Comments on ‘Bayesian evaluation of comparison data’

A recent paper by Ignacio Lira in this journal (2006 Metrologia 43 S231–4) proposes a new procedure to evaluate the data from a simple key comparison where a travelling standard of a stable value during the comparison is independently measured by the participants. In particular, Lira presents an expression, which he claims to be the kernel of a Bayesian posterior probability density function for the value of the measurand. Lira claims that his expression encodes the collective state of knowledge whether or not the results are consistent. Thus Lira offers an alternative to the method recommended by an Advisory Group on Uncertainties commissioned by the BIPM. We discuss Lira's procedure from the viewpoint of what we believe to be a good practice of Bayesian statistics.