Measurement and Statistical Issues in Multilevel Research on Schooling

Measurement and statistical problems in the analysis of multilevel data should be approached through an identification of the appropriate set of substantive research questions at and within various levels and the specification of appropriate models for analyzing multilevel data. Simply selecting a unit of analysis and a level of analysis does not adequately address such problems as contextual or group effects, the selection of an aggregate statistic to represent group effects, and adoption of a statistical procedure that will consider effects at all pertinent levels. A further concern is that different multilevel approaches in analysis can change the order of importance of constructs and alter the total variation accounted for.