Multilevel models for repeated binary outcomes: attitudes and voting over the electoral cycle

Yang, M., Goldstein, H., Heath, A.
Journal of the Royal Statistical Society: Series A, 48:2, 253-268

Hampshire LEA has carried out two longitudinal studies: from reception intake to the end of Year 2, and from the end of Year 2 to the end of Year 6. A total, respectively, of 161 and 114 schools and about 6400 and 4700 pupil records have been analysed. Test scores at baseline (entry to reception), Key Stage 1 and Key Stage 2 have been used, together with pupil and school level variables. Multilevel models have been fitted and show that schools differ along several dimensions, both by curriculum subject and by prior attainment of pupil. The research sought ways of communicating the results to head teachers in ways which were meaningful without destroying the underlying complexity of the relationships uncovered. The paper describes how this can be done, in ways which can assist the process of school improvement.

Number of levels
Model data structure
Response types
Multivariate response model?
Longitudinal data?
Substantive discipline

Key paper that shows how properly specified multilevel models can be used for educational accountability. Often sited in this literature.

Paper submitted by
Harvey Goldstein, Graduate School of Education, University of Bristol,
Edit this page