Cereal Chem 61:327 - 329.  |  VIEW ARTICLE
Mathematical Treatment of Near-Infrared Reflectance Data for the Estimation of Protein.

F. S. Lai, Y. Pomeranz, D. Traylor, and S. Afework. Copyright 1984 by the American Association of Cereal Chemists, Inc. 

Protein was determined by Kjeldahl and near-infrared reflectance methods on corn from two crop years, and in two-and six-row barley and malt from numerous locations from one crop year. Kjeldahl protein values and reflectance values obtained at six wavelengths on calibration samples were used to prepare linear regression lines (calibrations) by including all significant terms from multiple regression equations (up to six values) with and without interaction terms up to the sixth degree. A slight but consistent improvement was seen in correlation coefficients, prediction ranges, and standard errors of estimate when interaction terms were included. The improvement was valid for predicting protein content in corn samples from the same or from different crop years, in barley or malt of two-or six-row types, and for predicting protein in malt from protein in barley.