*2.5. Statistical Analysis*

A two-way ANOVA (environment × order) with a generalized linear model (GLM) and a post-hoc Tukey's Honest Significant Difference (HSD) test were used to asses significant differences among the evaluated environments (traditional booths, bright-restaurant, dark-restaurant, bright-VR, and dark-VR) for the hedonic ratings and intensity scores of the Cabernet Sauvignon wine. The order effect was included in the ANOVA model to test whether the position of the samples was a significant factor when tasting the wine. A penalty test on the JAR data was performed to determine the effects of sensory attributes on the hedonic liking. Mean drops for the "too much" and "too little" scores were calculated (differences between the liking mean for the JAR level minus the "too much" or "too little" levels). For the CATA frequency data, correspondence analysis and principal coordinate analysis were used to assess the differences among the evaluated environments relative to the selection of the emotion terms and overall liking levels. For the purchase intent, the Cochran's Q test and simultaneous confidence intervals testing were used for multiple comparisons. A principal component analysis (PCA) was applied to interpret relationships between the hedonic ratings and intensity scores of the wine in different environments. A product-attribute biplot was used for the illustration of the PCA. Data were analyzed at α = 0.05 using the XLSTAT Statistical Software version 2017 (Addinsoft, New York, NY, USA). All data were reported as mean values with standard errors.
