Good and Bad Controls: An Introduction to Causal Diagrams

Estimating causal effects in observational data is a key aim of applied social research. Unfortunately, it is also notoriously difficult. The analysis of observational data is fraught with biases that can obscure the true causal relationship between two variables. To resolve this, quantitative social scientists have traditionally been encouraged to control for all variables that introduce indirect correlations, sometimes termed ‘controls’. Unfortunately, not all potential controls are good controls, as some can introduce or amplify bias. While this problem has been long recognised, the task of distinguishing good and bad controls has recently been made much simpler with the emergence of causal diagrams, which have revolutionised the understanding of different sources of bias in the health sciences.

This short talk offers a friendly introduction to causal diagrams for social scientists and explains how they can be used to select appropriate controls when seeking to estimate causal effects in observational data.

This session has been organised by the Social Research Methods Centre at the University of Leeds, and will be led by Dr Peter Tennant, Associate Professor of Health Data Science. From 2018-2024, he was a Fellow of the Alan Turing Institute, and in 2026, he will be the John S Saden Visiting Professor at Yale University. Initially trained as an epidemiologist, his research is focused on adapting and translating contemporary causal inference methods into health and social science research.

This training session will be delivered in person at the University of Leeds.

Please note that bookings for this session are not being managed by the WRDTP. You can access the booking form and further information about this event by clicking the ‘BOOK NOW’ button below.

Please contact srm@leeds.ac.uk with any enquiries.