Instructor: Hsiu-Ting Yu
Time: MW 08:35-09:55
Location: LEA 210

Objectives:  Multilevel models (also known as hierarchical linear modeling or mixed modeling) provide an extremely flexible approach to the analysis of a wide array of social 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, longitudinal repeated measures nested within individuals, disease incidence nested within census tracts, etc. Traditional general linear models are not well suited for the analysis of these types of data, given the violation of the assumption of independence. In contrast, multilevel models are explicitly designed to analyze clustered data structures and can incorporate individual-level predictors, group-level predictors, and individual-by-group-level interactions.

Without this technique, researchers often resort to focusing separately on the micro-level units (e.g. children within a family) and/or on the macro units of analysis (e.g. families). Focusing on the micro-level units ignores the variability due to the macro-level units, and tends to produce false positive results. Collapsing data over micro-level units to focus on macro-level units ignores a lot of information. Multilevel modeling not only produces power and correct p values at all levels, but it also makes it possible to answer simultaneously questions at each level, between macro-level as well as between micro-level, using macro-level and micro-level predictors.

This seminar will provide a general introduction to a variety of applications of multilevel modeling in the social sciences. Topics to be covered include basic two-level and three-level univariate and multivariate linear regression models, designing multilevel studies and growth curve analysis. If time permits, advanced topics such as multilevel nonlinear models for binary, count, ordinal and multinomial outcomes, latent variable models, and nonstandard applications of multilevel models such as meta-analysis and social network analysis will also be considered.