Detection of inhomogeneities in sets of NIR spectra

Methods for the detection of outliers and clusters in data sets for multivariate calibration are discussed in this study. Replicate as well as sample outliers have been investigated. The Cochran test applied to the sum of absorbances is shown to identify outlying replicates. Sample outliers were detected on latent variables (PCs) or by using Rao's statistic combined with Grubbs' and Dixon's tests. The conclusions were then verified by the modelling. The tendency to cluster was evaluated by Hopkins statistic.