
Semester: Winter 2026/2027
Course Number: 300250
Lecturer: Marvin Bürmann
Location: X-D2-105
Time: tuesdays 14-16h c.t.
Notice:
All important information about the schedule of the seminar, the tools to be used and the requirements for the completion of study and examination achievements can be found in this document in the corresponding sections. If necessary, the document will be adapted during the semester. The current version of the document is always available at this URL.
URL (with “http://”" or without “www.”): lehre.mbuermann.de/syllabus_rmhd2627.html
Many questions in the social sciences are best answered with data that have a hierarchical structure, i.e., data available at different levels. For example, to investigate whether differences between countries are systematically related to individual attitudes, information is needed at the individual level (attitudes) and at the country level (operationalization of country differences). Similarly, a hierarchical data structure is needed to study the consequences of life course changes at the individual level. In this case, the additional level besides the individual is the temporal level. While data in which individuals are nested in different contexts - such as individuals in countries - are usually analyzed with so-called multilevel models, panel models are usually applied for changes in individual life courses. In cases where individuals are observed in different contexts over time, hybrid forms of the two types of models are also used.
The seminar will first discuss in which cases a “normal” regression model is not the best method for answering research questions and what advantages multilevel and panel models offer in comparison. The mathematical and statistical foundations of each of these model types will be explained and practical exercises will be conducted using teaching data sets. Emphasis is placed on practical applications and the differences and similarities among the model types.
Students should acquire the competencies to analyze hierarchical data sets adequately. This includes the correct modeling of hierarchical data in the context of multilevel and panel models as well as the model construction corresponding to a specific research question (selection of analysis steps and presentation of central results).
Students should have a good understanding of linear regression analysis in order to participate adequately in the course. However, lack of prior knowledge can be compensated by self-study during the first weeks. The practical exercises will be carried out with the help of guided exercise sheets. Appropriate knowledge of the syntax in “Stata” is helpful in order to be able to participate adequately in the practical exercises. In principle, the exercises can also be completed using “R”, but the solutions for the exercise sheets will be provided and discussed as “Stata” syntax.
The Moodle course of the seminar is one central tool. It can be accessed via the normal learning room (“Moodle-Kurse” tab on the left-hand side) and via the following URL: https://moodle.uni-bielefeld.de/course/view.php?id=16472. The Moodle course provides slides & excercises sheets for the respective meetings. The files for the credits are also to be submitted here.
Data as well as the solution files will be provided in a folder on the “CIPUX_veranstaltungen” drive (usually the “K” drive). In the folder (probably “buermann_rmhd2”) there will be a “readonly” and a “public” folder. In the “readonly” folder the data and the solution files will be provided. In the “public” folder you can create your own subfolder (e.g. with the name of your access code) in which you work during the practical meetings. As subfolders for this folder I recommend: “do”, “log”, “derived” & “results”. The access to the drive is also possible via remote-access (see next section). CIP-pool account names (not the password!) are needed for on- and off-site access to the folder, please tell it in the first session or send it via mail.
It is possible to use Stata in the faculty’s CIP pools outside of class hours (current scheduling charts are posted on the doors of the CIP pools). If you would also like to use Stata from home, you must request access to “Solabor” from the IT department of the Department of Sociology (if you are enrolled in another course where you are granted access to the course drive, you will automatically be granted access to “Solabor”). Once “Solabor” has been activated and set up, you can use Stata and all other CIP pool software programs, as well as your personal drives and course drives, from home. For more information on activating and setting up “Solabor,” please visit:
https://www.uni-bielefeld.de/fakultaeten/soziologie/fakultaet/soz-it/distance-learning/
Many services of the university are only available within the university network. This includes e.g. the access to paywalled literature. To access literature from home, a VPN connection is required. Appropriate instructions can be found on the following page of the university: https://www.uni-bielefeld.de/einrichtungen/bits/services/infra/vpn/
The study credits (Studienleistung (SL)) can be obtained by submitting solution files for the exercise sheets. The number of solution files to be submitted for the study credits can be reduced by a presentation on a scientific paper analyzing hierarchical data like the ESS. Without such a presentation, at least 4 of the 6 exercise sheets must be completed. A presentation (10-15 minutes) can reduce the number of solution dofiles to be submitted to 2. The solution files to the exercise sheets have to be uploaded to the corresponding submission folder of the Moodle course at the latest on Sunday evening before the seminar (the submission folder closes after 23:59!).
