Structural equation modeling in longitudinal research
calendar_month 29 Iul 2015, 00:00
Structural equation modeling (SEM) has become one of the most widely used techniques for analyzing longitudinal data in the social sciences. The advantage of using SEM (compared to traditional approaches) is that we can exploit the full power of SEM in the longitudinal setting: we can use a comprehensive measurement model for the (latent) constructs in our model, and test for measurement invariance over time; we can use mediational relationships (and possibly reciprocal relations) in addition to regression; we can use multiple groups; we can include time varying covariates, and we can use arbitrarily complex error structures to capture the dependency structure of the data. And last but not least, we get feedback (goodness-of-fit measures) about the adequacy of our model, so we are not maneuvering in the dark. The price for all this modeling power is that it may be hard to see the forest for the trees. If you need a comprehensive overview of both basic and advanced statistical techniques to analyze longitudinal data in a structural equation modeling framework, this course is for you. The course will start with a short refresher of the theory and practice of structural equation modeling, including a tutorial on how to use the (freely available) R package lavaan. We then proceed with an overview of several approaches to analyze longitudinal data. First, we will briefly discuss some traditional and non-SEM based approaches (the paired t-test, repeated measures ANOVA, MANOVA, the linear mixed model). Next, we will introduce the SEM approach, and illustrate how the traditional approaches (including the linear mixed model) can be seen as special cases of the SEM framework (if we have the same number of time points for each observation). We will discuss the SEM approach to analyze (observed or latent) repeated measures, the relationship with autoregressive and related panel models, the growth curve model, latent difference score models, the autoregressive latent trajectory model, and some other longitudinal SEM models that have appeared recently in the literature. Throughout the course, a non-technical hands-on approach will be used, and all analyses will be illustrated with the R package lavaan. Some basic knowledge of SEM is required. Some prior knowledge of R is recommended, but not required.

Course leader
Yves Rosseel, Ghent University, Belgium

Target group
Students, researchers, business professionals who need to develop good quality surveys and/or need to apply appropriate and up-to-date statistical methods.

Fee info
EUR 150: Student rate EUR 300: Academic rate (teachers & researchers)

Universitat Pompeu Fabra
Address: Department of Political and Social Sciences, Carrer Ramon Trias Fargas, 25-27 Campus Ciutadella
Postal code: 08005
City: Barcelona
Country: Spain
Phone: +3493 542 20 00