The objective of this study was a comparison of the performance of Fourier transformed near-infrared
(NIR) with Fourier transformed mid-infrared (MIR) spectroscopy using attenuated total reflectance for
the multivariate determination of compositional parameters in wheat bran samples. These parameters
were the contents of water, protein, ash, starch, soluble as well as insoluble dietary fibers, and lipids.
Partial least squares (PLS) regression was used to construct a calibration model. NIR was found to
perform better for ash, starch and soluble as well as insoluble dietary fiber, while MIR was superior only
for protein. The scores for water and fat were about equal. The prediction results for ash, insoluble dietary
fiber and fat were good. The analysis of soluble dietary fiber suffered from the inaccuracy of the
underlying wet-chemical reference method, which had a negative impact on the calibration. Starch was
prone to a large relative error despite a good coefficient of determination. Protein and water gave
acceptable relative errors but suboptimal goodness of fit. From the precision achieved with a limited
sample set, it can be concluded that infrared spectroscopy is an appropriate method to establish a rapid
analysis of wheat bran. In general, NIR seemed to be the superior and more robust method.