Multilevel Measurement Models For Group Collective Constructs

When studying groups and teams, researchers can include individual-level and/or group-level phenomena in their research design. Distinguishing between individual-and group-level constructs is necessary before choosing statistical analysis and interpretation. This article addresses multilevel measurement models for data collected from individuals within small discussion groups. Measurement models for this class of variables present several challenges because the data are nonindependent. We propose a conceptual framework for establishing a set of models a priori based on the notion of collective constructs and describe means for testing structures at both levels of analysis. Using two data sets, we discuss the conceptual and statistical problems associated with developing measurement models for group data collected at the individual level.