Three-way data analysis of a wheat growing experiment using near infrared spectral data

Abstract

An agricultural growing experiment cultivating 10 cultivars at five localities produced whole wheat flour samples that were analysed by Fourier transform near infrared (FT-NIR) spectroscopy. This produced a three-way data array: locality × cultivar × wavelength. By taking smoothing first derivatives and removing a noisy part of the spectra meaningful PARAFAC (PARAIIel FACtor analysis) models could be produced. Two factors could be interpreted spectroscopically, showing that cultivar properties are related in a not so trivial way to average wheat hardness values and that locality variations are straightforwardly related to average protein content. Copyright © 2008 John Wiley & Sons, Ltd.