Annual meeting of the American Educational Research Association location:New York date:24-28 March 2008
This paper investigates the surplus value of a multilevel exploratory factor analysis, compared to the use of a standard exploratory factor analysis plus aggregation. Until now, this alternative multilevel technique has never been used in educational effectiveness research. Both methods were used to construct school process variables, based on data from 1550 teachers within 113 cycle groups. Our results show broader and, to some extent, also distinct factors at the group level in comparison with the teacher level, irrespective of taking the dependency into account. Furthermore, while the total amount of variance explained is not larger at the group level, we observe higher internal consistencies of the group-level factors. Thus, these findings clearly support the use of a multilevel analysis.