Naïve regression requires weaker assumptions than factor models to adjust for multiple cause confounding

07/24/2020
by   Justin Grimmer, et al.
0

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 by 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.

READ FULL TEXT

page 1

page 2

page 3

page 4

research
05/17/2018

The Blessings of Multiple Causes

Causal inference from observation data often assumes "strong ignorabilit...
research
05/21/2018

Multiple Causal Inference with Latent Confounding

Causal inference from observational data requires assumptions. These ass...
research
04/16/2021

Sequential Deconfounding for Causal Inference with Unobserved Confounders

Using observational data to estimate the effect of a treatment is a powe...
research
11/09/2020

Identifying effects of multiple treatments in the presence of unmeasured confounding

Identification of treatment effects in the presence of unmeasured confou...
research
08/17/2022

Estimating individual treatment effects under unobserved confounding using binary instruments

Estimating individual treatment effects (ITEs) from observational data i...
research
01/11/2019

The Estimation of Causal Effects of Multiple Treatments in Observational Studies Using Bayesian Additive Regression Trees

There is currently a dearth of appropriate methods to estimate the causa...
research
01/17/2023

Causal Falsification of Digital Twins

Digital twins hold substantial promise in many applications, but rigorou...

Please sign up or login with your details

Forgot password? Click here to reset