Measurement uncertainty and optimized conformance assessment

This paper is concerned with the assessment from measurements of whether or not a product meets its specification. Given measurement data and a product specification, a finite number of choices have to be made, e.g., accept the product, re-measure the product with a more accurate measuring system, re-engineer the product, reject the product, etc. Acknowledging that the measurement information provides only partial information about the true product characteristics, we require decision rules that make best use of the measurement data and minimise the negative consequences associated with making a wrong decision. In this paper we apply Bayesian decision-making approaches to conformity assessment, using a loss function to quantify the cost of wrong decisions and deriving optimal decision rules that minimise the expected loss. One important aspect is that the presence of unknown systematic effects associated with the measurement system has a significant influence on the behaviour of the expected loss.