Tempo de vida de prateleira de produtos alimentícios levando em conta erros de avaliação

Abstract

This dissertation is a follow up of the work done by Freitas, M. A.; Borges, W.; Ho, L. L (2003) in which a Weibull model was proposed as the underlying distribution for the shelflife of food products based on sensory evaluation data. In this approach the model incorporates misclassification errors that can occur when applying sensory evaluationmethods for shelf-life determination. The study used Bayesian techniques. As a first approach to this problem, the exponential distribution was used to model the failure times (or shelflife times). A simulation study was implemented to evaluate the effect of misclassification errors on the credibility intervals for percentiles. This effect was studied using informative and non-informative (in this case Jeffreys distributions) priori distributions. Finally, theapproach was applied to the real data set presented in the work of the authors mentioned above.Key-words: misclassification errors; shelf life; Jeffreys priori; Bayesian inference; SIR