Modelagem de batalhas no Instagram via distribuição Beta e Inferência Bootstrap

Abstract

Competitions, contests and championships are commonly held in the most diverse areas - sports, beauty, skills, music, etc. - to compare and rank competitors. Particularly, some championships follow a key scheme, which can be single, double,..., to evolve the competitors until choosing the winner. In these cases, the confrontation takes place in pairs, that is, when two opponents face each other, only one emerges victorious. When it is not possible to have draws or repechage, these clashes are called “knockout”. At the end of championships of this type, the result is the grouping of opponents into classes: first place, second, third, fourth, fifth, etc. However, as each confrontation happens only once, we have a sample size 1 of a conceptual population of infinite repetitions of that combat. Therefore, the groupings resulting from such championships have an associated uncertainty, which is normally ignored. It would be useful to model the uncertainty involved in a championship, in order to know which winners can be considered truly superior and which won by mere chance. To answer this question, this work proposes a methodological approach that uses the Beta distribution and parametric Bootstrap inference. As examples, championships made on Instagram stories with popular cheeses in Brazil and country singers are used. In this way, the proposed methodology proved to be efficient in modeling the uncertainty associated with groups promoted by switching championships, revealing that the cheeses that faced each other cannot be said to be statistically different and that Mar´ılia Mendon¸ca, Henrique and Juliano, Jorge and Mateus and Z´e Neto and Cristiano are the country singers preferred by Brazilians.