On Multi-Cause Causal Inference with Unobserved Confounding: Counterexamples, Impossibility, and Alternatives
read the original abstract
Unobserved confounding is a central barrier to drawing causal inferences from observational data. Several authors have recently proposed that this barrier can be overcome in the case where one attempts to infer the effects of several variables simultaneously. In this paper, we present two simple, analytical counterexamples that challenge the general claims that are central to these approaches. In addition, we show that nonparametric identification is impossible in this setting. We discuss practical implications, and suggest alternatives to the methods that have been proposed so far in this line of work: using proxy variables and shifting focus to sensitivity analysis.
This paper has not been read by Pith yet.
Forward citations
Cited by 4 Pith papers
-
Shrinkage priors for Bayesian Substitute Confounders
Bayesian shrinkage priors on factor models produce sparse substitute confounders that support consistent regression-adjusted causal estimates under latent variable identification assumptions.
-
An adaptive variance estimator for relative sparsity
A new adaptive variance estimator for relative sparsity coefficients is introduced that fully utilizes the prior asymptotic normality theorem and incorporates variable selection effects.
-
Estimating the Effect of Timing on Coupon Effectiveness
Causal inference framework applied to natural experiments measures coupon timing effects on engagement, shown on company onboarding data and a public retention dataset.
-
Markovianity-Based Conditioning Depth Diagnostics for Hidden Confounding in Observational Datasets
A diagnostic that measures instability of constraint-based causal graphs over increasing conditioning depths to detect hidden confounding or incomplete state in time series observational data.
discussion (0)
Sign in with ORCID, Apple, or X to comment. Anyone can read and Pith papers without signing in.