Validation Of chemometric analysis of Osijek wheat cultivars

Chemometric model based on principal component analysis (PCA) of eleven Croatian wheat cultivars by evaluation of statistics of residuals is assessed. Evaluated are samples of Žitarka, Super Žitarka, Srpanjka, Barbara, Klara, Golubica, Monika, Kata, Ana and Demetra (selected at Agricultural Institute Osijek, Croatia) and cultivar Divana (Jost-Seeds Research, Križevci, Croatia) produced from harvest in 2000. The model is derived from experimental determination of the following 21 variables: High Molecular Weight Glutenin Subunit (HMW) proportions, chemical and physical properties of wheat and dough, and cultivar bread making qualities. Quality of bread crumbs are evaluated by computer image analysis. PCA analysis revealed that the samples can be projected to the subspace of two latent variables with an average residual error of 10 %. Applied is Q statistics for determination of cultivars which are projected outside the model subspace (i.e. which do not confirm to the model of the two latent variables), and T2 (Hotelling’ s) for evaluation of cultivars which confirm to the model based on the two latent variables, but have “ unusual” properties. Application of the chemometric model for improvement of selection of wheat cultivars and improvement of technology of production of various food products are proposed.