Advanced Spatial Analysis

This website is preserved as an Archive for the NIH-funded GISPopSci / Advanced Spatial Analysis Training Programs (2005–2013). Current resources in support of
Spatially Integrated Social Science
are now available at the following:


Multilevel Modeling Course Preparation

The workshop will be based on two sessions a day: morning and afternoon. Each session is usually composed of three parts:

      Practical Application

All the modelling will be done in the software environment MLwiN 2.11.

Outline of sessions:

Sessions 1 to 8 (Monday through Thursday): concepts, specification and interpretation of 2-level model, continuous response

Sessions 9 & 10 (Friday): IGLS/RIGLS estimation, introduction to MCMC estimation; 3-level model; discrete response

At the start of the course, building on standard single-level models, we develop the two-level model with continuous predictors and response. Examples include house-prices varying over districts, and pupil progress varying by school. Later, these models are extended to cover complex variation, both within and between levels, three-level models, and models with categorical predictors and response (the multilevel logit model). Throughout the course, we shall use graphical examples, verbal equations, algebraic formulation, class-based model interpretation, and practical modelling using MLwiN. At the end of the course we will discuss more technical issues. This will include full Bayesian estimation as well as maximum likelihood estimates (IGLS/RIGLS). We will also discuss and illustrate shrinkage, and precision weighted estimation. Finally, we will also look at multilevel logit and spatial models."

Course Objectives

On completion of the course, participants will be able to recognize a multilevel structure; distinguish between fixed and random classifications; specify a multilevel model with complex variation at a number of levels; and fit and interpret a range of multilevel models.

Course Preparation

Participants taking this course should have strong familiarity with regression modeling and inferential statistics, especially regression intercepts and slopes, standard errors, t-ratios, residuals, dummy variables, and the concepts of variance and covariance. Even so, the aim is not to cover mathematical derivations and statistical theory, but to provide a conceptual framework and 'hands-on' experience with MLwiN. Two pieces of preparation are strongly recommended if your knowledge is rusty

LEMMA: Multilevel Modelling online course: Courses 1 to 3 would be an ideal preparation for this course;

Working through provided Powerpoint slides Stat101: A reminder of some key ideas in statistics, p48-62 of the downloadable materials (see below)

How do I get hold of the materials

Note: The three main files are large and can take time to load on a browser.
The session files are small and the browser load time is decreased.

1) KJFullMLlectures2009.pdf (4.4MB) these are the 477 Powerpoint slides for the course; we strongly recommend these are printed off so you can easily annotate as course proceeds. I suggest that it is printed two slides to a page and back to back. It needs to be printed with a high a resolution as possible as some of the detail is small but important. We also suggest that is ring bound as a separate document from the training manual. It can be downloaded from KJ Full ML lectures 2009

View the KJ Full ML lectures 2009 (Multilevel modeling: concepts and applications) material by sessions

2) TrainingManual2009NewVersion (9.8MB) is a large PDF file. There are 328 pages. This is a script of the models we are going to fit. We strongly recommend that this is printed off so you can easily annotate as course proceeds. You will also need it to undertake the practical sessions. We also suggest that is ring bound as a separate document from the slide material. It can be downloaded from Training Manual 2009 New Version

View the Training Manual 2009 material by chapters

3) Multilevel readings (38MB) is a collection of papers that have been written using multilevel modelling. There is a mixture of expository papers and substantial applications. You can browse through these at your convenience. It can be downloaded from Multilevel readings

View the Multilevel readings material by article titles

  1. Context, composition and heterogeneity
  2. Contextual models of urban house price
  3. Revisiting Robinson: The perils of individualistic and ecologic fallacy
  4. The relevance of multilevel statistical methods for identifying causal neighbourhood effects
  5. Multilevel perspectives on modeling census data
  6. Multilevel Studies displays sideways
  7. Multilevel Methods, Theory, and Analysis
  8. Macrosocial Determinants of Population Health displays sideways
  9. Subramanian et al Respond to "Think Conceptually, Act Cautiously"
  10. Comparing Individual- and Area-based Socioeconomic Measures for the Surveillance of Health Disparities: A Multilevel Analysis of Massachusetts Births, 1989-1991
  11. Social Trust and Self-Rated Health in US Communities: A Multilevel Analysis
  12. Does the state you live in make a difference? Multilevel analysis of self-rated health in the US
  13. Using Multilevel Models to Model Heterogeneity: Potential and Pitfalls
  14. Levels of analysis for the study of environmental health disparities
  15. Multilevel glossary
  16. KJ : some publications