Multiple Regression and Multi-Level Modelling using PISA data

Eveline Gebhardt & Alex Daraganov, ACER (Melbourne, Australia)

The course will deal with multiple regression in the morning and multi-level modelling in the afternoon. Basic theory will be dealt with very briefly, because participants are expected to be familiar with it. In the morning, some example multiple regression analyses will be carried out using the SPSS module with the BRR replication method to estimate standard errors.

The subject in the afternoon will be multi-level analysis using PISA data. The focus will be on the advantages and disadvantages of multi-level models compared to multiple regression. In addition, issues related to multi-level modelling when using PISA data will be discussed in detail. In brief, the issues will be about weighting the data (at which level and what kind of weights are best to use), computing standard errors (replication or not), which software to use (SPSS/SAS versus multi-level specific software) and interpreting results. Participants will be provided with a copy of the PISA Data Analysis Manual for SPSS users which they can retain for their own use after the training session. Participants will also be provided with computers and the required software (SPSS and HLM) for the course.

The Training Session on the 3rd day of the conference (16 September 2009) will take place at the IPN and will run from 9 to 17:00. including coffee and lunch breaks. The Training Session is aimed at researchers who have extensive experience with PISA data. Please note that we can only accept applicants who at least

1.      have experience with the PISA database (plausible values, weights and replicate weights),

2.      can type commands in the SPSS syntax editor and

3.      have basic knowledge of multiple regression and multi-level analysis.

These areas will not be explained during the workshop.

25 participants have been selected and the course is now fully booked. Please do not apply for the training session any more.