Theory Construction and Causal Modeling

This workshop teaches participants strategies for constructing causal theories and casual models on a conceptual level, and provides participants with guidelines for how to statistically analyze data to gain perspectives on those theories. The first part of the workshop reviews conceptions of causality and then describes examples of the building blocks of causal theories. These include direct causal effects, indirect causal effects (mediation), moderated causal relationships, spurious relationships, reciprocal causality, and feedback loops. The presenter will also discuss the concepts of mediated moderation, moderated mediation, mediated mediation, and moderated moderation. The presenter will also describe a range of thought experiments and thinking strategies for generating research ideas using these concepts and illustrate how to use influence diagrams to summarize one’s theory. The second part of the workshop focuses on issues to consider when statistically analyzing data to test a causal model. This includes how to translate a path diagram into a set of equations that can then be tested using either structural equation modeling or general(ized) linear models. The presenter will distinguish between limited information estimation and full information estimation and discuss the basics of statistical modeling in each case. Presenter will also briefly consider some of the more common sources of model specification error, including measurement misspecification, left out variable error, and function misspecification. The emphasis is on providing participants with a non-technical appreciation for these issues and practical analytic strategies to deal with them.