Structural equation modelling

Introduction

Structural equation models (SEM) are used to test a theory or hypothesis about how multiple constructs are related to one another. A construct can be a cause of an outcome or the outcome itself. With SEM it’s possible to assess constructs that can be observed and measured directly, or constructs that cannot be observed and measured directly and are therefore ‘latent’.

For example, you can use SEM to test whether the relationship between higher body weight (observed variable) and more depressive symptoms (latent variable, assessed by several indicators of depression) is mediated by a more negative self-concept (latent variable, assessed by several indicators of self-concept).

The course will explain both the measurement part of the model (i.e. confirmatory factor analysis, linking indicators to latent variables) and the structural part of the model (also linking latent variables amongst each other). We will also discuss how SEM can be used to test hypotheses about mediation and moderation, and about change when longitudinal data are available.
 

Course information

ECTS: 2.5 
Number of sessions: 4
Hours per session: 3

Aims and working method

We will discuss both the theory and practice of SEM. Participants will learn how to fit a structural equation model to the data in the software program R (package lavaan). The meetings will consist of mini-lectures and the opportunity to practice with structural equation modelling in R using both exercise data and your own data.

Entry level

To attend this course properly participants should ideally have basic knowledge of the program R. If you do not have such knowledge yet, you can first follow our course on Data Analysis with R. If you doubt whether you have sufficient knowledge about R, please contact the lecturer, Marleen de Moor.