*2.5. Statistical Analysis*

A 3 × 3 factorial design was used in this research, which referred to three chocolate products and three contextual settings. The acceptability data for sensory attributes were tested for normality by the Shapiro–Wilk test procedure and all attributes that were found violating the normal distribution were transformed through the PROC RANK procedure of Statistical Analysis System-SAS (SAS, Cary, NC, USA) for the subsequent Friedman's test (RStudio, Version 1.1.456—© 2009-2018 RStudio, Inc.). To assess whether the reported sensory differences among chocolate type × environment were significant or not, we used the non-parametric Friedman and post-hoc Nemenyi tests. The Friedman test first establishes whether at least one of the treatments is significantly different from the rest, and if this was the case, we used the Nemenyi test to identify those treatments for which there is evidence of statistically significant differences. Nememyi test can be used [23,24], provided that data has an equal sample size between each group and Friedman-type ranking of the data. The advantage of this testing approach is that it does not impose any distributional assumptions and does not require multiple pairwise testing between treatments, which would distort the outcome of the tests. The package *tsutils* for RStudio implications were used to run the Nemenyi test (R package version 0.9.2.) [25]. Penalty analysis was applied to the JAR data in order to determine how much the overall liking and acceptance of the chocolate samples were influenced by their attributes. The effect of contextual settings was also considered in this research. The CATA emotional responses of chocolate samples under different settings were assessed by the Cochran's Q test, Correspondence analysis (CA), and principal coordinate analysis (PCoA) [22,26]. With regard to the purchase intent, it was analyzed for multiple comparisons based on Cochran's Q test and the simultaneous confidence intervals testing as well. Principal component analysis (PCA; Correlation Biplot) was used to analyze the relationship between the hedonic acceptability of ten attributes and chocolate samples under different environmental settings. The PCA results are presented as a product-attribute biplot. Chocolate samples under different settings were categorized by the Agglomerative Hierarchical Cluster (AHC) analysis. The dissimilarity of them was analyzed based on the Euclidean distance and the Ward's method. The responses given through an electronic questionnaire were collected by RedJade Sensory Software (Martinez, CA, USA). Data were analyzed using Minitab 18 (Minitab, LLC, State College, PA, USA) and XLSTAT Statistical Software 2016 (Addinsoft, New York, NY, USA).
