Description

Most questions of interest are fundamentally questions of causality, i.e. what is the effect of some variable x on some other variable y. This course introduces the statistical methods that are currently available for studying such questions. Modern causal analyses are based on either the potential outcomes framework or the structural equation framework. Advantages and disadvantages of both frameworks will be discussed. Rubin developed the potential outcome approach into a powerful formal framework, the Rubin Causal Model (RCM), for assessing causation in observational data. The resulting methods make only few assumptions, which adds credibility to the results obtained. Structural Equation Models (SEM) make more assumptions, and have been a dominant influence in management science and marketing for decades. More recently, it has captured the interest of the other social sciences. A Structural equation model is a causal framework that provides elegant and relatively easy-to-use methods to deal with many important issues that applied empirical researchers have to face. These include multi-level random effects, endogeneity, sample selection and missing data. Recent versions of Stata include implementations of these methods. This course aims to introduce the theory of the RCM and SEM and show how to apply the theory using Stata.

Organizers: