| Low-moisture dairy products and air-classified bean powders were analyzed by conventional methods and by near infrared reflectance (NIR). For 41 spray-dried milk samples prepared by premixing whole milk with skim milk at various proportions, the correlation coefficients (r) between the data of the two analytical procedures were 0.7889, 0.9672, 0.9892, 0.9437 and 0.9528 for moisture, protein, fat, ash and lactose contents, respectively. For six random spray-dried milk samples the corresponding r values were 0.8384, 0.9282, 0.9908, 0.9886 and 0.9882, respectively.; The r values of 31 samples prepared by mixing whole milk powder with skim milk powder were 0.8208, 0.9844, 0.9956, 0.9846 and 0.9959 for moisture, protein, fat, ash and lactose, respectively. For 6 unknown samples the r values were 0.6017, 0.9945, 0.9514, 0.8625 and 0.9628 for moisture, protein, fat, ash and lactose contents, respectively.; Eleven commercially dehydrated milk powders had r values 0.8507, 0.9878, 0.9992, 0.9949 and 0.9888 for moisture, protein, fat, ash and lactose contents, respectively.; For 40 commercial cheese powder samples the r values were 0.7494, 0.9352, 0.9803 and 0.7486 for moisture, protein, fat and ash contents, respectively.; Thirty air-classified bean flour samples were analyzed for moisture, protein, and ash contents using conventional methods and NIR. The r values were 0.8971, 0.9878 and 0.9571 for moisture, protein and ash contents, respectively. Six unknown samples had the corresponding r values: 0.7467, 0.9394 and 0.9699, respectively.; F-values obtained from the analysis of variance of conventional and NIR data for each food components were smaller than F-table values in all cases except for the moisture content of the random spray-dried milk samples. The F-values indicated that there was no significant statistical difference between NIR data and that obtained by the conventional procedures.; The NIR method is rapid (30 seconds), involves little sample preparation, and provides a direct read out. The test is not destructive and requires 8-12 g sample. Accurate calibration of the NIR instrument with a reliable reference method of analysis is necessary for consistent and meaningful data. |
