P.
MEASUREMENT UNCERTAINTY AND NIOSH METHOD ACCURACY RANGE by David L. Bartley, Stanley A. Shulman, and Paul C. Schlecht, NIOSH/DART CONTENTS 1. 2. 3.
4.
5. 6.
7. 8.
1.
Page
Introduction . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 208 ISO GUM . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 209 The Symmetric Accuracy Range as Used by NIOSH . . . . . . . . . . . . . . . . . . . 212 a. Definition and Approximation . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 212 b. Uses of the Symmetric Accuracy Range . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 213 Uncertainty and Analytical Lab Procedures . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 215 a. Validated Method Adoption . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 215 b. Pooled Quality Control Results . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 218 c. Continual Method Re-Evaluation . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 218 d. Limit of Detection and Detection Limit . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 218 Discussion . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 217 Technical Notes . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 219 Note 1: Example of accuracy range confidence limit . . . . . . . . . . . . . . . . . . . . . . 220 Note 2: Single-evaluation correction of bias . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 221 Note 3: Characterizing effects of uncorrected bias . . . . . . . . . . . . . . . . . . . . . . . 224 References . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 225 Terminology . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 227
INTRODUCTION Recently, the ISO Guide to the Expression of Uncertainty in Measurement (GUM) has come close to being universally adopted as the standardized way to characterize and document measurement uncertainty [1-5]. Since the mid-1970s, accuracy criteria have been an integral part of the evaluations of the sampling and analytical methods used by the National Institute for Occupational Safety and Health (NIOSH), the Occupational Safety and Health Administration (OSHA), the Mine Safety and Health Administration (MSHA), and others. NIOSH has previously published extensive discussions addressing the issue of accuracy as a factor in the development, evaluation, and characterization of analytical methodology. Both traditional method accuracy and new measurement uncertainty concepts are intended to communicate measurement limitations to laboratory clients. Naturally, laboratories are interested in how NIOSH accuracy requirements [6-7] relate to measurement uncertainty. This chapter provides guidance for achieving consistency in determining measurement uncertainty by those laboratories using NIOSH methods. Minor modifications to NIOSH accuracy measures, and an expansion of ISO GUM to cover situations unique to workplace atmosphere measurement can improve consistency and utility. ISO GUM proposes pooling estimated variance components from diverse error sources. The square root of the pooled variance estimate is termed the combined uncertainty uc . Multiplication of uc by a coverage factor k (generally in the range of 2 to 3) results in an expanded uncertainty U. The purpose of the expanded uncertainty is for each measurement to provide an interval bracketing the measurand (the true value of what is to be measured)
3/15/03
208
NIOSH Manual of Analytical Methods
