The effect of preprocessing methods in reducing interfering variability from near-infrared measurements of creams

This work is part of a study in which the possibility of NIR combined with some chemometrical methods is investigated as a suitable technique to classify clinical study samples of a cream. In this study, the influence of different preprocessing methods on the removal of spectral variations due to some variance sources has been investigated. The applied preprocessing methods are standard normal variate (SNV), detrend correction, offset correction, and first and second derivation. The investigated variance sources are different batches of ingredients, different samples of the same batch, different days and different positions of the sample cup in the sample drawer of the instrument. A nested ANOVA design has been applied in order to quantify the variances introduced by these variance sources. Since ANOVA is a univariate technique, the necessary variable (wavelength) selection has been performed by the Fisher criterion. The best results, i.e. largest reduction of interfering variability and clearest distinction between different clinical study samples, are obtained with the second derivative spectra.