Choosing which covariates to control for in regression models is a well-known challenge, and clear guidance on this issue is often lacking. Causal diagrams, specifically directed acyclic graphs (DAGs), offer a powerful framework for addressing this challenge. This workshop discusses the paper “A Crash Course in Good and Bad Controls” (Cinelli, Forney, & Pearl, 2022), which introduces the concepts of “good” and “bad” controls in statistical regression analysis and clarifies when adding a variable improves or undermines the accuracy of causal effect estimates.

Through a series of simulated examples, we will demonstrate how different variable structures can lead to bias reduction (good controls), bias amplification (bad controls), or neutral effects, and discuss how some neutral controls can nonetheless influence the precision of estimates. The workshop’s primary goal is to show how DAGs can be used systematically to identify appropriate control variables, enabling valid causal interpretations of regression models.

While the material is primarily conceptual, all simulations and DAG visualizations will be demonstrated in R, with additional discussion on how AI tools can be used to support this workflow.