Name: | Description: | Size: | Format: | |
---|---|---|---|---|
4.28 MB | Adobe PDF |
Authors
Advisor(s)
Abstract(s)
A natureza, a quantidade de variáveis envolvidas no processo de ensino aprendizagem, a
forma como estas variáveis estão agrupadas: variáveis do aluno, da sala, do professor, da
escola, da família, etc. nos impõe que qualquer estudo aplicado nesta área não deve
descurar esta hierarquia organizacional, pois se tal acontecer, as conclusões advenientes
correm o risco de serem desprovidas de rigor científico.
Nesta vertente, aparecem os modelos de regressão multiníveis que estatisticamente
trazem esta mais valia que é a possibilidade de analisar dados que possuem uma
organização hierárquica, o que na educação se verifica de uma forma bem óbvia, visto
termos alunos (com caraterísticas próprias) agrupados em turmas (um outro nível
hierárquico) que por sua vez estão agrupadas em escolas, que também podem pertencer
a agrupamentos escolares. Apesar desta evidente organização hierárquica, há uma regra
estatística que nos irá permitir avançar ou não com o ajustamento dos dados a um
modelo de regressão multinível: o Coeficiente de Correlação Intraclasse. Caso não seja
possível o ajuste dos dados pelo modelo multinível devemos pesquisar outros
procedimentos estatísticos que melhor se adequem aos dados.
Neste trabalho, procuramos modelar os dados de avaliação de desempenho dos alunos
das escolas secundárias da cidade do Porto Novo – Santo Antão – Cabo Verde dos anos
letivos 2016 a 2019, tendo como variáveis dependentes as classificações finais nas
disciplinas de Matemática e Português. Os dados não suportaram estatisticamente um
modelo multinível, pelo que decidimos recorrer a regressão múltipla, considerando a
Nota Final à Matemática e à Português como variáveis respostas. O estudo foi conduzido
com o apoio do software IBM SPSS (Satistical Package for Social Science) Versão 25.
The nature, the quantity of variables involved in the learning process, the way these variables are grouped: the students, the classroom, the teacher, the school, the family, etc requires us that any study applied in this area should not neglect this organizational hierarchy, for if it happens, the conclusions are likely to be devoid of scientific rigour. In this aspect, the multilevel regression models that statistically bring this added value appear, which is the possibility of analysing data that have a hierarchical organization, which in education is verified in a very obvious way, since we have students (with their own characteristics), grouped in classes (another hierarchical level), which in turn are grouped in classes, which also belong to school groupings. Despite this clear hierarchical organization, there is a statistical rule that will allow us to proceed or not with the adjustment of data to a multilevel regression model: the Intraclass Correlation Coefficient. In case it is not possible the adjustment of the data by the multilevel model we must research other procedures that better fit the data. In this work, we try to model the performance evaluation data of the students of the secondary schools in Porto Novo City - Santo Antão - Cape Verde from the school year 2016 to 2019, having as dependent variables the final classifications of the Mathematics and Portuguese subjects. The data did not statistically support a multilevel model, so we decided to use multiple regression, considering the Final Grade to Mathematics and Portuguese as the final response variable. The study was conducted with the support of IMB SPSS (Statistical Package for Social Science) Version 25.
The nature, the quantity of variables involved in the learning process, the way these variables are grouped: the students, the classroom, the teacher, the school, the family, etc requires us that any study applied in this area should not neglect this organizational hierarchy, for if it happens, the conclusions are likely to be devoid of scientific rigour. In this aspect, the multilevel regression models that statistically bring this added value appear, which is the possibility of analysing data that have a hierarchical organization, which in education is verified in a very obvious way, since we have students (with their own characteristics), grouped in classes (another hierarchical level), which in turn are grouped in classes, which also belong to school groupings. Despite this clear hierarchical organization, there is a statistical rule that will allow us to proceed or not with the adjustment of data to a multilevel regression model: the Intraclass Correlation Coefficient. In case it is not possible the adjustment of the data by the multilevel model we must research other procedures that better fit the data. In this work, we try to model the performance evaluation data of the students of the secondary schools in Porto Novo City - Santo Antão - Cape Verde from the school year 2016 to 2019, having as dependent variables the final classifications of the Mathematics and Portuguese subjects. The data did not statistically support a multilevel model, so we decided to use multiple regression, considering the Final Grade to Mathematics and Portuguese as the final response variable. The study was conducted with the support of IMB SPSS (Statistical Package for Social Science) Version 25.
Description
Keywords
Avaliação de desempenho Desempenho escolar Modelos multiníveis ou hierárquicos Modelos de regressão múltipla School performance Multilevel or hierarchical models Multiple regression models
Pedagogical Context
Citation
Pires, Emerson Andrade - Análise multinível [Em linha]: aplicação em avaliação de desempenho escolar. [S.l.]: [s.n.], 2020. 135 p.