Correction of non-linearities in spectroscopic multivariate calibration byusing transformed original variables. Part II. Application to principal component regression

The addition of non-linear transformations of the original variables to the original data set is proposed to model strong non-linearities with PCR, Th is yields always better results than the ones obtained by using only the or iginal variables. In some cases predictive ability increases, in other case s the same predictive ability is achieved with however smaller complexity, giving more parsimonious and robust models. A correct selection of the PCs to be included in the model improves the results obtained by using the top- down selection, giving equivalent models to those obtained by using polynom ial PCR, but generally with a lower needed complexity, or by using linear P LS regression with transformed variables. Three real data sets, with differ ent sources and degrees of non-linearity have been tested. (C) 1999 Elsevie r Science B.V. All rights reserved.