https://repositorio.ufba.br/handle/ri/34119 2026-05-03T10:40:13Z https://repositorio.ufba.br/handle/ri/36110 Título: Análise do tempo de permanência em cursos da UFBA: uma aplicação de modelagem com tempos discretos Autor(es): Góes, Jéssica Fagundes Primeiro Orientador: Brito, Edleide de Abstract: This work brought the analysis of censup data, the largest source of public data on higher education in Brazil, as a pillar in the identification of factors associated with the length of stay of students of the Statistics and Mathematics courses at UFBA. The proposal consisted of incorporating a study based on public supply data, which until then was not verified in the work done on the subject in the institution. Thus, it was possible to aggregate important information about this context from the perspective of these courses. Through the application of the methodology of competitive events and using covariates provided by CENSUP, the study was able to identify associated factors that are related to the length of stay of these students at UFBA. Editora / Evento / Instituição: Universidade Federal da Bahia Tipo: Trabalho de Conclusão de Curso 2022-08-19T00:00:00Z https://repositorio.ufba.br/handle/ri/36107 Título: Análise de fatores associados ao desempenho dos participantes do ENEM 2018 utilizando o sparklyr Autor(es): Almeida, Rafael Toledo Costa de Primeiro Orientador: Brito, Edleide de Abstract: The National High School Exam (ENEM) was created in 1988 with the purpose of evaluating the performance of high school graduates. The ENEM has become the largest Brazilian national exam using innovations in the registration process, in which it enabled a growing demand for registrations and the attendance of new test sites with realization throughout Brazil. This paper seeks to present the factors associated with the performance of all participants in ENEM 2018 through multiple correspondence analysis, presenting new ways to apply these methods together with the Spark framework. The database consists of 137 variables distributed in different aspects, such as educational aspects and socioeconomic factors of those enrolled. By preprocessing the data, using the Spark framework and the spaklyr package in R, control characteristics of the participant, school, tests and socioeconomic questionnaire were selected. The use of Spark Apache was necessary in the data preprocessing (Big Data) step. Correspondence analysis is one of the multivariate statistical techniques used for the study of categorical data that makes it possible to graphically evaluate the associations of these variables. The results of the descriptive and exploratory analysis of the data show that the participants from the states of the South and Southeast regions presented better scores compared to the participants from other states and regions. There is a contrast in education and consequently in the performance of participants in states that are farther from the economic center of Brazil. The results found in the analysis of multiple correspondence indicate that there is an association between the control variables of the participant, school, socioeconomic questionnaire and performance in objective and discursive tests. This relationship between socioeconomic aspects and the performance of enem 2018 participants is a situation that evidences a social, economic and cultural inequality that Brazil still faces and that influences other aspects, one of which is education. Editora / Evento / Instituição: Universidade Federal da Bahia Tipo: Trabalho de Conclusão de Curso 2021-06-14T00:00:00Z https://repositorio.ufba.br/handle/ri/36106 Título: Propriedades da distribuição Alpha Power Beta Weibull modificada Autor(es): Borges, Matheus dos Santos Primeiro Orientador: Brito, Edleide de Abstract: A new class of continuous distributions, a "Modified Alpha Power Beta Weibull Family", is introduced. This new class extends the modified Weibull beta distribution. Some of its mathematical properties are studied and two special cases are described. The distribution parameters are estimated by maximum likelihood and some of their inferential characteristics are investigated. Editora / Evento / Instituição: Universidade Federal da Bahia Tipo: Trabalho de Conclusão de Curso 2021-12-15T00:00:00Z https://repositorio.ufba.br/handle/ri/18001 Título: Modelos Aditivos Generalizados para Posição, Escala e Forma (GAMLSS) na modelagem paramétrica da taxa de congestionamento na fase de conhecimento Autor(es): Oliveira, Caio Batalha Dias; Carrasco, Jalmar Manuel Farfan Abstract: O presente trabalho tem como objetivo principal estudar a família dos Modelos Aditivos Generalizados para Posição, Escala e Forma (GAMLSS) no âmbito paramétrico. Esta classe de modelos foi proposta por Rigby R.A. e Stasinopoulos (2005), com o proposito de aumentar o leque de distribuições de probabilidade para a variável resposta, bem como permitir o uso de termos paramétricos e não paramétricos para as variáveis explicativas. Este trabalho também objetiva aplicar a família dos modelos GAMLSS paramétrico a um conjunto de dados. Este por sua vez, foi fornecido pelo Tribunal Regional do Trabalho da 5a região, que engloba o estado da Bahia. Uma aracterística deste conjunto de dados é que a variável resposta, taxa de congestionamento na fase de conhecimento é limitada no intervalo contínuo (0,1). Desta forma, uma distribuição de probabilidade que tem sido relevante para este tipo variáveis é a distribuição Beta. Técnicas bootstrap não paramétrico são utilizadas com o intuito de corrigir o viés dos estimadores. Encontramos também, diferentes tipos de intervalos de confiança bootstrap. Tipo: Trabalho de Conclusão de Curso 2015-07-28T00:00:00Z