Robust causal inference for point exposures with missing confounders

被引:0
|
作者
Levis, Alexander W. [1 ]
Mukherjee, Rajarshi [2 ]
Wang, Rui [2 ,3 ,4 ]
Haneuse, Sebastien [2 ]
机构
[1] Carnegie Mellon Univ, Dept Stat & Data Sci, Pittsburgh, PA 15213 USA
[2] Harvard Univ, Dept Biostat, Boston, MA 02115 USA
[3] Harvard Pilgrim Hlth Care Inst, Dept Populat Med, Boston, MA USA
[4] Harvard Med Sch, Boston, MA USA
关键词
Causal inference; missing data; multiply robust; semiparametric theory; BARIATRIC SURGERY;
D O I
10.1002/cjs.11832
中图分类号
O21 [概率论与数理统计]; C8 [统计学];
学科分类号
020208 ; 070103 ; 0714 ;
摘要
Large observational databases are often subject to missing data. As such, methods for causal inference must simultaneously handle confounding and missingness; surprisingly little work has been done at this intersection. Motivated by this, we propose an efficient and robust estimator of the causal average treatment effect from cohort studies when confounders are missing at random. The approach is based on a novel factorization of the likelihood that, unlike alternative methods, facilitates flexible modelling of nuisance functions (e.g., with state-of-the-art machine learning methods) while maintaining nominal convergence rates of the final estimators. Simulated data, derived from an electronic health record-based study of the long-term effects of bariatric surgery on weight outcomes, verify the robustness properties of the proposed estimators in finite samples. Our approach may serve as a theoretical benchmark against which ad hoc methods may be assessed. Les grandes bases de donn & eacute;es observationnelles sont souvent confront & eacute;es au probl & egrave;me des donn & eacute;es manquantes. De ce fait, les m & eacute;thodes d'inf & eacute;rence causale doivent g & eacute;rer & agrave; la fois les facteurs de confusion et les donn & eacute;es manquantes, un domaine & eacute;tonnamment peu explor & eacute;. Pour r & eacute;pondre & agrave; ce d & eacute;fi, les auteurs de cet article proposent un estimateur efficace et robuste de l'effet causal moyen du traitement. Cet estimateur est sp & eacute;cialement con & ccedil;u pour les & eacute;tudes de cohortes o & ugrave; les facteurs de confusion sont manquants de mani & egrave;re al & eacute;atoire. L'approche propos & eacute;e s'appuie sur une nouvelle factorisation de la vraisemblance. Contrairement & agrave; d'autres m & eacute;thodes, celle-ci permet de mod & eacute;liser les fonctions de nuisance de mani & egrave;re flexible, notamment gr & acirc;ce & agrave; des techniques avanc & eacute;es d'apprentissage automatique, et ce tout en maintenant les taux de convergence attendus des estimateurs finaux. Des simulations bas & eacute;es sur des dossiers m & eacute;dicaux & eacute;lectroniques concernant les effets & agrave; long terme de la chirurgie bariatrique sur le poids confirment la robustesse des estimateurs dans des & eacute;chantillons finis. Cette approche pourrait servir de r & eacute;f & eacute;rence th & eacute;orique pour & eacute;valuer d'autres m & eacute;thodes ad-hoc.
引用
收藏
页数:18
相关论文
共 50 条
  • [1] Causal inference with confounders missing not at random
    Yang, S.
    Wang, L.
    Ding, P.
    [J]. BIOMETRIKA, 2019, 106 (04) : 875 - 888
  • [2] Nonparametric causal inference with confounders missing not at random
    Shan, Jiawei
    Yan, Xinyu
    [J]. STATISTICA NEERLANDICA, 2024,
  • [3] SEMIPARAMETRIC INFERENCE OF CAUSAL EFFECT WITH NONIGNORABLE MISSING CONFOUNDERS
    Sun, Zhaohan
    Liu, Lan
    [J]. STATISTICA SINICA, 2021, 31 (04) : 1669 - 1688
  • [4] Robust Causal Inference for Recommender System to Overcome Noisy Confounders
    Zhang, Zhiheng
    Dai, Quanyu
    Chen, Xu
    Dong, Zhenhua
    Tang, Ruiming
    [J]. PROCEEDINGS OF THE 46TH INTERNATIONAL ACM SIGIR CONFERENCE ON RESEARCH AND DEVELOPMENT IN INFORMATION RETRIEVAL, SIGIR 2023, 2023, : 2349 - 2353
  • [5] Doubly robust estimation in missing data and causal inference models
    Bang, H
    [J]. BIOMETRICS, 2005, 61 (04) : 962 - 972
  • [6] Sequential Deconfounding for Causal Inference with Unobserved Confounders
    Hatt, Tobias
    Feuerriegel, Stefan
    [J]. CAUSAL LEARNING AND REASONING, VOL 236, 2024, 236 : 934 - 956
  • [7] Robust and efficient estimation for the treatment effect in causal inference and missing data problems
    Lin, Huazhen
    Zhou, Fanyin
    Wang, Qiuxia
    Zhou, Ling
    Qin, Jing
    [J]. JOURNAL OF ECONOMETRICS, 2018, 205 (02) : 363 - 380
  • [8] Improved double-robust estimation in missing data and causal inference models
    Rotnitzky, Andrea
    Lei, Quanhong
    Sued, Mariela
    Robins, James M.
    [J]. BIOMETRIKA, 2012, 99 (02) : 439 - 456
  • [9] Commentary: Causal Inference for Social Exposures
    Kaufman, Jay S.
    [J]. ANNUAL REVIEW OF PUBLIC HEALTH, VOL 40, 2019, 40 : 7 - 21
  • [10] Causal Inference: A Missing Data Perspective
    Ding, Peng
    Li, Fan
    [J]. STATISTICAL SCIENCE, 2018, 33 (02) : 214 - 237