The relationship between absorption in the near-infrared (NIR) spectral reg
ion and the target analytical parameter is frequently of the non-linear typ
e. The origin of the non-linearity can be widely varied and difficult to id
entify. In some cases, the relationship between absorption and the analytic
al parameter of interest is intrinsically non-linear owing to the very chem
ical nature of the sample or analyte concerned. In this work, various multi
variate calibration procedures were tested with a view to overcoming intrin
sic non-linearity in NIR reflectance. An approach to solving the problem is
suggested, Calibration was done, after transformation of spectra, by using
linear and non-linear techniques. The Linear calibration techniques used a
re partial least squares (PLS) regression (with and without variable select
ion), linear PLS with X projection (LP-PLS) and stepwise polynomial princip
al component (SWP-PC) regression. Non-linear calibration methods included p
olynomial PLS (PPLS) and artificial neural networks (ANNs). Results were co
mpared on the basis of NIR spectra for ampicillin trihydrate samples, where
the simultaneous presence of crystallization water and surface moisture gi
ves rise to intrinsic non-linearity that affects the determination of the t
otal water content in the sample. The best results were obtained by using t
he non-linear calibration techniques. (C) 1999 Elsevier Science B.V. All ri
ghts reserved.