30211

WiSe 19/20: Introduction to Multilevel Analysis and its Application in Stata and R

Jan Paul Heisig

Hinweise für Studierende

Seats are limited, so please register by email to jan.heisig@wzb.eu before enrolling.

Kommentar

Social scientists often work with multilevel data where lower-level units (e.g., individuals) are nested in one or several upper-level units (e.g., countries or neighborhoods). An important reason for analyzing such data is an interest in 'context effects', that is, in the effects of upper-level characteristics on lower-level outcomes and lower-level relationships. For example, one might wonder how a country's level of welfare spending affects the happiness of its citizens or whether the relationship between students' socio-economic status and academic achievement differs across school types (e.g., between public and private schools). The analysis of multilevel data to answer these and similar research questions entails some statistical challenges because lower-level units belonging to the same upper-level unit (e.g., citizens from the same country) tend to be more similar to each other than units that belong (e.g., citizens from different countries), thus violating the independence assumptions underlying traditional regression analysis. The course aims to provide students with a solid understanding of the challenges and potentials of multilevel analysis, to introduce them to the implementations of relevant methods in the software packages Stata and R, and to enable them to conduct independent research projects based on multilevel data. The main focus will be on so-called multilevel or mixed-effects models with random intercepts and slopes, which are the most common tool for analyzing multilevel data in sociology. In addition, we will discuss two-step methods and cluster-robust variance estimation as alternative approaches. Participants should have a solid understanding of conventional regression analysis and should also be familiar with at least one of the two software packages used in the course (R and Stata). Schließen

Literaturhinweise

Preparatory readings (not required and most will be discussed in class):

Bryan, M. L. and Jenkins, S. P. (2016). Multilevel Modelling of Country Effects: A Cautionary Tale. European Sociological Review, 32, 3–22.
Cameron, A. C. and Miller, D. L. (2015). A Practitioner’s Guide to Cluster-Robust Inference. Journal of Human Resources, 50, 317–372.
Heisig, J. P., Schaeffer, M. and Giesecke, J. (2017). The Costs of Simplicity: Why Multilevel Models May Benefit from Accounting for Cross-Cluster Differences in the Effects of Controls. American Sociological Review, 82, 796–827.
Lewis, J. B. and Linzer, D. A. (2005). Estimating Regression Models in Which the Dependent Variable is Based on Estimates. Political Analysis, 13, 345–364.
Raudenbush, S. W. and Bryk, A. S. (2002). Hierarchical Linear Models: Applications and Data Analysis Methods, Thousand Oaks: Sage.
Snijders, T. A. B. and Bosker, R. J. (2012). Multilevel Analysis: An Introduction to Basic and Advanced Multilevel Modeling, London: Sage.

Schließen