Exemples d'interprétation de résultats de dégustation

De nombreux auteurs ont proposé des méthodes et des tables pour l'interprétation de résultats de dégustation, notamment ROESSLER et al. (1978) pour les tests de différence et KRAMER (1956, 1974) pour les tests de classement. Ces dernières ont été remises en question par JOANES (1985) et SMITH (1985) préconise le test de FRIEDMAN. Les méthodes d'analyses multidimensionnelles (AC.P. et A.F.C.) permettent l'interprétation de dégustations descriptives (notation ou appréciation d'un vin sur plusieurs paramètres). Des exemples sont donnés pour développer l'utilisation de ces méthodes en oenologie. +++ Statistics seem to be essential in a sensory valuation, not only to interpret the results but also to define the experimental protocol, to distribute samples in a balanced way, to avoid the effect of sytematic mistakes (SAUVAGEOT 81). A lot of authors proposed methods of calculs and tables: KRAMER (1956), KRAMER et al. (1974), ROESSLER et al. (1978), LARMOND (1969), TOMASSONE et FLANZY (1977). A. JOANES (1985) has just called in question again validity of KRAMER's tables (very used in œnology) it was important for us to present with few examples, some methods of interpretation of difference and preference tests. We especially use « duo trio and triangle tests ». The last one, more difficult, must be done with a expert panel. ROESSLER and al. (1978) proposed tables giving critical values (number of correct answers) at different probability levels. So we advise the FRIEDMAN's test; it's a non parametric method, with which you can know if one wine is different or not. Two wines are different if the difference between their rank sums is rank above a value determined with the NEWMAN's test. The descriptive tasting (appreciation of a wine with several variables: colour, arome intensity, sourness, bitterness...) give more informations. Two methods of multidimensional analysis are proposed to take into account the entire information: - The principal components analysis if the results are presented as quantitative variables, - The correspondance analysis if the results are presented as qualitative variables. Examples of interpretations are proposed to develope the use of theses methods in œnology. That's easier with the introduction of the micro-computer.