Teacher

Prof. Tarani Chandola, University of Manchester

Timing and place

Tue 9.9.2014 10-17, SSKH IT-hall (Snellmaninkatu 12, 1st floor)

Registration

Registration for the course at WebOodi.

Aim

The aim of this unit is to teach students the concepts underlying logistic multilevel models and its application using MLwin software. Particular attention will be paid to Monte Carlo Markov Chain (MCMC) estimation. Datasets illustrating the application of such models to health and social research will be used. Basic knowledge on multilevel linear regression models using MLwin and single level logistic regression models will be assumed.

Objectives

After the course, students should be able to:

• Recognise when there is a need for more advanced modelling techniques
• Apply multilevel techniques to binary data
• Acquire knowledge on how to use the MLwiN software for fitting multilevel logistic regression models
• Understand why multilevel analysis may be more appropriate for certain data designs such as clustered designs
• Discuss the basic underlying theory of multilevel models
• Interpret in non-technical language the results from a multilevel analysis of a large dataset
• Use MLwiN software for multilevel logistic regression analysis
• Students will develop skills for using multilevel models for their own research and for reading journal papers that very often employ multilevel analysis

Course Content

This unit will revise the basic concepts of linear and logistic regression models before going on to introduce multilevel logistic regression models. A brief introduction to Bayesian statistics will be used to introduce MCMC estimation. The unit will pay particular attention to the trace diagnostics using MCMC estimation.

Teaching and Learning

The course will consist of a one day (6 hours) lecture-based sessions and practical sessions (MLwiN workshops).

• 10:00-10:45am Revision of multilevel linear regression models and single level logistic regression models
• 10:45-11:15am Introduction to multilevel logistic regression models
• 11:15-11:30 Break
• 11:30-13:00 Computer based practical using MLwin: multilevel logistic regression models
• 13:00-14:00 Lunch 14:00-15:15 MCMC Estimation in MLwin
• 15:15-15:30 Break
• 15:30-17:00 Computer based practical using MLwin: MCMC Estimation in MLwin

Key Reading

• Rasbash, J., Steele, F., Browne, W. and Goldstein, H. (2009) A user’s guide to MLwiN. Centre for Multilevel Modelling, University of Bristol

Key Online Resource

• LEMMA (Learning Environment for Multilevel Methods and Applications)
• Recommended reading BEFORE the unit: Module 7 – Multilevel Models for Binary Responses

Additional Reading

• Snijders, T.A.B. and Bosker, R.J. (2012). Multilevel Analysis. London: Sage.
• Dobson, A. (2002). An introduction to generalized linear models. Chapman and Hall
• Goldstein, H. (1995). Multilevel Statistical Models. London: Edward Arnold.