Demystifying Longitudinal Data Analyses Using Structural Equation Models in School Psychology

Structural equation models (SEM) are a method of latent variable analysis that offer a high degree of flexibility in terms of modeling methods for applied research questions. Recent advancements associated with longitudinal SEM have unlocked innovative ways to decompose variance and to estimate mean trends over time (e.g., Allison et al., 2017; Berry & Willoughby, 2017; Hamaker et al., 2015; McArdle & Nesselroade, 2014). However, these longitudinal methods are not necessarily readily accessible to scholars seeking to advance theory and practice in school psychology. Importantly, not all longitudinal data are the same and not all longitudinal SEMs are the same; thus, analytic approaches must be appropriately matched to specific research aims to meaningfully inform school psychology theory and practice. The present article highlights recent advances in longitudinal SEMs, clarifies their similarities to other—perhaps more familiar—methods, and matches their applications to specific types of research questions. The intent of this work is to promote careful thinking about the correspondence between estimands, developmental theory, and practical applications to foster specificity in testing quantitative questions in school psychology research and advance a more rigorous evaluation of longitudinal trends relevant to research and practice in the field.