140.656.01 Multilevel Statistical Models in Public Health
- 4th term
- 4 credits
- Academic Year:
- 2017 - 2018
- East Baltimore
- Class Times:
- M W, 10:30 - 11:50am
- Lab Times:
Wednesday, 9:00 - 10:20am
Explores conceptual and formal approaches to the design, analysis, and interpretation of studies with a “multilevel” or “hierarchical” (clustered) data structure (e.g., individuals in families in communities). Develops skills to implement and interpret random effects, variance component models that reflect the multi-level structure for both predictor and outcome variables. Topics include: building hierarchies; interpretation of population-average and level-specific summaries; estimation and inference based on variance components; shrinkage estimation; discussion of special topics including centering, use of contextual variables, ecological bias, sample size and missing data within multilevel models. STATA and SAS software are supported.
- Learning Objectives:
- Define multilevel data
- Implement and interpret results associated with Multi-level Statistical Models (MLMs)
- Identify when and why MLMs can or should be used when they are unnecessary or possibly dangerous
- Describe the implications of centering, contextual variables, missing data and ecological bias within MLMs
- Methods of Assessment:
Student evaluation based on a lab materials (short multiple choice quiz plus graphics/model specification/fit), two homework assignments (three short answer questions, a short abstract and peer assessment) and a final exam which is also an analysis of a multilevel data set, presentation of the results, and a written scientific report of the analysis methods and results.
The course grade is labs (40%), homeworks (40%) and final exam (20%).
- Instructor Consent:
No consent required