Evaluation of analytical figures of merit when multivariate calibration is employed in atomic spectrometry

Abstract

Electrothermal atomic absorption spectrometry (ETAAS) is a common choice to measure trace metals in complex materials. However, its combination with experimental design and multivariate calibration to exploit all the information contained within atomic spectra present is uncommon. A reason may be that it was not straightforward to evaluate the corresponding figures of merit of the resulting methodology. In this work, multivariate figures of merit which consider the risk of false negatives and false positives are calculated for the first time in the atomic spectrometry field following the latest ISO and European guidelines: critical level, minimum detectable value, trueness and precision, multivariate analytical sensitivity (or multivariate capability of discrimination) and sample-specific confidence intervals. The practical case study is about quantifying Cu in lubricating oils (a complex problem although relevant to trace machinery wearing) using partial least squares (PLS).