Variable selection is an inevitable step in observational research. When new data are to be collected, researchers are tasked with identifying which relevant variables should be measured. Similarly, when analyzing available secondary data, researchers are faced with the decision of which variables to consider in the analysis.

To inform strategies for proper estimation of causal effects, modern causal inference approaches dictate that Directed Acyclic Graphs (DAGs) should be used to explicitly model the assumptions underlying causal relationships. DAGs (see figure) are compact, intuitive, visual tools that, despite their simple appearance, entail complex mathematical relationships among the included variables and are therefore useful to inform design and analysis approaches to answer causal questions. 

By applying the principle of “The Wisdom of Crowds” - the idea that the aggregation of judgements of a group of people is often as good as or better than the judgement of the best performing ‘expert’ in the group - we aim to summarize independent, decentralized opinions and knowledge about stroke-related processes from a ‘crowd’ of experts in the topic from various research disciplines. Ultimately, we hope to unveil the true underlying data generation processes and generate a so-called consensus DAG that can inform approaches to answer meaningful clinical questions in stroke research.

Study participants are experts with experience in stroke research. They take part in a series of causal inference-focused lectures, followed by a Q&A session and hands-on practice building their own DAGs using Causalify.

Further information:

Hernán, Miguel A., and James M. Robins. 2020. “Causal Inference: What If.” Boca Raton: Chapman & Hill/CRC 2020.

Surowiecki, James. 2005. The Wisdom of Crowds. Anchor.

Talbot, Denis, and Victoria Kubuta Massamba. 2019. “A Descriptive Review of Variable Selection Methods in Four Epidemiologic Journals: There Is Still Room for Improvement.” European Journal of Epidemiology 34 (8): 725–30.