Eureka! Institutional Repository

In the present study, the reliability of the sensory evaluation of wines was investigated. The reliability was related to the tasters' evaluation and the bottling evaluation for each wine separately (42 wine codes in total). Evaluation agreement was applied using Cronbach's α coefficient. Multiple regression analysis was used to investigate the potential influence of objective organoleptic variables on hedonic organoleptic variables. Principal component analysis (PCA analysis) was applied to find the most important sensory variables characterizing each type of wine. While the Cluster Analysis was used to examine the classification of wines into common groups (clusters) based on their common characteristics. According to the results for the evaluation of the tasters, using the Cronbach's α coefficient of agreement, most of the wines and bottlings (except rosé and sparkling wine) obtained a reliable value (α>0.700) without the application of a corrective measure. One wine code did not obtain a satisfactory Cronbach's α value when evaluating the bottlings, and one when evaluating the tasters even after corrective measures were applied. The technique of excluding taster scores as a corrective measure imposition, except in one case, positively increased the Cronbach's α coefficient in the bottling evaluation. The results of the Multiple Regression Analysis on the wines showed that the majority of the hedonic variables of the wine types were satisfactorily described by the objective variables. The hedonic variables were also influenced by nonhomologous objective variables. Common objective variables were also found to influence hedonic variables in most wine types. PCA was used to discriminate the main sensory variables of each wine type, while Cluster Analysis was used to classify wines into groups based on their related characteristics by type.