Selection of calibration set samples in determination of olive oil acidity by partial least squares–attenuated total reflectance–Fourier transform infrared spectroscopy

Abstract A chemometric method has been applied for the determination of the free fatty acid (FFA) concentration in commercial olive oil samples of different types an origins by using Fourier transform infrared spectroscopy (FTIR) attenuated total reflectance (ATR) measurements. Different methods for selecting the training set, including hierarchical cluster analysis, were applied and compared. The prediction capabilities of partial least squares (PLS) multivariate calibration methods, net analyte signal (NAS) preprocessing followed by PLS or classical least squares (CLS) regression method of ATR–FTIR data were evaluated. Several aspects, like spectral range to be considered, different preprocessing alternatives (mean centering, multiplicative scattering correction, standard normal variate (SNV)), together with a critical evaluation of the calibration set were made on using the mean square error of cross-validation and prediction, as control parameters. Using a calibration set of 16 samples the properties of 28 samples were predicted with relative precision of triplicates of 0.017 wt.%. The mean difference between predicted and actual values and the standard deviation of mean differences were −0.001 and 0.037 wt.%, respectively. The limit of detection (LOD), sensitivity and selectivity of the methodology developed were evaluated in terms of the net analyte signal, being found a limit of detection of 0.072 wt.%, a sensitivity value of 0.077 in terms of analytical signal per unit of concentration, being expressed that in  wt.%, and a linear relationship ( R 2 =0.9963) between selectivity and FFA concentration (equivalent to 0.24% for a sample containing 1 wt.% of FFA).