engleski

Doubly Latent Multilevel Structural Equation Modeling: An Overview of Main Concepts and Empirical Illustration

Many previous studies that investigated the association between classroom climate variables and student outcomes have suffered from methodological flaws, such as relying exclusively on self-reports, failing to consider the appropriate level of statistical analysis, and inadequately controlling for potential measurement and sampling errors. Such analytical strategies typically lead to confounding the true effects at student and classroom levels as well as to biased estimates. The present chapter provides an overview of main concepts of doubly latent multilevel structural equation modeling (DL-MSEM) that enables testing theoretically relevant relationships at proper level of analysis (i.e., class, teacher, school) and controlling for measurement (by using multiple indicators for latent variables at student and teacher levels) and sampling errors (by incorporating the scores for different students in the same class as multiple indicators of latent variables at teacher level). In addition, an empirical illustration of data analysis with the DL-MSEM is provided by using data based on multilevel design (i.e., students nested within teachers) and drawn from multiple sources (i.e., teachers’ and students’ reports). More specifically, to assess the climate effects, each student within a class directly rated the instructional behavior of their teacher, thus making a teacher (rather than a student) the referent. In addition, teacher self-reports of their personal characteristic were combined with student reports of their teachers’ instructional behavior and student self-reports of their motivational processes. The results were interpreted in relation to the main concepts of the DL-MLSEM method.

doubly latent multilevel structural equation modeling ; measurement and sampling errors ; multilevel mediation ; climate vs. contextual effects

nije evidentirano