Invariant Representation Learning for Treatment Effect Estimation (2020-11-24T00:00:00.000000Z)

TL;DR

Nearly Invariant Causal Estimation (NICE) is developed, which uses invariant risk minimization (IRM) to learn a representation of the covariates that, under some assumptions, strips out bad controls but preserves sufficient information to adjust for confounding.

Abstract

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 \emph{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. NICE is appropriate in the following setting. i) We observe data from multiple environments that share a common causal mechanism for the outcome, but that differ in other ways. ii) In each environment, the collected covariates are a superset of the causal parents of the outcome, and contain sufficient information for causal identification. iii) But the covariates also may contain bad controls, and it is unknown which covariates are safe to adjust for and which ones induce bias. 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 existing methods that adjust for all the covariates.

Authors

Victor Veitch

5 papers

D. Blei

8 papers

Claudia Shi

3 papers

TL;DR

Abstract

Authors

References49 items

The Risks of Invariant Risk Minimization

A Calculus for Stochastic Interventions: Causal Effect Identification and Surrogate Experiments

Domain Adaptation As a Problem of Inference on Graphical Models

Making sense of sensitivity: extending omitted variable bias

Invariant Risk Minimization

Causal Regularization

Adapting Neural Networks for the Estimation of Treatment Effects

Review of Causal Discovery Methods Based on Graphical Models

Invariance, Causality and Robustness

GANITE: Estimation of Individualized Treatment Effects using Generative Adversarial Nets

Outcome‐adaptive lasso: Variable selection for causal inference

Overlap in observational studies with high-dimensional covariates

Deep IV: A Flexible Approach for Counterfactual Prediction

Domain Adaptation by Using Causal Inference to Predict Invariant Conditional Distributions

Invariant Causal Prediction for Nonlinear Models

Invariant Causal Prediction for Sequential Data

Causal Effect Inference with Deep Latent-Variable Models

Instrumental variables as bias amplifiers with general outcome and confounding

The Magnitude and Direction of Collider Bias for Binary Variables

Estimating individual treatment effect: generalization bounds and algorithms

Multi-Level Cause-Effect Systems

Invariant Models for Causal Transfer Learning

Recursive partitioning for heterogeneous causal effects

Causal inference by using invariant prediction: identification and confidence intervals

Adam: A Method for Stochastic Optimization

Visual Causal Feature Learning

Confounding Equivalence in Causal Inference

A New Criterion for Confounder Selection

An Introduction to Propensity Score Methods for Reducing the Effects of Confounding in Observational Studies

Doubly robust estimation of causal effects.

Bayesian Nonparametric Modeling for Causal Inference

An Introduction to the Augmented Inverse Propensity Weighted Estimator

Confounding Equivalence in Causal Inference

Should observational studies be designed to allow lack of balance in covariate distributions across treatment groups?

Covariate adjustment in randomized trials with binary outcomes: Targeted maximum likelihood estimation

Do Instrumental Variables Belong in Propensity Scores?

A Linear Non-Gaussian Acyclic Model for Causal Discovery

Gender Differences in Mate Selection: Evidence From a Speed Dating Experiment

Improving Quantitative Studies of International Conflict: A Conjecture

Some surprising results about covariate adjustment in logistic regression models

The central role of the propensity score in observational studies for causal effects

Comment on “Causal inference using invariant prediction”

Modelling Gene Expression Data using Dynamic Bayesian Networks

Nonparametric estimation of average treatment effects under exogeneity: A review.’

Causality : Models , Reasoning , and Inference

Overt Bias in Observational Studies

Constructing Bayesian Network Models of Gene Expression Networks from Microarray Data

A new approach to causal inference in mortality studies with a sustained exposure period—application to control of the healthy worker survivor effect

The Statistical Implications of a System of Simultaneous Equations

Field of Study

Journal Information

Name

Volume

Venue Information

Name

Type

URL

Alternate Names