Multilevel Models: Pooled and Clustered Data (Chapel Hill, NC)

Instructor(s):

Multilevel models (also known as hierarchical linear models or mixed models) provide an extremely flexible approach to the analysis of a wide array of social and behavioral science data. Multilevel modeling allows for the analysis of non-independent or "clustered" data that arise when studying topics such as siblings nested within families, students nested within classrooms, clients nested within therapists, or voters nested within media markets. Such models also accommodate data clustered or pooled across units and/or time periods--(e.g. 50 states measured annually over 30 years)--often labeled pooled time series data, time series cross-sectional data, or panel data. Multilevel models are explicitly designed to analyze clustered or pooled data structures and can incorporate individual-level predictors, group-level predictors, and individual-by-group-level interactions.

This course provides a general introduction to a variety of applications of multilevel modeling in the social sciences. Equal emphasis is placed on the underlying statistical model and on the estimation and interpretation of empirical data. The course explores methods for so-called robust standard errors in the face of clustered data, along with traditional methods for pooled time series and multi-level models. However, the primary goal is to demonstrate a general solution that encompasses these more narrowly focused approaches. We will consider models for continuous and categorical dependent variables. Time permitting, we will conclude with an introduction to Bayesian multi-level modeling.

Each day will consist of a mixture of lectures and hands-on computer exercises and examples. We will include examples both in R and STATA. We will make substantial use of computer simulations to explore the statistical properties of multi-level models. Participants should already be familiar with multiple regression and OLS. Some familiarity with Generalized Linear Models and Maximum Likelihood estimation is also helpful. No previous knowledge of Bayesian statistics is expected. Some prior exposure to R and STATA is helpful, but not required.

The course will be taught at a level similar to the book Data Analysis Using Regression and Multi-level/Hierarchical Models, by Andrew Gelman and Jennifer Hill (2006), Cambridge University Press. Participants are encouraged to bring their own data and projects with them to the class, as some lab time later in the course can be devoted to helping students address their specific needs.

Fee: Members = $1700; Non-members = $3200

Tags: Multilevel Models

Course Sections