Politik i simplexrummet

Sammanfattning: In this thesis we discuss a potential problem of negative correlation arising when using ordinary statistical methods on compositional data. We discuss a less known statistical method, compositional data analysis, that stands on the pillars of the works of John Aitchison from the 1980-s and Pawlowsky-Glahn, Egozcue and Delgado from the 2015. Using results from the 2022 national parliament elections divided into 290 Swedish municipalities, we present how the method, together with multivariate regression, can be used to explore the effects of different social and economic aspects on how people tend to vote. We find compositional data analysis to be a powerful and underutilized tool for analyzing aggregate voter behavior, allowing an application of regression analysis to elections with results that are easy to visualize and present.