PASEC 2014 Data

Students' scores were estimated at the aggregate level, but also for various categories (e.g., girls versus boys, rural students versus urban students, etc.) for equity analysis purposes. These analyses:

• Distinguish the differences in scores between the modalities of a variable
• Were carried out using regression methods (rather than a simple means tests comparison), thus correcting for violations of the test implementation hypotheses (in means comparison)
• Were done in Stata Software using replicated weights.

In addition, some multi-varied analyses were undertaken to analyze student scores in relation to a wider range of variables.

• Using hierarchical linear modelling (HLM), the analysis focused on the relationship between a student’s score and various variables taken together, such as age of student, gender, socioeconomic capital of the student’s family, and conditions for learning at home and at school.
• These analyses were carried out using HLM software with weighted data.
• The multivariate analysis method involves the implementation of multilevel models.
  • Classically, a model without explanatory variables is estimated, allowing to determine the percentages of the variance of the scores that is attributable to each level (students and schools).
  • A block of variables is then added.
  • The characteristics of the students, then those of the teachers and the class, and finally those of the principal and the school are successively introduced.
  • For each of the estimated submodels, it is possible to calculate the variance reduction that results from the addition of the respective block of variables.

For all analyses, procedures for plausible values were used; 5 plausible values were drawn for each student.