JURE EARLI (Junior Researchers - European Association for Research on Learning and Instruction) location:Leuven (Belgium) date:8-11 July 2008
Until now, the multilevel confirmatory factor analysis has rarely been applied in educational effectiveness research to construct measures for school process variables. Therefore, this study investigates the surplus value of a multilevel confimatory factor analysis, compared to the use of a standard confirmatory factor analysis plus aggregation. Both methods are used to develop school process variables, based on data from 1550 teachers within 82 schools or 113 cycle groups. Differences between teachers from the first cycle and teachers from the second and the third cycle are taken into account by means of a multiple group analysis if necessary. We evaluate the interpretation of the measures as well as their internal consistency. Interestingly, both methods yield equivalent results at the teacher level, thus irrespective of taking the nesting within cycles into account. Moreover, with regard to the social and orderly school climate, identical measures are also constructed at the cycle group level. On the contrary, regarding organizational aspects like the cooperation between teachers and the decision making process at the school, more global measures are detected at the cycle group level in comparison with the teacher level. Noteworthy, concerning all constructed measures, we observe considerably higher internal consistencies at the cycle group level. Therefore, our findings support the use of a multilevel factor analysis when confronted with nested data.