Extending the Scope of Analysing Data of Iea Studies: Applying Multilevel Modelling Techniques to Analyse Timss Data

The traditional approach in analysing IEA data has been to use single-level statistical models. This paper argues that multilevel modelling offers a wider spectrum in analysing IEA data because of multi-stage sampling, since students are nested within classes, classes within schools, and schools within countries. Two illustrative examples are used to provide support to this argument. In the first example, TIMSS 1999 data concerning students’ achievement and their self-perceptions in two subjects (i.e., Mathematics and Science) were analysed using single and multi-level analyses. While single-level analysis resulted in positive correlations at the microlevel and negative correlations at the macro-level, the findings of multi-level analysis did not support such contradictory relationships. In the second example multi-level modelling techniques were used to analyse TIMSS data related to student achievement in Mathematics. A large number of variables taken from the student, teacher and school questionnaires and which were associated with the theoretical models of Educational Effectiveness Research were added to the model, explaining about 40% of the variance of student achievement. The results of the study are discussed in terms of the importance of using multi-level modelling in analysing IEA data. Methodological issues concerning the design of comparative studies in education are also raised.