Klaus-Dieter Sommer, Bernd Siebert, Anna-Lisa Hauswaldt
NOVEL AND ESTABLISHED CONCEPTS FOR CONSIDERING CORRELATION IN UNCERTAINTY EVALUATION
Modern uncertainty evaluation is based on both the knowledge about the measuring process and the input quantities contributing to the measurement result. Very often, two or more of the input quantities are not independent from each other. The combined uncertainty can be enhanced or decreased by such correlation. In everyday practice, however, correlation is often ignored since the relevant uncertainty documents do not provide ready-for-use procedures for proper treatment of correlation. The paper provides practical techniques for identifying and quantifying correlation in measurements. Starting from a systematic modelling procedure, a concept is presented that allows to easily include correlation in the measurement model and to properly estimate correlation coefficients or correlated fractions of the related input quantities either from existing (statistical) data or from other (non-statistical, logical) knowledge. Three possible ways to take correlation into consideration when evaluating measurement uncertainty are described and discussed.