Chemometrics and food chemistry: data validation

Chemometrics has often been applied in food chemistry to cluster and classify samples or to produce models for food quality. In recent years, data on food composition have become important for public health protection and food trades. The quality of the available chemical information on foods is a problem; data obtained with the newer analytical methods is scarce and general knowledge about food composition is poor, judged by published tables on food composition. Moreover, agreement between results obtained by different analytical methods is very poor. To overcome this critical problem, several countries have decided to create data banks on food composition. The analytical results to be stored must be validated. Chemometric modelling is useful for this purpose. Interlaboratory studies allow standardization of methods and the preparation of food reference materials. The classical computation of repeatability and reproducibility does not extract all the available information so that a multivariate approach is necessary to improve the quality of a data bank on food composition.