Multilevel Modeling
Course Description
Statistical methods for analyzing multilevel data (e.g., datasets with clustered, nested, or longitudinal structures) using a framework known variously as multilevel, mixed-effects, hierarchical linear, or random-coefficient modeling. The course reviews ordinary least squares regression and then introduces: the challenges posed by multilevel data, tools for processing multilevel data for analyses, and both maximum-likelihood and Bayesian estimation. Core topics include partial pooling, random intercepts, fixed and random slopes, cross-level interactions, model comparison, troubleshooting, applications to repeated-measures and longitudinal data, and thorough diagnostics and reporting. Emphasis is placed on conceptual understanding paired with hands-on analysis in statistical software. Students apply methods to real datasets and complete an independent project.
Learning Outcomes
At the conclusion of this course, students will be able to:
- Explain the issues that clustered data cause for the linear model
- Contrast the pros and cons of different approaches for clustered data
- Prepare a clustered dataset for multilevel analysis in R
- Estimate various types of multilevel models in R
- Flexibly specify different random effects structures
- Interpret the fixed and random effects estimates
- Explain the rationale and consequences of partial pooling
- Identify the interpretation implied by each centering option
- Detail and assess the assumptions made by a multilevel model
- Decompose repeated and longitudinal measures into components
- Discuss the idea of Bayesian estimation for multilevel models
- Discuss the ideas of nested and cross-classified models
Instructor
- , Instructor
- , Teaching Assistant
Class Time and Location
- M 11:00AM - 12:15PM in Wescoe 4035
- F 02:00PM - 03:45PM in Wescoe 4035
Office Hours
- T 11:00AM - 12:00PM (Dasha, FR 528)
- T 03:00PM - 04:00PM (Jeff, FR 454)
- R 10:00AM - 11:00AM (Dasha, FR 528)
- F 12:00PM - 01:00PM (Jeff, FR 454, odd weeks)
- F 01:00PM - 02:00PM (Jeff, FR 454, even weeks)