Multilevel Analysis

Day 1 will provide a gentle, non-technical introduction to GLMMs by means of a data example of clustered hierarchical data involving a binary outcome, where a single-level logistic regression would ignore the multilevel nature of the observations. Analogously to the basic module, we will look into estimation methods to estimate fixed effects and distributional parameters related to the random effects distribution, as well as quantify the degree of clustering by means of the intraclass correlation coefficient. Emphasizing application and interpretation rather than the underlying mathematical underpinning, we will discover how to adequately draw conclusions about the population based on these models and discuss similarities and differences with Linear Mixed Models as a special case of GLMMs. Afterwards, GLMMs for count and ordinal outcomes will be discussed. 

Day 2 will introduce Generalized Estimating Equations (GEEs) as an alternative approach for analysing correlated longitudinal and clustered data. Starting again from practical examples, we will explore how GEEs account for the association between repeated or clustered observations through the specification of a working correlation structure. Particular attention will be paid to the interpretation of population-averaged effects and how these differ from the subject-specific effect estimates obtained using GLMMs. We will discuss the role of robust standard errors, the choice of working correlation structure, and practical considerations when fitting and evaluating GEE models. Applications will include binary and count outcomes, with emphasis on model specification, interpretation of regression parameters and appropriate statistical inference. Throughout the day, GEEs and GLMMs will be compared to highlight when the two approaches answer different research questions and to provide practical guidance on choosing an appropriate modelling strategy for longitudinal and clustered data. 

This advanced course is intended for researchers, data analysts, and professionals who already have experience with multilevel analysis. Participants should be familiar with the principles of mixed-effects models for continuous outcomes and have prior experience performing these analyses in R.In addition, familiarity with generalized linear models is expected.

The course is suitable for participants who have previously completed the Introduction to Multilevel Analysis course or who have an equivalent level of knowledge and practical experience. Familiarity with fitting, interpreting, and evaluating linear mixed-effects models in R is assumed. For background knowledge on generalized linear models, our Categorical Data Analysis course suffices.

As this is an advanced course, introductory concepts such as the rationale for multilevel modelling, model specification, and the interpretation of basic mixed-effects models will not be covered. The focus is on extending these concepts to generalized linear mixed models (GLMMs) and generalized estimating equations (GEEs) for the analysis of non-continuous outcomes.

December 17-18, 2026 from 9.30 until 15.00 at the latest, with a one hour lunch break around 12.00

This course is fully ONLINE

Prof. dr. Steven Abrams and Jesse Berwouts

For this course, we offer the possibility to take an exam. 

For the PhD students in the faculties IOB and Applied Economics, this is a requirement to obtain a credits for these courses, but people from other faculties are allowed as well. 

If you are interested in taking the exam, check the wants-to-take-exam-box in the registration form. 

Participating in the exam costs 10€, which is deduced automatically from your educational credit.