Different possibilities of treatment for NIR spectra were analysed, including usual parametric statistical methods in order to ascertain the affiliation of an individual measurement to a certain data set for the same product. These methods were compared with the results obtained by using NSAS procedures for qualifying individual samples. Both the training set used for generating the fingerprint of a product and individual measurements were represented in a polar co-ordinate system to permit a quick view of fingerprint zone and position of “suspicious” individual measurement.
