TABLE OF CONTENTS This chapter is concerned with providing quantitative measures of the quality of test results. Two approaches are discussed and then brought together. The first approach is concerned with estimating the reproducibility of analytical measurements. We saw in Chapter 1 that reproducibility is a measure of the magnitude of the difference that might be expected between two measurements made on the same material, one in each of two laboratories. This concept is important for commercial and regulatory purposes in that it explicitly recognizes that differences are to be expected between measurements generated in different laboratories and it represents an approach to quantifying the magnitude of expected differences. In order to calculate the reproducibility, estimates of the sizes of the within and between-laboratory variation must be obtained; the different components of the total variation are measured by appropriately defined standard deviations. Estimates of these standard deviations may be obtained from an inter-laboratory collaborative trial. This is a long-established approach to providing measures of the fitness for purpose of analytical results.
In recent years a different approach to assessing measurement quality has been gaining momentum. This is concerned with evaluating the 'uncertainty' of measurements, where that word is given a specific technical meaning: 'a parameter associated with the result of a measurement, that characterizes the dispersion of the values that could reasonably be attributed to the measurand' - the parameter is usually a standard deviation. The method essentially involves breaking down the measurement procedure into a series of steps, assessing the...