The exam points (Prüfungsleistung(PL)) can be obtained by a seminar paper (Hausarbeit) on an own research question using the methods of analysis discussed in the seminar.
The data must be obtained, prepared and analyzed by the student. Information on the adequate preparation of seminar papers at the AG Kroh can be found here.
| Session | Content | Participation- points |
|---|---|---|
| 13.10.2026 Session 1 |
Organizational stuff, how to earn credits, seminar schedule, brief introduction to hierarchical data | - |
| 20.10.2026 Session 2 |
Lecture: Which hierarchy? an introduction to hierarchical data structures: different types, but similar benefits and problems |
- |
| 27.10.2026 Session 3 |
Lecture: Using full variance vs. ignoring level differences how to apply “normal” OLS models and how to estimate LSDV models for hierarchical data |
- |
| 03.11.2026 Session 4 |
Exercise: Set up and estimating OLS & LSDV models using OLS and LSDV models for hierarchical data |
Excercise 1/6 |
| 10.11.2026 Session 5 |
Lecture: Does the context affect lower-level associations? how to model different effect sizes in fixed-effects models via interaction terms and why one should be cautious with this analytical approach |
|
| 17.11.2026 Session 6 |
Exercise: Cross-level interactions in fixed-effects models application with re-introduced means |
Excercise 2/6 |
| 24.11.2026 Session 7 |
Lecture: Actively modelling level differences between contexts how variance is decomposed and explained in (random intercept) multilevel models and how to use them (and why they may not be a good choice for panel data) |
- |
| 01.12.2026 Session 8 |
Exercise: Estimating multilevel (random effects) models Using commands dedicated to multilevel (mixed) and panel data (xtreg) |
Excercise 3/6 |
| 08.12.2026 Session 9 |
Lecture: Explaining lower-level associations with context characteristics. how to introduce random slopes into multilevel models and how to explain them via cross-level interactions |
- |
| 15.12.2026 Session 10 |
Exercise: From the empty to the full multilevel model practical exercises on the construction and successive estimation of nested multilevel models |
Excercise 4/6 |
| 05.01.2027 Session 11 |
Lecture: Whats the center of the variable? how different types of mean-centering affect estimates in multilevel models |
- |
| 12.01.2027 Session 12 |
Excercise: Different effects for different centerings calculating different types of mean-centering and comparing the estimates |
Excercise 5/6 |
| 19.01.2027 Session 13 |
Lecture: The best of the two worlds? how hybrid (between-within) models work and how to apply them |
- |
| 26.01.2027 Session 14 |
Exercise: Applying hybrid models specification and estimation of hybrid models |
Excercise 6/6 |
| 02.02.2027 Session 15 |
Summary, feedback on seminar, feedback on term paper ideas. | - |
Hox, Joop J.; Moerbeek, Mirjam; van de Schoot, Rens (2010): Multilevel analysis: Techniques and applications: Routledge.
Hox, Joop J.; Roberts, J. Kyle (Hg.) (2011): Handbook of advanced multilevel analysis. ebrary, Inc. New York: Routledge (European Association of Methodology Series).
Rabe-Hesketh, Sophia; Skrondal, Anders (2008): Multilevel and longitudinal modeling using Stata: Stata Press.
Snijders, Tom; Bosker, Roel (1999): Multilevel analysis. An introduction to basic and applied multilevel analy-sis: London: Sage.
Allison, Paul D. 2009. Fixed effects regression models: SAGE Publications.
Wooldridge, Jeffrey M. 2010. Econometric analysis of cross section and panel data: MIT press.