DATA ANALYTICAL METHODOLOGIES IN THE DEVELOPMENT OF A VOCABULARY FOR EVALUATION OF MEAT QUALITY

Abstract

ABSTRACTPork patties from M. longissimus dorsi and M. psoas major were used by a sensory panel to develop a descriptive vocabulary to describe warmed‐over flavor development. The patties were made from meat from animals supplemented with one of four dietary treatments, (I) control diet, (II) supplemental iron (7 g iron (II) sulfate/kg feed), (III) supplemental vitamin E (200 mg dl‐α‐tocopherol acetate/kg of feed) and (IV) supplemental vitamin E + supplemental iron. The sample set used for training reflected the variation in muscle type and degree of warmed‐over flavor development (day 1 and 5). The sensory terms selected had to be relevant to the samples, discriminate between the samples, have cognitive clarity and be nonredundant. Based on these selection criteria an initial training list of 36 sensory terms consisting of odors, flavors, basic tastes and aftertastes was developed in a pretrial using experts with product knowledge. This list constituted the start of training the panel. Principle Component Analysis (PCA) and assessor suggestions were used for reducing the number of terms and after 5 days of training this list was reduced to 21 sensory terms. The discriminative ability of the sensory panel improved over the course of sensory training and was quantified by using the mean assessor signal to noise ratios (S/N) for the sensory terms for each training session. This signal to noise ratio is essentially the total initial variance divided by the residual cross‐validated variance. Subsequent detailed multivariate analysis found that the bilinear descriptor leverage was a particularly efficient method for term reduction.