Comparison of traditional and multivariate calibration techniques applied to complex matrices using inductively coupled plasma atomic emission spectroscopy

Univariate and multivariate (partial least squares 1, PLS1) calibration tec hniques are compared for the determination of Pt, Pd and Rh in a complex sa mple matrix. The univariate techniques utilised either pure standards, pure standards with interelement correction factors applied, or matrix matched standards. Univariate calibration yielded relative root mean square errors (RRMSEs) of prediction for Pt, Pd and Rh of 19, 15, and 54% and 13, 18, and 88% for test solutions and real samples, respectively. Univariate calibrat ion with matrix matching proved the most accurate method with RRMSEs for Pt , Pd and Rh of 2.5, 3.7 and 2.4% for a series of synthetic test solutions, and 12, 2.3 and 8.0% for real samples, respectively. In comparison, the mul tivariate calibration method yielded relative root mean errors for Pt, Pd a nd Rh of 5.8, 3.0 and 3.5% in the test solutions, and 32, 7.4 and 76% in th e real samples. The relative error for the matrix matching and PLS1 techniq ues was dependent upon analyte concentration.