Analysis of the Application of Type B Correlation in Measurement Uncertainty Evaluation

Two cases of the occurrence of a correlation between the results of measuring two input quantities are considered: when both input quantities are simultaneously observed in one measurement experiment (type A correlation) and if the same standards, measuring instruments, and standard data that have significant uncertainty are used to estimate their values (type B correlation). It has been shown that ignoring correlation can lead to significant biases in the measurement uncertainty evaluation. In general form, a measurement equation containing a type B correlation is written. It is shown that the statement declared by GUM that the need to take into account type B correlation can be avoided by redefining the set of input quantities on which the measured quantity depends, by including in it additionally as independent input quantities such that influence two or more input quantities of the original set does not give an equivalent value of measurement uncertainty. An example is analyzed that shows the fallacy of this statement.