Statistical Indices for the Selection of Food Sorption Isotherm Models

Abstract

A series of statistical indices were proposed for the adequate selection of a food sorption isotherm model. These indices are based on four rejection criteria, which measure the statistical significance and precision of the parameter estimates and the assessment of the regression assumptions, whereas a lumped index quantifies the fitness quality of the models. The suitability of these statistical indices for the model selection was tested through the mathematical description of two sets of sorption isotherm data taken from literature for almond powder and yogurt and one set of our own data for chayote. Eight sorption isotherm models were used to describe the experimental data. The statistic for evaluating the randomness of the moisture residual series together with the optimality criterion for measuring the size of the joint confidence region and the significance test of the model constants were shown to exhibit the highest model discrimination properties among the applied model selection criteria. The results demonstrated that an appropriate model for the mathematical description of sorption isotherm data may be chosen on the basis of the proposed indices. Furthermore, the results suggest that the proposed indices may be used in other food modeling applications.