Data pooling and key comparison reference values

Abstract

Key comparison data are often reported, as a matter of policy, with respect to a reference value calculated from the pool of the participants' results. Consideration of the comparison data sets, particularly with regard to the consequences and implications of such data pooling, can justify the choice of a simple statistical reference value.Recent key comparisons in thermometry are examined to illustrate the nature of the problem, and the utility of a simple approach to creating pooled data distributions. An objective criterion is introduced to help determine when a comparison can safely be described as a single population having a key comparison reference value (KCRV). Graphical techniques of data pooling are presented as an aid to harnessing expert opinion on this question.

Comparison of KCRV and Its Uncertainty of CCM Key Comparisons by Monte Carlo Method

“…For example, A. Steele et al applied the pair-difference Chi squared tests with MCM for KCs analysis. We agree that the application of MCM can provide physical insight into some of the occasions when it is inappropriate to average measurement results statistically [19].…”

Key comparisons: the chance for discrepant results and some consequences

“…Table 1 reports a synthesis of the data collected. The corresponding typical overall mixture [3]- [5] (or pooled [6]) distributions of several CCT KCs is shown in Figure 1, together with the distributions of three common location parameters. Figure 2 shows a similar situation for a CCQM KC [7].…”

“…The terms 'error' and 'uncertainty' should be used properly and care taken to Actually, they are compensations for 'deviations' of measured from a 'reference condition' of specified quantities, 5 due to perturbations caused by other influence quantities producing off-set conditions, called systematic effects (the term 'bias' is not used in the GUM). 6 In short, the reference condition can be defined as the one where all the perturbations caused by other influence quantities are null (thus no corrections).…”

“…Notice that this definition of "influence quantity" is not the VIM3 one (2.52) "quantity that, in a direct measurement, does not affect the quantity that is actually measured, but affects the relation between the indication and the measurement result" (emphasis added), but the GUM one, clause B.2.10 "quantity that is not the measurand but that affects the result of the measurement". 6 By definition, once the influence quantities have been identified in each specific case, no other quantity outside the set of the input quantities plus the quantities responsible for systematic effects/errors can significantly influence the value of the measurand.…”

Simplicity with advanced mathematical tools for metrology and testing

“…It is also referred to as the all-pairsdifference reduced chi-squared, v 2 APD [11,2,9]. Although Eq.…”