Chapter 13: Multilevel Structural Equation Modeling with Complex Sample Data

As researchers become aware of the wealth of large-scale datasets available at the national and international level (within the United States, for example, data are collected by the National Center for Education Statistics, the National Science Foundation, and the Center for Disease Control), questions of how to analyze such data are becoming more frequent. The choice of analysis depends first and foremost on the research question being asked. Once decided, the researcher typically needs to utilize special techniques in order to estimate model parameters and standard errors appropriately because the data were not collected using a simple random sampling procedure. In this chapter, I introduce the types of complex sampling designs that are used in national and international studies and illustrate the structural equation modeling (SEM) analysis tools that are available to researchers using data collected with such designs. In particular, because complex sampling designs often call for collecting data in multiple stages, for example, students nested within schools, the focus will be on analyzing multilevel models. But researchers should be aware that multilevel models answer specific research questions and are not always necessary when data are nested. Multilevel models allow the researcher to examine relations among variables within a nested structure (such as patients within hospitals) as well as relations at a group level (in this case, hospitals). For other research questions, a single-level modeling approach may be appropriate and strategies for such modeling will be briefly discussed.