Characterizing School Climate Using Student-Level Measures of Attitudes, Behaviors, and Emotions: An Application of Multilevel Factor Analysis

• In Event: Multilevel Latent Variable Models: The Next Generation

Abstract

Interest in understanding how the contexts surrounding youth shape their health and development has grown considerably. School environments, namely school climate, is one of the most promising contexts studied to date. Although prevention-oriented scientists often implement interventions in school-based studies and seek to understand the role of the school context on youth outcomes, these efforts may fall short of their full potential based on the limited ways in which school climate is understood. This paper presents an illustration of an emerging analytic method called multi-level factor analysis (MLFA) that provides an alternative strategy to conceptualize, measure, and model environments, including school climate. MLFA decomposes the total sample variance-covariance matrix into within-group (i.e. student-level) and between-group (i.e. school-level) matrices and simultaneously models distinct latent factor structures at each level. Using data from 82,186 students nested in 126 schools in the National Longitudinal Study of Adolescent Health, we use MLFA to show how 22 items that capture student’s attitudes, behaviors, and emotions (e.g. feelings of safety in school; alcohol/drug use; ability to pay attention in school) provide information about both students (within-level) and their school environment (between-level). We identified five factors at the within-level: (1) difficulty meeting the demands of the school environment; (2) externalizing symptoms; (3) internalizing symptoms; (4) not belonging to the school; and (5) poor self view. Three factors were identified at the between-level: (1) hidden school social disorder, (2) culture of school disengagement, and (3) culture of school collective hopelessness. High intraclass correlation coefficients (ICC) for some items underscore the need to account for clustering. Researchers can extend the MLFA method to other nested relationships, such as youth in neighborhoods, in an effort to understand associations between environments and individual outcomes and how to best implement preventive interventions.

Authors

• Erin Dunn, Harvard University

• Katherine E. Masyn, Harvard University

• Stephanie Margaret Jones, Harvard University

• S.V. Subramanian, Harvard University

• Felton Earls, Harvard University

• Karestan Koenen, Harvard University