*2.4. Data Analysis*

Overall liking for each sample on the hedonic scale, purchase intent and willingness to pay were analyzed with ANOVA, and Fisher's LSD was used for paired comparisons, with a significance level of 0.05. The CATA analysis for emotions and emojis was performed with the Cochran's Q test and correspondence analysis using XLSTAT (Addinsoft, New York, NY, USA: version 2020.1.1). Correspondence analysis was carried out by the software using the variables generated in the Cochran's Q test. Fisher's exact test was applied to 2 × 2 cross-tables for comparison of the terms in CATA analysis between the two cultures. ANOVA was carried out on biometric emotions, and the PCA biplots were constructed using XLSTAT (Addinsoft, New York, NY, USA: version 2020.1.1). The linear mixed model for each method was created using REML in GENSTAT (VSN International, Hemel Hempstead, UK; version 16), after selecting the top descriptors using all-subsets regression

(*p* < 0.05). Overall liking scores generated from hedonic liking were taken as response variables, and the emotion/emoji/biometric variables were taken as fixed factors. The random factors in the mixed model were "participant codes" and "product codes". If the fitting of the model did not converge numerically, one or both random factors were omitted to obtain convergence. A multi-factor analysis (MFA) was carried out for a combination of CATA emotions, CATA emojis, FER and overall liking using XLSTAT, with the mean values for each of the factors.
