Compositional analysis of nonfat dry milk by using near infrared diffuse reflectance spectroscopy

Proximate data from 82 nonfat dry milk (NFDM) samples were correlated with near infrared reflectance (NIR) measurements. The best wavelengths for determining constituent concentrations were chosen from 19 preselected filters by using linear regression analysis. The correlation coefficient (r) was 0.971 and the standard error of prediction (SEP) was 0.274 when the predicted values (from NIR measurements) using the 3 wavelengths selected for determining moisture content were compared with laboratory values; r and SEP were 0.961 and 0.099, respectively, when the predicted values using the 4 wavelengths selected for fat content were compared with laboratory results; 0.887 and 0.594, respectively, using the 4 wavelengths selected for lactose content; 0.905 and 0.438 using the 8 wavelengths selected for protein (micro-Kjeldahl) content; and 0.911 and 0.509 using the 7 wavelengths selected for protein (dye binding). These data indicate that NIR can be used to estimate moisture, fat, lactose, and protein content of NFDM.