Structural Equation Models (SEM)


  1. Multilevel SEM
  2. SEM with R and lavaan

Multilevel SEM

Together with Bruno Castanho Silva and Levente Littvay, we wrote in 2019 a short Sage book introducing Multilevel Structural Equation Models to a social science audience. Though the models presented in the book are estimated using Mplus 8, where possible we also supply lavaan code, along with all TikZ code for generating the path diagrams we use.

All replication data and code are available here.

SEM with R and lavaan

This 3-day workshop in 2021 at the University of Bamberg introduced participants to the structural equation modeling (SEM) framework, and to how a set of basic but flexible SEM specifications can be estimated with the help of the R statistical environment and the lavaan package.

In the 3 days devoted to this topic, we covered the fundamentals of observed-variable (path) models, confirmatory factor analysis (CFA) models, and full structural regression models. During this progression to ever-more complex specifications we touched upon aspects of graphical presentation of such models, estimation, local and global fit testing, as well as sample size considerations. In the final day of the workshop, we discussed how the SEM and the MLM frameworks can be joined to understand how latent traits impact outcomes differentially across different contexts.

All materials (data sets, slides, and code files) are made available here.