planning measurements, data processing and interpretation of results especially important and meaningful. When processing the data of interlaboratory comparisons, the main mathematical tool is methods of the theory of probability and math statistics. There exists a great variety of methods developed for evaluation of data of the interlaboratory studies. So it is important to choose the method correctly proceeding from the purpose of a given interlaboratory studies and taking into account available a priori information, as well as the final task of the studies. The paper contains a comparative analysis of widely used methods of data evaluation and discussion of new methods. Particular attention is paid to formation of a measurement model taking into account available a priori information, and to further usage of the measurement results. The methods considered are applicable not only in the evaluation of interlaboratory comparison data, but also in the general case when joint processing of data from different groups is performed.