The increasing search for a reduction in air emissions has provided a challenging analytical scenario. This paper discusses metrologically, the best practices relating to analytical curves for the determination of very low concentrations of sulfur content in diesel S-10 by monochromatic wavelength dispersive X-ray fluorescence spectrometry (MWDXRF). For weighted linear regression, a new strategy to correlate the weights with the output variable is proposed, which is lacking in the literature. The analytical curve 0-500 mg kg-1, built in compliance with international standards, featured heteroscedastic behavior. The results showed that weighted linear regression is the most suitable choice for assessing concentrations at the beginning of an analytical curve, such as 10 mg kg-1 (product specification limit) rather than simple linear regression which holds that the measured results must be closest to the centroid of the regression line. Finally, it was observed that, when a shorter analytical curve is built, it loses both in accuracy and in precision for very low concentrations, as compared to t simple and weighted linear regression.
Keywords
metrological evaluation of the analytical curve; simple linear regression; weighted linear regression; very low concentrations of sulfur content in diesel S-10; MWDXRF
