Online Causal Inference Seminar

Free and open to the public
All seminars are on Tuesdays at 8:30 am PT (11:30 am ET / 4:30 pm London / 5:30 pm Berlin).

Tuesday, November 10, 2020 [Link to join] (webinar ID: 942 3597 2707 password: 007080)

• Speakers: Justin Grimmer (Stanford), Dean Knox (Wharton), Brandon Stewart (Princeton)
• Discussants: Ilya Shpitser (Johns Hopkins), Betsy Ogburn (Johns Hopkins), Eric Tchetgen Tchetgen (Wharton)
• Title: Naïve regression requires weaker assumptions than factor models to adjust for multiple cause confounding
• Abstract: The empirical practice of using factor models to adjust for shared, unobserved confounders, Z, in observational settings with multiple treatments, A, is widespread in fields including genetics, networks, medicine, and politics. Wang and Blei (2019, WB) formalizes these procedures and develops the “deconfounder,” a causal inference method using factor models of A to estimate “substitute confounders,” Ẑ, then estimating treatment effects—regressing the outcome, Y , on part of A while adjusting for Ẑ. WB claim the deconfounder is unbiased when there are no single-cause confounders and Ẑ is “pinpointed.” We clarify pinpointing requires each confounder to affect infinitely many treatments. We prove under these assumptions, a naïve semiparametric regression of Y on A is asymptotically unbiased. Deconfounder variants nesting this regression are therefore also asymptotically unbiased, but variants using Ẑ and subsets of causes require further untestable assumptions. We replicate every deconfounder analysis with available data and find it fails to consistently outperform naïve regression. In practice, the deconfounder produces implausible estimates in WB’s case study to movie earnings: estimates suggest comic author Stan Lee’s cameo appearances causally contributed $15.5 billion, most of Marvel movie revenue. We conclude neither approach is a viable substitute for careful research design in real-world applications.
  - [Paper]