9am - 5:45pm
Monday 18 February - Wednesday 20 February 2019
Multilevel modelling for social scientists
This course provides a thorough discussion of multilevel models and demonstrates how they can be deployed to answer social science questions.
Complex structures exist in the social world and can influence the experiences of individuals – for example, the school you attend can have an impact on the grades you achieve and future life chances, and life expectancy varies dramatically across neighbourhoods, even in the same city. This short course introduces statistical methods for dealing effectively with these types of data structures, enabling us to make robust inferences about the effects of groups, individuals, and the effects of being in a particular group on different individuals.
We will start by covering some of the basic concepts in multilevel modelling and the fundamentals of random intercept and random coefficient models. We will then move on to consider more advanced topics including: nonlinear models for binary responses, repeated measures, and cross-classified models. Throughout the course, the emphasis will be on the practical issues involved in multilevel modelling and the critical interpretation of results, rather than on the underlying statistical derivations. Computer exercises in R will accompany the formal teaching sessions.
The content will cover:
- Multilevel data structures
- Random intercept models
- Random coefficient models
- Context effects and cross-level interactions
- Multilevel models for binary responses
- Longitudinal modelling
- Cross-classified data structures
Hands on practical workshops will provide you with experience of:
- Fitting multilevel models to real world data
- Models designed to deal with linear and binary responses
- Analysing longitudinal data
On successful completion of this course, participants will be able to:
- Have a critical understanding of the ideas behind multilevel modelling, and to know when their use is appropriate (C,K)
- Be able to fit multilevel models to continuous and binary response data (C,P)
- Have a comprehensive understanding of more advanced topics including binary response models, and methods for longitudinal data (K)
- Be able to engage with existing research studies using multilevel models, highlighting their key strengths and weaknesses (C,T)
- Be able to interpret the results from multilevel models critically (C,T)
Key: C-Cognitive/Analytical; K-Subject Knowledge; T-Transferable Skills; P- Professional/ Practical skills
Professor Brunton-Smith is a quantitative social scientist with particular expertise in multilevel modelling, survey methodology, and missing data. He has taught multilevel modelling workshops nationally and internationally.
Level of study
Basic (some basic knowledge of simple linear regression will be an advantage)
Varies according to status:
- £595 - Government/commercial sector
- £495 - Educational/charitable sector
- £395 - Students.
Hox, J J. (2002) Multilevel analysis: Techniques and applications. Second edition. Routledge. Chapters 1 and 2
Brunton-Smith, I., and Sturgis, P. (2011) ‘Do neighborhoods generate fear of crime?: An empirical test using the British Crime Survey’. Criminology, 49(2): 331-369.
For participants unfamiliar with R, the following introductory materials will be of use.
No prior knowledge is required, but it is assumed that participants will have a basic understanding of regression. All computing workshops will be in R, using the GUI RStudio. For a basic introduction to R for data manipulation and analysis, please click here.
Participants on the course will include some students completing the MSc in Social Research Methods