Year
1994
Abstract
All measurements include some error. Whether measurements are used for accountability, environmental programs or process support, they are of little value unless they are accompanied by an estimate of the measurement's uncertainty. This fact is often overlooked by the individuals who need measurements to make decisions. This paper will discuss the concepts of measurement, measurement errors (accuracy or bias and precision or random error), physical and error models, measurement control programs, examples of measurement uncertainty, and uncertainty as related to measurement quality. Measurements are comparisons of unknowns to knowns, estimates of some true value plus uncertainty; and are no better than the standards to which they are compared. Direct comparisons of unknowns that match the composition of known standards will normally have small uncertainties. In the real world, measurements usually involve indirect comparisons of significantly difference materials (e.g., measuring a physical property of a chemical element in a sample having a matrix that is significantly different from calibration standards matrix). Consequently, there are many sources of error involved in measurement processes that can affect the quality of a measurement and it associated uncertainty. How the uncertainty estimates are determined and what they mean is as important as the measurement. The process of calculating the uncertainty of a measurement itself has uncertainties that must be handled correctly. Examples of chemistry laboratory measurement uncertainties will be reviewed in this report and recommendations made for improving measurement uncertainties.