Tutorial review. Deconvolution of mixtures by factor analysis

The deconvolution of mixtures using chemometric techniques is discussed using a simple simulated numerical example and graphical and numerical output from a spreadsheet (Excel). Topics covered include notation, the distinction between observed, predicted, true and expected data, errors, signal-to-noise ratios, integration, quality of reconstructions, principal components analysis (PCA), pre-processing, calculation and significance of eigenvalues, windowing methods, estimation of composition of regions of spectra and chromatograms, projection graphs, variable selection, determining factors without using PCA and determining factors using principal components regression. Considerable emphasis is placed upon method choice including the effect of centring.