Uncertainty Determination for Nondestructive Chemical Analytical Methods Using Field Data and Application to XRF Analysis for Lead

Abstract

Air sampling and analytical methods are developed to provide a basis for decision making. They are evaluated in the laboratory against prescribed fitness-for-use criteria even though laboratory validation does not take into account all possible sources of uncertainty in field application. Field evaluation would be preferable but is complicated by the lack of controlled conditions, which limits the ability to compare analytical methods and to recognize outliers and assess variance homogeneity across the range of interest. The specific situation of evaluating nondestructive field analytical methods against their reference laboratory equivalent is considered here, since the difficulty of providing replicates is obviated in this case. A portable X-ray fluorescence (XRF) analyzer was used to determine the lead content of air filter samples from several workplaces where lead is used or is a contaminant of the process material. The portable XRF method has the advantage of allowing for faster decisions compared with the alternative of submitting the air samples to an off-site laboratory for analysis. Since the XRF method is nondestructive, the same air samples were also subjected to the reference laboratory-based method of analysis. Two statistical approaches were developed specifically to deal with non-normal elements of the data in evaluating the results. The ISO GUM method identifies outliers and then calculates an accuracy range about the true concentration for the remainder of the data. This coverage is then adjusted to account for the rate of outlier occurrence. The bootstrap procedure uses a large number of computer-generated data points that are sampled, with replacement, from the original set including outliers to determine the coverage. No significant difference is seen between the two statistical approaches. Both approaches result in similar coverage and support the adoption of method acceptance criteria specific to field evaluation (a symmetric accuracy range of 35%). The portable XRF analyzer met this criterion when used with several different sampling methods and thus could be used as a method for routine evaluation of compliance with lead limit values. As the method is nondestructive, further analysis of air samples with analytical results near decision points is possible.