ON THE EFFICIENCY OF ALGORITHMS FOR MULTIVARIATE LINEAR CALIBRATION USED IN ANALYTICAL SPECTROSCOPY

The most common multivariate calibration algorithms now in frequent use for analytical spectroscopic methods are reviewed. The full-spectrum techniques which are considered often suffer from spectrum collinearity. This leads to ill-conditioned equation systems for which standard least-squares have to be replaced by estimators such as principal component regression (PCR) and partial least squares (PLS). The efficiency in obtaining optimum prediction models with these two common algorithms is discussed using practical examples.