Multilevel and Longitudinal Models

EDU S043 Subject & Catalog Number

Description

Data often have structure that needs to be modeled explicitly. For example, when investigating students' outcomes we need to account for the fact that students are nested inside classes that are in turn nested inside schools. If we are watching students grow, we need to account for the dependence of measurements across time. If we do not account for such structure, our inferences will tend to be wrong. We also lose the opportunity to understand variation across time or groups and make more insightful inferences. This course provides an overall framework—the multilevel model—for thinking about and analyzing these forms of data. We will also deeply investigate specific versions of multilevel models for common forms of longitudinal and clustered data. We focus on applied work, using real data sets and the statistical software R. We will emphasize how to think about the applicability of the methods studied, how they might fail, and what one might do to protect oneself in such circumstances. Applications of multilevel models include random intercept, random slope, and cross-classified models, with extensions to measurement, meta-analysis, and Bayesian methods. We conclude with a final project; this can include thesis or research work.
Prerequisites: S-052, Stat 139, or an equivalent.