The following is an abstract for the selected article. A PDF download of the full text of this article is available here. Members may download full texts at no charge. Non-members may be charged a small fee for certain articles.
Raman and NIR Spectroscopic Methods for Determination of Total Dietary Fiber in Cereal Foods: Utilizing Model Differences
Volume 52, Number 1 (Jan. 1998) Page 32-41
Archibald, D.D.; Kays, S.E.; Himmelsbach, D.S.; Barton, F.E.
This work evaluates the complementarity in the predictive ability of three Raman and three near-infrared reflectance (NIRR) partial least-squares regression (PLSR) models for total dietary fiber (TDF) determinations of a diverse set of ground cereal food products. For each spectral type (R or N), models had previously been developed from smoothed (D0), first-derivative (D1), or second-derivative (D2) spectral data. The NIRR and Raman models tend to have very different sets of outliers and uncorrelated errors in TDF determination. For a single spectral type, the prediction errors of various preprocessing methods are partially complementary. The samples are very diverse in terms of composition, but the main problem groups were narrowed to high-fat, high-bran, and high-germ samples, as well as and those containing synthetic fiber additives. Raman models perform better on the high-fat samples, while NIRR models perform better with high-bran and high-synthetic samples. Raman models were better able to accommodate a wheat germ sample, even though this sample type was poorly represented by the calibration set. Two methods are presented for utilizing the complementarity of the spectral and processing techniques: one involves simple averaging of predictions and the other involves avoidance of outliers by using statistics generated from the sample spectrum to choose the best model(s) for determination of the TDF value. The single best model (N-D1) has a root-mean-squared prediction error of 2.4% TDF. The best model of prediction averages yields an error of 1.9% (combining N-D0, N-D1, N-D2, R-D0, and R-D1). An error of 1.9% was also obtained by choosing a single prediction from the six models by using statistics to avoid outliers. With the selection of the best three models and averaging their predictions, an error of 1.5% was achieved.