A note on overadjustment in inverse probability weighted estimation

Andrea Rotnitzky, Lingling Li, Xiaochun Li

Research output: Contribution to journalArticle

12 Citations (Scopus)

Abstract

Standardized means, commonly used in observational studies in epidemiology to adjust for potential confounders, are equal to inverse probability weighted means with inverse weights equal to the empirical propensity scores. More refined standardization corresponds with empirical propensity scores computed under more flexible models. Unnecessary standardization induces efficiency loss. However, according to the theory of inverse probability weighted estimation, propensity scores estimated under more flexible models induce improvement in the precision of inverse probability weighted means. This apparent contradiction is clarified by explicitly stating the assumptions under which the improvement in precision is attained.

Original languageEnglish
Pages (from-to)997-1001
Number of pages5
JournalBiometrika
Volume97
Issue number4
DOIs
StatePublished - Dec 2010

Fingerprint

Propensity Score
standardization
Standardization
Weighted Mean
Probability Theory
Epidemiology
observational studies
Observational Study
Observational Studies
epidemiology
Weights and Measures
Propensity score
Model

Keywords

  • Causal inference
  • Propensity score
  • Standardized mean

ASJC Scopus subject areas

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

Cite this

A note on overadjustment in inverse probability weighted estimation. / Rotnitzky, Andrea; Li, Lingling; Li, Xiaochun.

In: Biometrika, Vol. 97, No. 4, 12.2010, p. 997-1001.

Research output: Contribution to journalArticle

Rotnitzky, Andrea ; Li, Lingling ; Li, Xiaochun. / A note on overadjustment in inverse probability weighted estimation. In: Biometrika. 2010 ; Vol. 97, No. 4. pp. 997-1001.
@article{7fbe626ec075464ea85e953b60aee285,
title = "A note on overadjustment in inverse probability weighted estimation",
abstract = "Standardized means, commonly used in observational studies in epidemiology to adjust for potential confounders, are equal to inverse probability weighted means with inverse weights equal to the empirical propensity scores. More refined standardization corresponds with empirical propensity scores computed under more flexible models. Unnecessary standardization induces efficiency loss. However, according to the theory of inverse probability weighted estimation, propensity scores estimated under more flexible models induce improvement in the precision of inverse probability weighted means. This apparent contradiction is clarified by explicitly stating the assumptions under which the improvement in precision is attained.",
keywords = "Causal inference, Propensity score, Standardized mean",
author = "Andrea Rotnitzky and Lingling Li and Xiaochun Li",
year = "2010",
month = "12",
doi = "10.1093/biomet/asq049",
language = "English",
volume = "97",
pages = "997--1001",
journal = "Biometrika",
issn = "0006-3444",
publisher = "Oxford University Press",
number = "4",

}

TY - JOUR

T1 - A note on overadjustment in inverse probability weighted estimation

AU - Rotnitzky, Andrea

AU - Li, Lingling

AU - Li, Xiaochun

PY - 2010/12

Y1 - 2010/12

N2 - Standardized means, commonly used in observational studies in epidemiology to adjust for potential confounders, are equal to inverse probability weighted means with inverse weights equal to the empirical propensity scores. More refined standardization corresponds with empirical propensity scores computed under more flexible models. Unnecessary standardization induces efficiency loss. However, according to the theory of inverse probability weighted estimation, propensity scores estimated under more flexible models induce improvement in the precision of inverse probability weighted means. This apparent contradiction is clarified by explicitly stating the assumptions under which the improvement in precision is attained.

AB - Standardized means, commonly used in observational studies in epidemiology to adjust for potential confounders, are equal to inverse probability weighted means with inverse weights equal to the empirical propensity scores. More refined standardization corresponds with empirical propensity scores computed under more flexible models. Unnecessary standardization induces efficiency loss. However, according to the theory of inverse probability weighted estimation, propensity scores estimated under more flexible models induce improvement in the precision of inverse probability weighted means. This apparent contradiction is clarified by explicitly stating the assumptions under which the improvement in precision is attained.

KW - Causal inference

KW - Propensity score

KW - Standardized mean

UR - http://www.scopus.com/inward/record.url?scp=78651325910&partnerID=8YFLogxK

UR - http://www.scopus.com/inward/citedby.url?scp=78651325910&partnerID=8YFLogxK

U2 - 10.1093/biomet/asq049

DO - 10.1093/biomet/asq049

M3 - Article

VL - 97

SP - 997

EP - 1001

JO - Biometrika

JF - Biometrika

SN - 0006-3444

IS - 4

ER -