*2.8. Statistical Analysis*

The hedonic scores of sensory attributes and the liking of the energy drinks during the tasting session were analysed using an analysis of variance (ANOVA). A Tukey test was used as a post hoc data analysis technique to determine the differences between the samples. The significance level (α) was set at 5%. To analyse the relationship between energy drinks and familiarity, the chi-square test for homogeneity was performed. EsSense Profile® data were analysed using the XLSTAT Statistical Software (XLSTAT Version 2019.4.2, Addinsoft, New York, NY, USA). The frequency counts of 21 emotion words which described the samples were calculated. Cochran's *Q* test was used to find the difference between the samples by evaluating each emotion word used in the CATA questionnaire. The values of purchase intent were statistically analysed for multiple comparisons using the Cochran's *Q* test. Facial expression data were collected through the Affectiva Affdex (Affectiva Inc., Waltham, MA, USA) software. The analysis of variance (ANOVA) technique, through Minitab® 18 (Version 10.0.17763 Build 17763, State College, PA, USA), was used to locate significant differences among emotions. Pearson correlation coefficients (*r*) among the sensory attributes, familiarity, and facial expressions were calculated and plotted using a correlation matrix. Multivariate analysis of variance (MANOVA) was used to determine the significant difference among facial expressions and correlations among all sensory attributes were tested. Based on the MANOVA results, a principal components analysis (PCA) bi-plot was made. The relationship between each emotion method using overall liking as the response variable was analysed using a linear regression model.
