Accurate measurements of analytes at low to very low concentrations require accurate corrections for chemical blanks or other sources of background. While there are several different statistical approaches for obtaining an unbiased blank-corrected concentration estimate, not all of these approaches are equally good for estimating the uncertainty. The difficulty in uncertainty estimation arises because the blank or background measurements typically do not depend on the mass or volume of a sample, as the concentration does, and because it is impractical to use samples of exactly identical mass or volume. This talk outlines a statistical approach to blank correction based on linear regression that easily handles the estimation of the analyte concentration and the assessment of the uncertainty in the measurement result arising from the determinations of the analyte in both samples and blanks.
Statistical Engineering Division/ITL
Robert D. Vocke Jr., Jacqueline L. Mann, and W. Robert Kelly
Analytical Chemistry Division, NIST