Sequential Deconfounding for Causal Inference with Unobserved Confounders

Abstract

Using observational data to estimate the effect of a treatment is a powerful tool for decision-making when randomized experiments are infeasible or costly. However, observational data often yields biased estimates of treatment effects, since treatment assignment can be confounded by unobserved variables. A remedy is offered by deconfounding methods that adjust for such unobserved confounders. In this paper, we develop the Sequential Deconfounder, a method that enables estimating individualized treatment effects over time in presence of unobserved confounders. This is the first deconfounding method that can be used in a general sequential setting (i.e., with one or more treatments assigned at each timestep). The Sequential Deconfounder uses a novel Gaussian process latent variable model to infer substitutes for the unobserved confounders, which are then used in conjunction with an outcome model to estimate treatment effects over time. We prove that using our method yields unbiased estimates of individualized treatment responses over time. Using simulated and real medical data, we demonstrate the efficacy of our method in deconfounding the estimation of treatment responses over time.

References in corpus (14)

• Inferring causal impact using Bayesian structural time-series models
• Gaussian Processes for Big Data
• Artificial Intelligence Framework for Simulating Clinical Decision-Making: A Markov Decision Process Approach
• Causal Effect Inference with Deep Latent-Variable Models
• Reliable Decision Support using Counterfactual Models
• Assessing lack of common support in causal inference using Bayesian nonparametrics: Implications for evaluating the effect of breastfeeding on children's cognitive outcomes
• The Deconfounded Recommender: A Causal Inference Approach to Recommendation
• Deconfounding Reinforcement Learning in Observational Settings
• Implicit Causal Models for Genome-wide Association Studies
• Causal Inference using Gaussian Processes with Structured Latent Confounders
• Towards Clarifying the Theory of the Deconfounder
• The Medical Deconfounder: Assessing Treatment Effects with Electronic Health Records
• Decomposing feature-level variation with Covariate Gaussian Process Latent Variable Models
• Discussion of "The Blessings of Multiple Causes" by Wang and Blei

Cited by in corpus (1)

• Generalizing Off-Policy Learning under Sample Selection Bias