England was one of 59 countries participating in TIMSS 2007 (Trends in International Mathematics and Science Survey). In addition to the international reports of the survey (Martin et al, 2008; Mullis et al, 2008), England produced its own national report (Sturman et al, 2008). This summarised key findings from the international reports and also reported on additional national analysis. The aims of the national report were to communicate TIMSS findings in England, with a view not only to informing policy and research, but also to supporting improvements in teaching and learning. One key area of national analysis included in the report was ‘multi-level modelling’ outcomes, which identified factors significantly associated with attainment, both positively and negatively (that is, factors that were associated with higher attainment, and factors associated with lower attainment). The data in the international TIMSS reports provides some indications of relationships between attainment and other variables (for example, showing that the higher the value that students place on their learning, the more highly they attain). The advantage of multi-level modelling is that it considers all selected variables and can identify factors that are significant over and above others. Whilst the modelling cannot say whether the variable caused the level of attainment or whether the reverse is true, it can provide a useful starting point for considering factors that might have causal influence. This can then provide the impetus for conversations about how to improve teaching and learning. In TIMSS 2007 in England, attainment models were run for each subject (mathematics and science) and each age group (9-10 year olds and 13-14 year olds) and easily readable charts were produced showing the significant variables. The process began with the identification of potentially relevant background variables, with ‘factor analysis’ conducted where appropriate. Most variables were derived from the TIMSS questionnaires completed by students, teachers and headteachers. Others were sourced from England’s National Pupil Database (NPD), which includes variables such as postcode and eligibility for free schools meals (both acting as proxies for socio-economic status), prior attainment, and so on. The multi-level models resulted in several factors being identified as significantly associated with attainment, and these findings were presented in graphic form in order to be accessible to readers with little or no statistical background. - --continues in Methodology---

--- This presentation will describe how the variables in the model were derived and will outline the factors identified as significant in each of the four attainment models. Information will be provided on how to interpret the findings (as well as the dangers of over-interpretation) and the potential usefulness of the findings for classroom practice will be explored. The presentation will be pitched at a level requiring little or no technical expertise, but a statistician will be available should any of those attending wish to know more about the technical aspects of the analysis.