The blessings of multiple causes

DOI10.1080/01621459.2019.1686987zbMATH Open1428.62059arXiv1805.06826OpenAlexW2983831776MaRDI QIDQ5208062FDOQ5208062

Authors: Yixin Wang, David M. Blei 

Publication date: 15 January 2020

Published in: Journal of the American Statistical Association (Search for Journal in Brave)

Abstract: Causal inference from observational data often assumes "ignorability," that all confounders are observed. This assumption is standard yet untestable. However, many scientific studies involve multiple causes, different variables whose effects are simultaneously of interest. We propose the deconfounder, an algorithm that combines unsupervised machine learning and predictive model checking to perform causal inference in multiple-cause settings. The deconfounder infers a latent variable as a substitute for unobserved confounders and then uses that substitute to perform causal inference. We develop theory for the deconfounder, and show that it requires weaker assumptions than classical causal inference. We analyze its performance in three types of studies: semi-simulated data around smoking and lung cancer, semi-simulated data around genome-wide association studies, and a real dataset about actors and movie revenue. The deconfounder provides a checkable approach to estimating closer-to-truth causal effects.

Full work available at URL: https://arxiv.org/abs/1805.06826

Recommendations

• Comment: The challenges of multiple causes
• Causal inference through a witness protection program
• Causal inference in statistics: an overview
• Statistics and causal inference: A review. (With discussion)
• Causal statistical inference in high dimensions

zbMATH Keywords

causal inferencemachine learningunobserved confoundingprobabilistic modelsunconfoundedness

Mathematics Subject Classification ID

Learning and adaptive systems in artificial intelligence (68T05) Causal inference from observational studies (62D20)

Cites Work

• Automatic differentiation variational inference
• Causal inference by using invariant prediction: identification and confidence intervals. With discussion and authors' reply
• Title not available (Why is that?)
• Title not available (Why is that?)
• Estimation and Inference of Heterogeneous Treatment Effects using Random Forests
• Double/debiased machine learning for treatment and structural parameters
• Causal inference for statistics, social, and biomedical sciences. An introduction
• A Generalization of Sampling Without Replacement From a Finite Universe
• Bayesianly justifiable and relevant frequency calculations for the applied statistician
• Comment on J. Neyman and causal inference in experiments and observational studies: On the application of probability theory to agricultural experiments. Essay on principles. Section 9 [Ann. Agric. Sci. 10 (1923), 1--51]
• Estimation of causal effects with multiple treatments: a review and new ideas
• Mixed membership stochastic blockmodels
• Dealing with limited overlap in estimation of average treatment effects
• The central role of the propensity score in observational studies for causal effects
• Accurate Approximations for Posterior Moments and Marginal Densities
• Statistics and Causal Inference
• Foundations of Modern Probability
• The role of the propensity score in estimating dose-response functions
• Characterizing Selection Bias Using Experimental Data
• Title not available (Why is that?)
• Sensitivity Analysis for the Assessment of Causal Vaccine Effects on Viral Load in HIV Vaccine Trials
• Causality. Models, reasoning, and inference
• Elements of causal inference. Foundations and learning algorithms
• Semi-supervised learning in causal and anticausal settings
• The Probability Approach in Econometrics
• Probabilistic Principal Component Analysis
• 10.1162/jmlr.2003.3.4-5.993
• Title not available (Why is that?)
• Learning the parts of objects by non-negative matrix factorization
• Causal Inference With General Treatment Regimes
• A simulation-intensive approach for checking hierarchical models
• On the definition of a confounder
• Title not available (Why is that?)
• Bayesian checking of the second levels of hierarchical models
• Title not available (Why is that?)
• Causal Inference Using Potential Outcomes
• Identification, inference and sensitivity analysis for causal mediation effects
• Identifying the consequences of dynamic treatment strategies: a decision-theoretic overview
• Title not available (Why is that?)
• Population structure and cryptic relatedness in genetic association studies
• Title not available (Why is that?)
• Structured latent factor analysis for large-scale data: identifiability, estimability, and their implications
• Flexible sensitivity analysis for observational studies without observable implications
• Robust Causal Structure Learning with Some Hidden Variables
• The factorized self-controlled case series method: an approach for estimating the effects of many drugs on many outcomes
• Confounder detection in high-dimensional linear models using first moments of spectral measures

Cited In (21)

• Convex non-parametric least squares, causal structures and productivity
• Nonlinear Causal Discovery with Confounders
• Comment: The challenges of multiple causes
• Covariate balancing functional propensity score for functional treatments in cross-sectional observational studies
• Comment on “Blessings of Multiple Causes”
• The impacts of innovation and trade openness on bank market power: the proposal of a minimum distance cost function approach and a causal structure analysis
• Quantifying direct and indirect effect for longitudinal mediator and survival outcome using joint modeling approach
• Dynamic network data envelopment analysis with a sequential structure and behavioural-causal analysis: application to the Chinese banking industry
• Exploring causal mechanisms and quantifying direct and indirect effects using a joint modeling approach for recurrent and terminal events
• Causal inference in genetic trio studies
• The limits of piecemeal causal inference
• De-confounding Causal Inference Using Latent Multiple-Mediator Pathways
• Inference in High-Dimensional Multivariate Response Regression with Hidden Variables
• Rejoinder on: ``Hierarchical inference for genome-wide association studies: a view on methodology with software
• Incorporating causal modeling into data envelopment analysis for performance evaluation
• Why Are Big Data Matrices Approximately Low Rank?
• Comment on: “The Blessings of Multiple Causes” by Yixin Wang and David M. Blei
• Deep latent variable models for generating knockoffs
• Doubly debiased Lasso: high-dimensional inference under hidden confounding
• Characterization of a robust probabilistic framework for brain magnetic resonance image data distributions
• Identifying Effects of Multiple Treatments in the Presence of Unmeasured Confounding

Uses Software

• Stan
• BayesDA
• Edward
• Eigenstrat
• ADVI

This page was built for publication: The blessings of multiple causes

Report a bug (only for logged in users!)Click here to report a bug for this page (MaRDI item Q5208062)