Multilevel Structural Equation Modeling
In person in Calgary | July 21-22, 2023, during Rocky Mountain Methodology Academy
$849 (Canadian dollars) + 5% GST (10% discount for graduate students)
Multilevel structural equation modeling (MSEM) is a general modeling framework that borrows features from multilevel modeling (MLM) and structural equation modeling (SEM), and contains both as special cases. MSEM is becoming more popular as a general approach for modeling clustered (nested) data—for instance, students nested within classrooms, patients within classrooms, voters within districts, or repeated measures within individuals—because it overcomes common limitations of both MLM and SEM. For example, it is difficult to model latent variables or multiple outcomes with traditional MLM. Similarly, it is difficult to model clustered data or include level-specific predictors with traditional SEM. MSEM provides an intuitive, flexible framework for combining all the best abilities of MLM and SEM into a single modeling approach. Participants in this course will build on their foundational knowledge of either SEM or MLM and have opportunities to ask questions specific to their own projects. This in-person workshop emphasizes practical application of MSEM methods that are applicable to data from the contexts of education and psychology, medicine, business, and allied fields.
INSTRUCTOR: Kristopher Preacher, PhD
In this course, you will learn about the underlying principles and the practical applications of multilevel structural equation modeling. The topics to be covered include:
- Review of MLM and SEM
- Multivariate MLM
- Introduction to multilevel SEM
- MSEM equations and path diagrams
- Orientation to Mplus for MSEM
- Multilevel path analysis
- Multilevel confirmatory and exploratory factor analysis
- Multilevel SEM with latent variables
- Three-level models in MLM vs. MSEM
- Multilevel reliability estimation
- Model fit in MSEM
- Mediation in MSEM
- Multiple group multilevel models
- Moderation (interaction effects) in MSEM
- MSEM with discrete dependent variables
- Power analysis for MLM and MSEM
This course is delivered in person and will take place over two days at the University of Calgary during Rocky Mountain Methodology Academy in July 2023
This workshop not only will cover the statistical and mathematical basis for MSEM, but also will emphasize practical application of this family of models using Mplus. Although access to Mplus is not necessary to benefit from the workshop, it is strongly encouraged to bring a laptop to class with a recent installation of Mplus with either the multilevel or combination add-on. A large library of example Mplus code will be made available to participants.
Multilevel SEM is useful for testing simple and advanced models using nested data in a variety of fields—including psychology, education, epidemiology, organizational research, etc. This workshop will be particularly useful for students and researchers already somewhat familiar with either multilevel modeling or structural equation modeling, and who wish to expand their repertoire of modeling options. No particular software proficiency will be assumed. An introduction to the relevant features of Mplus will be included in the course materials.
Upon completing this course, you will
- master some advanced topics in MLM and SEM
- gain a deeper understanding of the relationships between SEM and MLM
- be able to use MSEM to test complex hypotheses
- understand when it is appropriate to use MLM, SEM, and MSEM
- know how to make informed choices when specifying and evaluating a model
- understand and be able to fluently translate among path diagrams, model equations, and Mplus code for MLM, SEM, and MSEM models
A certificate of completion from the Canadian Centre for Research Analysis and Methods is provided at the end of the course.
Dr. Preacher is the Lois Autrey Betts Chair in Education & Human Development in the Quantitative Methods program within Vanderbilt University’s Department of Psychology & Human Development. His research concerns the use (and combination) of structural equation modeling and multilevel modeling to model correlational and longitudinal data. Other interests include developing techniques to test mediation and moderation hypotheses, bridging the gap between substantive theory and statistical practice, and studying model evaluation and model selection in the application of multivariate methods to social science questions. He serves on the editorial boards of Multivariate Behavioral Research, Behavior Research Methods, Communication Methods and Measures, and Journal of Educational & Behavioral Statistics.