Comparación de dos modelos de calibración para la determinación de grasas y proteínas en carne de bovino congelada mediante espectroscopia de infrarrojo cercano | Comparison of two models of calibration for determination of fat and protein in frozen bovine meat by near infrared spectroscopy

The content of fat and protein in frozen bovine meat (n = 547) were determined by near infrared spectroscopy (NIRS), under a reflectance regimen. Two models for the calibration of this technique were applied: partial least squares regression (PLSR) and principal component regression (PCR), which were verified by cross validation. For reference, analytical wet determinations were performed according to international norms. In both calibration models, the multiple correlation coefficients (r) and the standard error of calibration (SEC) were satisfactory for each of the components. There were not statistically significant differences between references and predicted values. As a result, any of calibration models (PLSR, PCR) is suitable for the determination of the contents of fat and protein by NIRS, and this technique could be an alternative for conventional analytical methods