Causal Inference With Selectively-Deconfounded Data

Given only data generated by a standard confounding graph with unobserved confounder, the Average Treatment Effect (ATE) is not identifiable. To estimate the ATE, a practitioner must then either (a) collect deconfounded data; (b) run a clinical trial; or (c) elucidate further properties of the causal graph that might render the ATE identifiable. In this paper, we consider the benefit of incorporating a large confounded observational dataset (confounder unobserved) alongside a small deconfounded observational dataset (confounder revealed) when estimating the ATE. Our theoretical results show that the inclusion of confounded data can significantly reduce the quantity of deconfounded data required to estimate the ATE to within a desired accuracy level. Moreover, in some cases---say, genetics---we could imagine retrospectively selecting samples to deconfound. We demonstrate that by actively selecting these samples based upon the (already observed) treatment and outcome, we can reduce sample complexity further. Our theoretical and empirical results establish that the worst-case relative performance of our approach (vs. a natural benchmark) is bounded while our best-case gains are unbounded. Finally, we demonstrate the benefits of selective deconfounding using a large real-world dataset related to genetic mutation in cancer.

[1]  Catherine P. Bradshaw,et al.  The use of propensity scores to assess the generalizability of results from randomized trials , 2011, Journal of the Royal Statistical Society. Series A,.

[2]  Roman Vershynin,et al.  High-Dimensional Probability , 2018 .

[3]  Rajeev Dehejia,et al.  Propensity Score-Matching Methods for Nonexperimental Causal Studies , 2002, Review of Economics and Statistics.

[4]  Uri Shalit,et al.  Estimating individual treatment effect: generalization bounds and algorithms , 2016, ICML.

[5]  Eric J. Tchetgen Tchetgen,et al.  Multiply robust causal inference with double‐negative control adjustment for categorical unmeasured confounding , 2018, Journal of the Royal Statistical Society. Series B, Statistical methodology.

[6]  Z. Geng,et al.  Identifying Causal Effects With Proxy Variables of an Unmeasured Confounder. , 2016, Biometrika.

[7]  P. Holland.  Statistics and Causal Inference , 1985 .

[8]  D. Rubin.  Estimating causal effects of treatments in randomized and nonrandomized studies. , 1974 .

[9]  A. Knudson,et al.  Two genetic hits (more or less) to cancer , 2001, Nature Reviews Cancer.

[10]  M. J. Laan,et al.  Targeted Learning: Causal Inference for Observational and Experimental Data , 2011 .

[11]  Mihaela van der Schaar,et al.  Bayesian Inference of Individualized Treatment Effects using Multi-task Gaussian Processes , 2017, NIPS.

[12]  G. Imbens,et al.  Efficient Estimation of Average Treatment Effects Using the Estimated Propensity Score , 2002 .

[13]  Chiara Sabatti,et al.  Causal inference in genetic trio studies , 2020, Proceedings of the National Academy of Sciences.

[14]  Stefan Wager,et al.  Estimation and Inference of Heterogeneous Treatment Effects using Random Forests , 2015, Journal of the American Statistical Association.

[15]  D. Rubin,et al.  The central role of the propensity score in observational studies for causal effects , 1983 .

[16]  G. Imbens,et al.  Efficient Estimation of Average Treatment Effects Using the Estimated Propensity Score , 2000 .

[17]  J. Pearl.  Causality: Models, Reasoning and Inference , 2000 .

[18]  O. Papaspiliopoulos.  High-Dimensional Probability: An Introduction with Applications in Data Science , 2020 .

[19]  Uri Shalit,et al.  Removing Hidden Confounding by Experimental Grounding , 2018, NeurIPS.

[20]  Max Welling,et al.  Causal Effect Inference with Deep Latent-Variable Models , 2017, NIPS 2017.

[21]  Ian Tomlinson,et al.  A panoply of errors: polymerase proofreading domain mutations in cancer , 2016, Nature Reviews Cancer.

[22]  J. Pearl.  Causal diagrams for empirical research , 1995 .

[23]  Roland R. Ramsahai,et al.  From sample average treatment effect to population average treatment effect on the treated: combining experimental with observational studies to estimate population treatment effects , 2015 .

[24]  J. Pearl,et al.  Measurement bias and effect restoration in causal inference , 2014 .

[25]  Elias Bareinboim,et al.  A General Algorithm for Deciding Transportability of Experimental Results , 2013, ArXiv.

[26]  G. Imbens,et al.  Estimating Treatment Effects using Multiple Surrogates: The Role of the Surrogate Score and the Surrogate Index , 2016, 1603.09326.

[27]  D. McCaffrey,et al.  Propensity score estimation with boosted regression for evaluating causal effects in observational studies. , 2004, Psychological methods.

[28]  E. Montecino-Rodriguez,et al.  Causes, consequences, and reversal of immune system aging. , 2013, The Journal of clinical investigation.