Compositional Data Analysis in E-Tourism Research

Why some call the “worst” what most consider the “best”?: an analysis of tourist complaints at the wonders of the world

“…Once contents were categorised, the sum per topic complaints was computed to ultimately obtain the number of reviews referring to each of the topics per attraction out of all the low rating reviews. Statistically speaking, the data we processed in this study is count data (phenomenon count), a common type of data in compositional data analysis literature (Coenders and Ferrer-Rosell, 2020) and, thus, this method is considered as an important and relevant complementary tool for content analysis (Ferrer-Rosell et al, 2022) which allows observing attractions' patterns. Since the focus is on the relative importance, and the aim is to find differences about the most common complaint topics between attractions, this proportionality must be considered, and CoDA serves for that.…”

“…It jointly represents the proportionality between components (complaint topics) and the compositions (attractions). Vectors that are far from one another indicate that, if the proportion of a component increases in a given composition, the proportion of the other component decreases (Ferrer-Rosell et al, 2022). In sum, a CoDa biplot enables both the visualisation of the approximate importance of each component (topic) for each composition (attraction) in relative terms, and the identification of topics that help to make an attraction stand out from others.…”

Complex thinking and social entrepreneurship. An approach from the methodology of compositional data analysis

“…Ignoring the compositional nature of our data and using them in raw form with standard statistical tools could cause out-of-sample-space predictions, sub-compositional inconsistency, and spurious correlations [ 10 , [37] , [38] , [39] , [40] , [41] , [42] ]. Other implications can be found in Aitchison [ 43 , 44 ], Chayes [ 45 ], and Pearson [ 46 ], among others.…”