Multilevel Modeling 2500-PL-PS-SP15-19

Many phenomena investigated by social scientists have a multilevel structure. For example, pupils are nested in classes and schools, employees in companies, citizens in countries, or longitudinal measurements in persons. With this type of data, it is crucial to recognize that each level of analysis may serve as a distinct source of variance. Namely, pupils may differ in their language skills not only because of the variability in individual features such as SES but also because they belong to different classes and/or attend different schools. If such interdependence between observations is not modeled properly, statistical analyses may lead to incorrect conclusions.

Multilevel modeling (MLM) is a statistical technique designed to handle nested datasets. This approach is not only capable of accounting for observations’ interdependence but may also be used as a tool for testing hypotheses on the interplay between individuals and the context in which the latter are embedded.

This course aims to introduce students to multilevel analysis. Throughout the consecutive classes, participants will be presented with different kinds of questions that may be asked and answered with multilevel models. We will start with an overview of MLM applications. Next, we will consider 2-level models for continuous, binary, and count data. Finally, we will cover contextual effects. The classes will involve a combination of lectures and lab sessions focusing on the specification, estimation, and interpretation of multilevel models. All analyses would be performed in R.