A uniform approach to determine measurement uncertainty in routine chemical forensic casework.

Abstract

In routine forensic chemical casework where measurements are performed on reference materials, determination of measurement uncertainty is described in several guidelines. The proposed methods often have the drawback that they are not derived from a statistical framework and may lead to conservative confidence intervals. Furthermore, the formulas involved may vary considerably for different types of reference material. The current study presents a method for determination of measurement uncertainty that is both statistically sound and uniform for different types of reference material, namely material from proficiency tests, certified reference material, and noncertified reference material. The statistical model used for measurements uses relative standard deviations. The method is based on analysis of two sources of uncertainty, namely the random variation in the measurement itself and uncertainty connected to the bias of the process. It is explained how the method works when there is bias correction of results. It is also described how the method works if no correction is applied. Furthermore, the results are compared using simulation experiments, showing a better performance of the proposed method when compared with commonly used alternatives.