Revisiting fixed- and random-effects models: some considerations for policy-relevant education research
Authors:
Open Access
- 10 September 2013
- journal article
- research article
- Published by Taylor & Francis in Education Economics
- Vol. 23 (3) , 259-277
Abstract
The use of fixed (FE) and random effects (RE) in two-level hierarchical linear regression is discussed in the context of education research. We compare the robustness of FE models with the modelling flexibility and potential efficiency of those from RE models. We argue that the two should be seen as complementary approaches. We then compare both modelling approaches in our empirical examples. Results suggest a negative effect of special educational needs (SEN) status on educational attainment, with selection into SEN status largely driven by pupil level rather than school-level factors.Keywords
This publication has 17 references indexed in Scilit:
- Parental choice of primary school in England: what types of school do different types of familyreallyhave available to them?Policy Studies, 2011
- The Role of Sample Cluster Means in Multilevel ModelsMethodology, 2011
- The impact of support staff on pupils’ ‘positive approaches to learning’ and their academic progressBritish Educational Research Journal, 2011
- Multilevel Statistical Models2010
- Evaluating Kindergarten Retention PolicyJournal of the American Statistical Association, 2006
- Microeconometrics2005
- Regressor and random‐effects dependencies in multilevel modelsStatistica Neerlandica, 2004
- Inferring Program Effects for Special Populations: Does Special Education Raise Achievement for Students with Disabilities?Review of Economics and Statistics, 2002
- Methods in School Effectiveness ResearchSchool Effectiveness and School Improvement, 1997
- Statistical Modelling Issues in School Effectiveness StudiesJournal of the Royal Statistical Society. Series A (General), 1986