3.1.1. Hedonic Ratings

The sensory attributes measured on 9-point interval scale were analyzed under the null hypothesis that all treatment groups were taken from populations with the same median, and they do not evoke different liking in consumers. Friedman's two-way nonparametric analysis of variance (ANOVA) was used to evaluate the effect of context environment and chocolate type on the response. By using the ANOVA procedure in conjunction with the PROC RANK procedure, data was ranked 'not' by block to analyze the effect of context, whereas traditional method of rank assigning was used, i.e., rank within blocks for the chocolate type. As Friedman's test cannot be used for interaction, another method was used, where the non-normal distributions were transformed by NORMAL = BLOM function of PROC RANK [27]. This function computes normal scores from the ranks using the formula yi = (1/P) (ri-3/8)/(*n* + 1/4), and after normalization the data were evaluated by using PROC GLM (general linear model) procedure for main and interaction effects. Each response was grouped by two categorical factors, namely context type, and chocolate type, and ranks were assigned by PROC RANK in SAS. In a nutshell, the chocolate type effect was statistically significant, whereas the environment effect was not significant at 95% confidence level. Furthermore, the interaction between chocolate types and environments did not significantly affect the liking scores of the evaluated sensory attributes at

95% confidence. Interestingly, the cocoa flavor was affected by the environment effect at about 88% confidence level. *Foods* **2020**, *9*, x FOR PEER REVIEW 6 of 19

Consumers that tasted the different chocolate types showed a significant difference in the liking of sensory attributes and thus, accepting the alternate hypothesis that all treatments groups do not belong to the same population median. Nemenyi test, similar to Tukey-HSD (honestly significant difference), helped identify the treatments responsible for the significance of the chocolate effect by comparing the medians of distribution (Figure 2). In general, milk chocolate had the highest liking scores of the evaluated sensory attributes, followed by white chocolate. However, the two types of chocolate products (milk and white) were not significantly different (sharing the same color, brown line) regarding the liking scores of most attributes except for cocoa flavor (Figure 2d). Milk chocolate under PVR and white chocolate under B scored 6.51 ± 0.20 and 5.06 ± 0.24, respectively, for the liking of cocoa flavor, which were significantly different from each other. Dark chocolate was the least liked chocolate type (aqua color line), and its liking scores of taste/flavor, dairy flavor, texture, smoothness, aftertaste, and overall liking were significantly different from other two chocolate types regardless of the environments. Considering the effect of environments, there were no significant differences among PVR, NVR, and B within the same type of chocolate product (*p* > 0.05). The liking scores of the evaluated attributes under PVR were similar and generally high (but not significant, *p* > 0.05) for both milk chocolate and white chocolate, whereas generally high (but not significant, *p* > 0.05) liking scores of evaluated attributes were obtained under NVR for dark chocolate. Consumers that tasted the different chocolate types showed a significant difference in the liking of sensory attributes and thus, accepting the alternate hypothesis that all treatments groups do not belong to the same population median. Nemenyi test, similar to Tukey-HSD (honestly significant difference), helped identify the treatments responsible for the significance of the chocolate effect by comparing the medians of distribution (Figure 2). In general, milk chocolate had the highest liking scores of the evaluated sensory attributes, followed by white chocolate. However, the two types of chocolate products (milk and white) were not significantly different (sharing the same color, brown line) regarding the liking scores of most attributes except for cocoa flavor (Figure 2d). Milk chocolate under PVR and white chocolate under B scored 6.51 ± 0.20 and 5.06 ± 0.24, respectively, for the liking of cocoa flavor, which were significantly different from each other. Dark chocolate was the least liked chocolate type (aqua color line), and its liking scores of taste/flavor, dairy flavor, texture, smoothness, aftertaste, and overall liking were significantly different from other two chocolate types regardless of the environments. Considering the effect of environments, there were no significant differences among PVR, NVR, and B within the same type of chocolate product (*p* > 0.05). The liking scores of the evaluated attributes under PVR were similar and generally high (but not significant, *p* > 0.05) for both milk chocolate and white chocolate, whereas generally high (but not significant, *p* > 0.05) liking scores of evaluated attributes were obtained under NVR for dark chocolate.

**Figure 2.** *Cont.*

*Foods* **2020**, *9*, x FOR PEER REVIEW 7 of 19

**Figure 2.** Nemenyi test (post hoc) results comparing medians at 5% significance level for sensory attributes, (**a**) taste, (**b**) sweetness, (**c**) bitterness, (**d**) cocoa flavor, (**e**) dairy flavor, (**f**) texture, (**g**) hardness, (**h**) smoothness, (**i**) aftertaste, and (**j**) overall. Treatments sharing color implies not **Figure 2.** Nemenyi test (post hoc) results comparing medians at 5% significance level for sensory attributes, (**a**) taste, (**b**) sweetness, (**c**) bitterness, (**d**) cocoa flavor, (**e**) dairy flavor, (**f**) texture, (**g**) hardness, (**h**) smoothness, (**i**) aftertaste, and (**j**) overall. Treatments sharing color implies not statistically different.
