|Barton Ii, Franklin|
Submitted to: Diffuse Reflectance International Conference Proceedings
Publication Type: Abstract Only
Publication Acceptance Date: 6/20/1996
Publication Date: N/A
Citation: Interpretive Summary:
Technical Abstract: Food labeling legislation has increased the need for more rapid and efficient methods of analyzing the constituents of foods, particularly dietary fiber. An NIR model, was previously developed, using partial least squares regression, to predict total dietary fiber (TDF) in dry milled cereal products (N=90, range in TDF =<1- 52%). The standard error of cross svalidation (SECV) and multiple coefficient of determination were 1.58% and 0.99, respectively. The standard error of performance (SEP) and coefficient of determination for the prediction of TDF in an independent set of cereal samples (n=29) were 1.51% and 0.99, respectively. The model was expanded to include cereal products with high fat (>10% fat, N=20) and high sugar (>20% sugar, N=39). The SECV and multiple coefficient of determination of the new model were 1.70% and 0.98, with 3 factors explaining 92% of the variation. The model was validated using the original independent validation set combined with high fat (N=11) and high sugar (N=10) cereal samples (total N=50). The SEP was 1.45% and coefficient of determination was 0.98 for prediction of TDF in the new independent validation set. Thus, the model for prediction of TDF in cereal and grain products has been expanded to include samples with high fat and high sugar content without loss in prediction accuracy.