Bounded, efficient and doubly robust estimation with inverse weighting

Research output: Contribution to journalArticle

112 Citations (Scopus)

Abstract

Consider estimating the mean of an outcome in the presence of missing data or estimating population average treatment effects in causal inference. A doubly robust estimator remains consistent if an outcome regression model or a propensity score model is correctly specified. We build on a previous nonparametric likelihood approach and propose new doubly robust estimators, which have desirable properties in efficiency if the propensity score model is correctly specified, and in boundedness even if the inverse probability weights are highly variable. We compare the new and existing estimators in a simulation study and find that the robustified likelihood estimators yield overall the smallest mean squared errors.

Original languageEnglish (US)
Pages (from-to)661-682
Number of pages22
JournalBiometrika
Volume97
Issue number3
DOIs
StatePublished - Sep 1 2010

Fingerprint

Propensity Score
Robust Estimators
Robust Estimation
Weighting
Nonparametric Likelihood
Average Treatment Effect
Estimator
Causal Inference
Missing Data
Mean Squared Error
Boundedness
Likelihood
Regression Model
Simulation Study
Weights and Measures
Model
Population
Robust estimation
Propensity score
Robust estimators

All Science Journal Classification (ASJC) codes

  • Statistics and Probability
  • Mathematics(all)
  • Agricultural and Biological Sciences (miscellaneous)
  • Agricultural and Biological Sciences(all)
  • Statistics, Probability and Uncertainty
  • Applied Mathematics

Keywords

  • Causal inference
  • Double robustness
  • Inverse weighting
  • Missing data
  • Nonparametric likelihood
  • Propensity score

Cite this

@article{6399360985044f328e8059d7dd4b0378,
title = "Bounded, efficient and doubly robust estimation with inverse weighting",
abstract = "Consider estimating the mean of an outcome in the presence of missing data or estimating population average treatment effects in causal inference. A doubly robust estimator remains consistent if an outcome regression model or a propensity score model is correctly specified. We build on a previous nonparametric likelihood approach and propose new doubly robust estimators, which have desirable properties in efficiency if the propensity score model is correctly specified, and in boundedness even if the inverse probability weights are highly variable. We compare the new and existing estimators in a simulation study and find that the robustified likelihood estimators yield overall the smallest mean squared errors.",
keywords = "Causal inference, Double robustness, Inverse weighting, Missing data, Nonparametric likelihood, Propensity score",
author = "Zhiqiang Tan",
year = "2010",
month = "9",
day = "1",
doi = "10.1093/biomet/asq035",
language = "English (US)",
volume = "97",
pages = "661--682",
journal = "Biometrika",
issn = "0006-3444",
publisher = "Oxford University Press",
number = "3",

}

Bounded, efficient and doubly robust estimation with inverse weighting. / Tan, Zhiqiang.

In: Biometrika, Vol. 97, No. 3, 01.09.2010, p. 661-682.

Research output: Contribution to journalArticle

TY - JOUR

T1 - Bounded, efficient and doubly robust estimation with inverse weighting

AU - Tan, Zhiqiang

PY - 2010/9/1

Y1 - 2010/9/1

N2 - Consider estimating the mean of an outcome in the presence of missing data or estimating population average treatment effects in causal inference. A doubly robust estimator remains consistent if an outcome regression model or a propensity score model is correctly specified. We build on a previous nonparametric likelihood approach and propose new doubly robust estimators, which have desirable properties in efficiency if the propensity score model is correctly specified, and in boundedness even if the inverse probability weights are highly variable. We compare the new and existing estimators in a simulation study and find that the robustified likelihood estimators yield overall the smallest mean squared errors.

AB - Consider estimating the mean of an outcome in the presence of missing data or estimating population average treatment effects in causal inference. A doubly robust estimator remains consistent if an outcome regression model or a propensity score model is correctly specified. We build on a previous nonparametric likelihood approach and propose new doubly robust estimators, which have desirable properties in efficiency if the propensity score model is correctly specified, and in boundedness even if the inverse probability weights are highly variable. We compare the new and existing estimators in a simulation study and find that the robustified likelihood estimators yield overall the smallest mean squared errors.

KW - Causal inference

KW - Double robustness

KW - Inverse weighting

KW - Missing data

KW - Nonparametric likelihood

KW - Propensity score

UR - http://www.scopus.com/inward/record.url?scp=77955858058&partnerID=8YFLogxK

UR - http://www.scopus.com/inward/citedby.url?scp=77955858058&partnerID=8YFLogxK

U2 - 10.1093/biomet/asq035

DO - 10.1093/biomet/asq035

M3 - Article

AN - SCOPUS:77955858058

VL - 97

SP - 661

EP - 682

JO - Biometrika

JF - Biometrika

SN - 0006-3444

IS - 3

ER -