Análise causal exploratória aplicada aos dados do Enem 2021: Revelando a estrutura causal dos fatores explicativos da nota final
Loading...
Date
Authors
Journal Title
Journal ISSN
Volume Title
Publisher
Universidade Federal de Juiz de Fora (UFJF)
DOI
Abstract
This work aims to present the causal relation between socioeconomic variables and the mean grade on the Exame Nacional do Ensino Médio (Enem). The Enem was created in 1998 as a tool to measure the performance of students after concluding basic education, and has been used as an entry method for higher education since 2009. Previous studies showed that family income and the parents' education level are influential on the performance of the student. As for the investigation on the relation between variables, the Exploratory Causal Analysis (ECA), or causal discovery, is the process of inferring a causal model, that is, the cause-effect relation of the variables on a dataset, a complement for Causal inference. Causal Inference, on the other hand, aims to identify the cause level of a variable as the effect on another. The cause-effect relationship is presented as Directed Acyclic Graphs (DAGs), where those relations are represented as arrows, pointing the cause-effect direction. Using constraint-based and score-based algorithms for causal discovery, applied to the Enem 2021 dataset, we proposed DAGs that identified factors which can explain the students’ average grade on Enem. After dataset manipulation, we select the variables most associated with school performance: type of school, income class, parental education, sex, color/race, access to the internet at home, location of school, and macroregion. The tests for difference between means returned significant differences between the groups of the selected variables, but testing conditional independence using likelihood tests for the constraint-based algorithms, sex, and location of school were not related to the other variables, and access to the internet at home being only related to the macroregion. For the three algorithms used, the models proposed direct influence of the type of school and income class on the Enem grade, the latter being influenced by, or mediating, the other variables. After removing variables sex, location of school, and access to the internet at home, the mother’s education also had a direct effect on the student’s Enem grade, along with income class and type of school, results that agree with previous studies.
Description
Citation
Collections
Endorsement
Review
Supplemented By
Referenced By
Creative Commons license
Except where otherwise noted, this item's license is described as Acesso Aberto

