| Abstract: | Re‐issuing the same test material, its true identity unknown to participants, in two rounds of the GeoPT proficiency testing programme 18 years apart has demonstrated remarkable similarity, and therefore stability, of consensus values independently estimated for over fifty measurands in each round. A comparison of the two data sets, GeoPT5 (AMH‐1) from 1999 and GeoPT41 (ORA‐1) from 2017, shows that corresponding consensus values, rigorously derived by current procedures for complete compatibility, are for the most part statistically indistinguishable when account is taken of their associated uncertainties. Inferences that may be drawn from this exceptional agreement include: (a) the consensus estimation procedure was robust and gave consistent results; (b) overall, the balance of the contributed data was unaffected by significant changes in: (i) the populations of laboratories participating, (ii) the proportions of laboratories producing results by different techniques, (iii) the manner in which systems were calibrated, or (iv) the aspirations of laboratories to attain higher quality results over time; and (c) unsurprisingly perhaps, that the test material itself had remained stable and unchanged during that time. Additionally, a statistical comparison of the original values for AMH‐1 with consensus values derived by current procedures demonstrates that they are also effectively indistinguishable. |
