The Blessings of Multiple Causes

By Yixin Wang, David M. Blei. 2018

320 citations recorded by the source corpus.

Abstract Causal inference from observational data is a vital problem, but it comes with strong assumptions. Most methods assume that we observe all confounders, variables that affect both the causal variables and the outcome variables. This assumption is standard but it is also untestable. In this article, we develop the deconfounder, a way to do causal inference with weaker assumptions than the traditional methods require. The deconfounder is designed for problems of multiple causal inference: scientific studies that involve multiple causes whose effects are simultaneously of interest. Specifically, the deconfounder combines unsupervised machine learning and predictive model checking to use the dependencies among multiple causes as indirect evidence for some of the unobserved confounders. We develop the deconfounder algorithm, prove that it is unbiased, and show that it requires weaker assumptions than traditional 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 is an effective approach to estimating causal effects in problems of multiple causal inference. Supplementary materials for this article are available online.

Subfields: Artificial Intelligence, Statistics and Probability

How researchers see this paper

The researchers whose trusted networks rank it highest, strongest advocate first. Every researcher gets a different order.

In Artificial Intelligence

In Statistics and Probability

Citations in the corpus

Algorithmic recourse under imperfect causal knowledge: a probabilistic approach (2020, background): “An alternative approach could address causal sufficiency violations by relying on latent variable models to estimate confounders from multiple causes [61] or proxy variables [31], or to work with bounds on causal effects instead [4, 49].”
Time Series Deconfounder: Estimating Treatment Effects over Time in the Presence of Hidden Confounders (2019, methodology): “The replicated treatment assignments are compared with the actual treatment assignments, at,val, using the test statistic T (at): T (at) = EZ [log p(at | Zt, Xt)], (9) related to the marginal log likelihood (Wang & Blei, 2018).”
Causal Confusion in Imitation Learning (2019, background): ““Causal discovery” approaches allow causal inference from pre-recorded observations under constraints [54, 17, 29, 15, 30, 31, 26, 14, 34, 57].”
Multiple Causes: A Causal Graphical View (2019, methodology): “We revisit the deconfounder algorithm in Wang & Blei (2018).”
From Real‐World Patient Data to Individualized Treatment Effects Using Machine Learning: Current and Future Methods to Address Underlying Challenges (2020, methodology): “Wang and Blei 47 observed that the dependencies in the assignment of multiple treatments can be used to infer latent variables that render the treatments independent and thus be used as substitutes of potential hidden confounders present in the observational dataset.”
Multi-cause causal inference with unmeasured confounding and binary outcome. (2019, background): “ome. To address this non-identiﬁability problem, D’Amour (2019) suggests applying classical ideas such as negative controls or sensitivity analysis to the shared confounding setting. Along this line, Wang and Blei (2019b) show that the deconfounder algorithm of Wang and Blei (2019a) is valid given a”
Structural Agnostic Modeling: Adversarial Learning of Causal Graphs (2018, background): “, 2015), or creating shared noise variables (Janzing and Schölkopf, 2018), or via dimensionality reduction (Wang and Blei, 2018).”
Using Text Embeddings for Causal Inference (2019, background): “Work on causal inference with hidden confounding and many treatments is in the same vein [ WB18; RP18; D’A19 ] .”
Multiple Causal Inference with Latent Confounding (2018, methodology): “A formulation of multiple causal inference was also proposed by (Wang and Blei, 2018); they take a model-based approach in the potential outcomes framework that leverages predictive checks.”