Invariant Representation Learning for Treatment Effect Estimation
This work is significant for researchers and practitioners in causal inference, as it provides a method to mitigate bias caused by unknown 'bad controls' when estimating treatment effects from observational data, which is an incremental improvement over existing adjustment methods.
This paper addresses the challenge of confounding in causal inference from observational data, specifically when 'bad controls' are present among the covariates. The authors developed Nearly Invariant Causal Estimation (NICE), which uses invariant risk minimization to learn a representation of covariates that removes bad controls while retaining confounding information. NICE outperforms adjusting for all covariates on synthetic and semi-synthetic data when unknown collider variables and other bad controls are present.
The defining challenge for causal inference from observational data is the presence of `confounders', covariates that affect both treatment assignment and the outcome. To address this challenge, practitioners collect and adjust for the covariates, hoping that they adequately correct for confounding. However, including every observed covariate in the adjustment runs the risk of including `bad controls', variables that induce bias when they are conditioned on. The problem is that we do not always know which variables in the covariate set are safe to adjust for and which are not. To address this problem, we develop Nearly Invariant Causal Estimation (NICE). NICE uses invariant risk minimization (IRM) [Arj19] to learn a representation of the covariates that, under some assumptions, strips out bad controls but preserves sufficient information to adjust for confounding. Adjusting for the learned representation, rather than the covariates themselves, avoids the induced bias and provides valid causal inferences. We evaluate NICE on both synthetic and semi-synthetic data. When the covariates contain unknown collider variables and other bad controls, NICE performs better than adjusting for all the covariates.