Project CEMAPRE internal
|Title||Statistical modelling in Education: multilevel models applied to large and complex data|
|Participants||Maria Eugénia Ferrão (Principal Investigator)|
|Summary||For the past 30 years multilevel modelling has become the mainstream statistical approach in|
educational research, recognizing the fact that schools differ in terms of their pupils'
achievements. Such statistical models explicitly incorporate the hierarchical structure of the
population under analysis. In this project we carry out both cross-section and longitudinal data
analyses. Specifically, data modelling includes PISA 2012 (OECD), Geres 2005, and ENEM (INEP,
Brazil), in order to enlighten three main research objectives. First, our aim is to explore research
questions related to the variability of grade repetition among schools, and to assess the students'
educational gain over an additional year at the same level of primary education. For that purpose,
we apply random coefficient models to the longitudinal data survey conducted in five municipalities
of Brazil (Geres 2005). Secondly, we also investigate what school characteristics may be related to
student's performance by analyzing the Portuguese and Brazilian data of PISA 2012. A logistic
multilevel model is used. Finally, we want to contribute to a better understanding of the role of
schools and regional authorities in mitigating inequalities in the distribution of education in the
Brazilian preuniversity students, using the ENEM data from 2009 to 2012. Methods applied include the
Lorenz curve and the variance component model with three levels, i.e. students grouped in schools,
and schools grouped in federal units.