On the choice of comparison reference values for the purpose of pair-wise comparison of laboratories

A reference value is sought that provides unbiased estimates of the differences between any pair of laboratories. It is found that in general no single-valued reference value exists that yields an unbiased estimate of the difference for all possible pairs of laboratories. In all cases the best estimate of laboratory difference is evaluated strictly in terms of differences in results. It is concluded that the concept of a comparison reference value is both unnecessary and unsuited to the purpose of pair-wise comparisons of laboratories.