Structural Equation Modeling for Longitudinal and Panel Data

About

The course will expand on the course “Introduction to Structural Equation Modeling: Confirmatory Factor Analysis with Mplus” and show how to apply the SEM approach to longitudinal data using the Mplus computer program. In the first part of the course, we will introduce the autoregressive model (ARM) and the cross-lagged panel model (CLPM) to study the stability of a single variable over time as well as reciprocal effects and 'causal' predominance between two variables over time. Each type of model will be discussed as a single-indicator as well as a multiple-indicator model. All models will be applied to data from a longitudinal study on authoritarianism and anomia in Germany. In the second part of the course we will focus on modeling development and change over time with the latent growth model (LGM) applying the same dataset. We will begin with the univariate case to analyze growth of a single variable and extend the basic model to the multivariate case to analyze parallel processes of two variables over time. Finally, two 'hybrid-models' will be discussed: the autoregressive latent trajectory model (ALT) and the more recently developed random-intercept cross-lagged panel model (RI-CLPM). Topics in both parts include parameterization of autocorrelations, Socratic effects, latent means, and MIMIC models. Furthermore, we will extend our discussion to multiple group comparisons as well as the issue of (longitudinal) measurement invariance as a prerequisite for comparisons across groups and over time.

Course leader

Dr. Daniel Seddig is a research assistant at the Institute of Sociology and Social Psychology, University of Cologne, statistical consultant at the Institute of Psychology, University of Zurich, and chair of the “European Working Group on Quantitative Met

Target group

Participants will find the course useful if they are interested in:
- assessing and explaining change over time,
- the relationship between processes of change in different variables,
- assessing the degree of reciprocity in the relationship between variables over time,
- causal predominance of one variable over another,
- conducting meaningful comparisons of (latent) variables over time, (or across groups).

Prerequisites
Basic knowledge of and basic experience with confirmatory factor analysis and structural equation modeling. This could be acquired in the course “Introduction to Structural Equation Modeling: Confirmatory Factor Analysis with Mplus“ in week 1.
We will use the software package Mplus. We will briefly introduce Mplus during the first exercise. Mplus will also be used in the course “Introduction to Structural Equation Modeling“ in week 1. The short course “Introduction to Data Analysis Using Mplus” in week 0 will explicitly focus on how to use Mplus effectively.

Course aim

By the end of the course participants will:
- know how to specify autoregressive and cross-lagged structural equation models to test for stability and reciprocity in manifest and latent variables over time as predicted by a social scientific theory,
- know how to test for measurement invariance of latent variables across time (and groups),
- know how to examine various forms of change with different specifications of the latent growth model,
- be able to specify all models with the software package Mplus;
- (if time allows), learn to separate interpersonal from intrapersonal variance in the variables of interest and run a random-intercept cross-lagged panel model test causal predominance as predicted by a social scientific theory.

Fee info