Selection of calibration mixtures and wavelengths for different multivariate calibration methods

Abstract

A comparative study to select calibration mixtures and wavelengths in multivariate calibration methods was made. The methods studied were classical least squares (CLS), inverse least squares (ILS), partial least squares (PLS), principal component regression (PCR) and Kaiman filter. For each method the calibration samples were selected from a total random population of 37 calibration standards, taking into account the standard error of prediction (SEP). The selection of analytical wavelengths for each method was carried out using different criteria: the condition number for CLS and Kalman filter methods, the signal-to-noise ( S N ) ratio and the condition number for ILS method, and all the previous criteria and the full spectrum for PCR and PLS methods. The best results were obtained using the condition number as criterion to select the analytical wavelengths. The study has been applied to the spectrophotometric determination of four priority pollutant chlorophenols.