Prediction of rice sensory texture attributes from a single compression test multivariate regression, and a stepwise model optimization method

Sensory texture characteristics of cooked rice (92 samples) were predicted using a compression test and a novel multivariate analysis method (that is, Partial Least Squares Regression optimized by a stepwise method), 11 senso ry texture characteristics were evaluated via a trained descriptive panel, and 14 instrumental parameters from a compression test were used in combina tion with Partial Least Squares Regression to evaluate predictive models fo r each of the sensory attributes studied. Among the texture attributes evaluated by the panel, 7 (cohesion of bolus, adhesion to lips, hardness, cohesiveness of mass, roughness of mass, toothp ull, and toothpack) were satisfactorily predicted after the optimization by the stepwise method (optimized Rcal > 0.6).