Abstract:
|
Few measurements are only made for their own sake but instead often provide an invaluable quantitative basis for making decisions. While much effort has been expended in the last decade or so in giving guidance about how to evaluate measurement uncertainty [1], arguably less attention has been paid to the treatment of uncertainty in conformity assessment and many questions remain as yet unanswered, such as a clear and objective rule for decision-making. In seeking a criterion for "fitness of purpose" in analytical measurement, a decision theory approach has been introduced by Thompson and Fearn [2], where the costs of analysis are balanced against the costs associated with the consequences of incorrect decision-making. This is considered to be an important step, not only in analytical measurement but also in a wider measurement context, towards establishing clearer procedures for setting and specifying tolerances and associated uncertainties, and in facilitating acceptance of conformity by both customer and supplier. Traditional treatment of the risk associated with incorrect decision-making caused by measurement uncertainty, including concepts such as "shared risk" and "guard-banding" in percentage terms, are arguably not always readily understood by a layman. An optimised uncertainty methodology in economic terms instead offers in this respect increased clarity for the decision-maker, be he end-user, supplier or consumer. What appears to be statistically an equal sharing of risks between consumer and supplier in purely percentage terms may be rather unfair when the different economic consequences are weighed in. The loss function approach has recently been introduced [3] to the optimisation of measurement capability and sampling in conformity assessment and testing of measuring instruments in the framework of legal metrology. Legal metrology covers the measurement in society of many important quantities, such as the utilities (electricity, water, heat, gas, etc) but has recently been extended to other important societal measurements, such as the environment in the form of exhaust gas analysers. Results include a new optimised sampling uncertainty methodology which extends traditional attribute sampling plans to include economic assessments of the costs of measuring, testing and sampling together with the costs of incorrect decision-making. [1] ISO et al., Guide to the Expression of Uncertainty in Measurement, ISO, Geneva, 1995. [2] M Thompson and T Fearn, Analyst, 121, 275, 1996. [3] L Pendrill and H. Källgren, Advanced Mathematical and Computational Tools in Metrology, Lisbon, June 2005, World Scientific, Singapore. In press.
|