PSY9160 – Multilevel modelling
Schedule, syllabus and examination date
Multilevel modelling, also known as Hierarchical Linear Modelling, is increasingly used for data analysis in applied fields of Psychology. For example the Journal of Applied Psychology regularly publishes papers with multilevel modelling.
Multilevel analysis is used for data collected in clustered samples for which sampling decisions were taken in several steps (e.g. first choosing firms, then employees in firms). Examples for clustered samples are:
- Individuals in groups (e.g. classes, countries, companies, teams)
- Several measurements (longitudinal) of individuals.
If data of a clustered sample is analysed by means of regression analysis, the assumption is violated that measurements are equally independent from one another. As a consequence standard errors are underestimated and tests produce too often significant results. Multilevel analysis respects groupings as levels of analysis.
The course aims to provide basic knowledge about the design and analysis of multilevel models, including models with 2 or more levels, multivariate and longitudinal multilevel models. After the course, participants are able to design a multilevel model and to conduct basic analyses. They know where to find resources for computing more complex models. They know how to document a multilevel analysis in a research paper.
Basic knowledge in regression analysis and analysis of variance is necessary for following the course. Basic knowledge in Structural Equation Modelling is helpful, but not required.
During the course, we work with SPSS and therefore focus on multilevel models with continuous outcomes. Data for practical exercises will be provided, but the option exists that participants work with their own data.
Day 1: Introduction to multilevel models
General introduction to multilevel models:
- Arguments for application
- Basic principles of analysis: strategy of data analysis, fit statistics
Discuss and define models regarding the following criteria:
- Number of levels
- Dependent variable and predictors (on which level?)
- Alternatively use provided data set
- Calculate 2-level model
Day 2: Advanced multilevel models
Main characteristics of more complex multilevel models and calculation with SPSS:
- 3-level model
- Multivariate model
Day 3: Longitudinal multilevel models
Main characteristics of longitudinal multilevel models and calculation with SPSS:
- Characteristics of a longitudinal analysis
- Inspection of longitudinal data
- Inspection of the error covariance matrix
- Finally: longitudinal analysis
Introduction to multilevel analysis:
Hox, J. (2010). Multilevel analysis. Techniques and applications (10th ed.). Mahwah, NJ: Lawrence Erlbaum. The second chapter can be downloaded from Joop Hox’s homepage http://joophox.net/
Multilevel modelling with SPSS:
Tabachnick, B. G., & Fidell, L. S. (2007). Using multivariate statistics. Boston: Pearson Education. Chapter 15.
Longitudinal multilevel models:
Singer, J. D., & Willett, J. B. (2003). Applied longitudinal data analysis. New York: Oxford University Press. Chapters 1-7.
For å melde deg på, må du fylle ut et nettskjema som gjelder for det semesteret du skal ta kurset.
To apply for this course please fill out the proper form online
Use of participants’ own data set
Participants who plan to use their own data are asked to contact the course leaders before the course starts and send the following information regarding their data set:
- Number of elements (individuals/measurements) on level 1 (=lowest level)
- Number of levels
- Number of groups on higher levels
- Dependent variable (measurement level)
- Independent variables/predictors (measurement level, level in the multilevel model)
Formal prerequisite knowledge
Admission to Ph.d.
Seminar of 21 hours over three full days. Maximum 20 participants.
After each course day, 1) participants work on an analysis (data will be provided) and hand in the SPSS output, 2) answer some questions related to the reading in written form.
3 ECTS points for course, assignments and additional analysis, 2 ECTS points for course and assignments without additional analysis
Additional analysis: Participants design a model during the course and discuss that with the course leader. It is recommended that participants use their data and test an analysis which is close to a research project or a publication. Data can be provided for participants without data. The report should be written according to the APA standard and should be structured as usual: short introduction with hypotheses (approx. 2 pages), methods section with description of measures and methods of data analysis (3-4 pages), results (2-3 pages, tables), short conclusion with methodical strengths and limitations (1-2 pages). Please email the first version of the report to Sabine until 04.03.2011.
Explanations and appeals
If you need confirmation on passing this course, you must do this through studentweb and use the description on this web page for information. We do not give out course confirmations in other ways.