Chemometric utilities to achieve robustness in liquid NIRS calibrations: Application to pig fat analysis

Previous research has demonstrated the ability of Near Infrared Reflectance Spectroscopy (NIRS) technology to predict fatty acids in fats and oils with a high degree of precision and accuracy. Nonetheless, it has been noted that when calibrations are applied to external validation samples, considerable deviations from expected values have been detected, despite sample analytical characteristics being well represented in the calibration set. Deviations appear to be greater as the time elapsing between calibration development and NIRS analysis of new unknown samples increases. This problem, which is common to other spectroscopy applications, appears mainly to affect NIRS analysis of oils and fats. The aim of this study was to evaluate the performance of the repeatability file algorithm for minimising the sources of variation that may prompt unexpected predictions of fatty acids in the NIRS analysis of pork fat. A total of 188 samples of Iberian pork fat used as the calibration set, and 12 further samples as the validation set, were analysed in a NIR (Near Infrared Reflectance) monochromator, using folded transmission cups. The results showed that although the use of a repeatability file did not improve the calibration statistics obtained, it enhanced predictive capacity for new samples whose spectra were recorded two months after calibrations were obtained.