Canal
Practical Guidelines for Instrumental Variables Methods (303726)
*Luke J. Keele, University of PennsylvaniaKeywords: instrumental variables, causal inference, diagnostic tests
The instrumental variables (IV) framework has seen increasing use as method for estimating causal effects. Critically, the IV framework allows for consistent estimates of causal effects in the presence of unobserved confounding. While IVs are a powerful form of analysis, they also require careful evaluation of the key assumptions, and there are a wide variety of diagnostics tests that should be applied to evaluate potential instruments. In this presentation, I will first review the key IV assumptions. I will provide a detailed overview of key diagnostic tests for IV applications. I will also outline a checklist that can help guide the evaluation of instruments. The presentation will include a real-world application of IV to illustrate concepts and help the audience understand how to evaluate IV based-analyses. The application uses an instrument to estimate the effect of transfer to critical care on mortality in the UK.