Main content area

Approach to the quantification of milk mixtures by partial least-squares, principal component and multiple linear regression techniques

Rodriguez-Nogales, J.M.
Food chemistry 2006 v.98 no.4 pp. 782-789
milk, ewe milk, goat milk, food composition, standards of identity, product authenticity, quantitative analysis, blended foods, least squares, principal component analysis, regression analysis, calibration, casein, capillary electrophoresis, mathematical models, prediction, seasonal variation, model validation
Four of the most widely employed multivariate calibration methods, partial least-squares regressions (PLS-1 and PLS-2), principal component regression (PCR) and multiple linear regression (MLR) were applied to predict the percentages of ternary mixtures of cow's, ewe's and goat's milk based in the analysis of casein fraction by capillary electrophoresis. The prediction models were calculated by using three batches of 10 milk mixtures each prepared in three different seasons and were validated by applying them to the analysis of nine milk mixtures. All the models were good for the prediction of percentages of milk of each species. However, it was found that MLR led to more precise predictions than the other multivariate calibration methods with a root square error under 1.2%.