Simple and efficient multivariate calibration of nonlinear absorption spectroscopic data

This paper presents a simple way to extend the repressor variables in partial least-squares (PLS). The method is particularly suited for nonlinear absorption spectra or any other measurements of the soft saturation type. Absorption below a certain limit is described as linear and above this limit as nonlinear. The extension of the regressor variables consists of the absorbances above the linear range. The proposed method is compared with standard PLS. The proposed method gives more parsimonious models and better predictions than standard PLS for the two data sets used in this study, ultraviolet-visible with absorbances in the region 0-2.9.