A comparison of multiple imputation and fully augmented weighted estimators for Cox regression with missing covariates

Lihong Qi, Ying Fang Wang, Yulei He

Research output: Contribution to journalArticle

17 Citations (Scopus)

Abstract

Several approaches exist for handling missing covariates in the Cox proportional hazards model. The multiple imputation (MI) is relatively easy to implement with various software available and results in consistent estimates if the imputation model is correct. On the other hand, the fully augmented weighted estimators (FAWEs) recover a substantial proportion of the efficiency and have the doubly robust property. In this paper, we compare the FAWEs and the MI through a comprehensive simulation study. For the MI, we consider the multiple imputation by chained equation and focus on two imputation methods: Bayesian linear regression imputation and predictive mean matching. Simulation results show that the imputation methods can be rather sensitive to model misspecification and may have large bias when the censoring time depends on the missing covariates. In contrast, the FAWEs allow the censoring time to depend on the missing covariates and are remarkably robust as long as getting either the conditional expectations or the selection probability correct due to the doubly robust property. The comparison suggests that the FAWEs show the potential for being a competitive and attractive tool for tackling the analysis of survival data with missing covariates.

Original languageEnglish (US)
Pages (from-to)2592-2604
Number of pages13
JournalStatistics in Medicine
Volume29
Issue number25
DOIs
StatePublished - Nov 2010

Fingerprint

Missing Covariates
Cox Regression
Multiple Imputation
Imputation
Estimator
Bayes Theorem
Censoring
Survival Analysis
Proportional Hazards Models
Linear Models
Software
Cox Proportional Hazards Model
Consistent Estimates
Model Misspecification
Survival Data
Conditional Expectation
Bayesian Methods
Linear regression
Proportion
Simulation Study

Keywords

  • Accelerated failure time model
  • Augmented inverse probability weighted estimators
  • Doubly robust property
  • Missing data
  • Proportional hazards model
  • Survival analysis

ASJC Scopus subject areas

  • Epidemiology
  • Statistics and Probability

Cite this

A comparison of multiple imputation and fully augmented weighted estimators for Cox regression with missing covariates. / Qi, Lihong; Wang, Ying Fang; He, Yulei.

In: Statistics in Medicine, Vol. 29, No. 25, 11.2010, p. 2592-2604.

Research output: Contribution to journalArticle

@article{bbced16a1d84412fa0977716a5c367ce,
title = "A comparison of multiple imputation and fully augmented weighted estimators for Cox regression with missing covariates",
abstract = "Several approaches exist for handling missing covariates in the Cox proportional hazards model. The multiple imputation (MI) is relatively easy to implement with various software available and results in consistent estimates if the imputation model is correct. On the other hand, the fully augmented weighted estimators (FAWEs) recover a substantial proportion of the efficiency and have the doubly robust property. In this paper, we compare the FAWEs and the MI through a comprehensive simulation study. For the MI, we consider the multiple imputation by chained equation and focus on two imputation methods: Bayesian linear regression imputation and predictive mean matching. Simulation results show that the imputation methods can be rather sensitive to model misspecification and may have large bias when the censoring time depends on the missing covariates. In contrast, the FAWEs allow the censoring time to depend on the missing covariates and are remarkably robust as long as getting either the conditional expectations or the selection probability correct due to the doubly robust property. The comparison suggests that the FAWEs show the potential for being a competitive and attractive tool for tackling the analysis of survival data with missing covariates.",
keywords = "Accelerated failure time model, Augmented inverse probability weighted estimators, Doubly robust property, Missing data, Proportional hazards model, Survival analysis",
author = "Lihong Qi and Wang, {Ying Fang} and Yulei He",
year = "2010",
month = "11",
doi = "10.1002/sim.4016",
language = "English (US)",
volume = "29",
pages = "2592--2604",
journal = "Statistics in Medicine",
issn = "0277-6715",
publisher = "John Wiley and Sons Ltd",
number = "25",

}

TY - JOUR

T1 - A comparison of multiple imputation and fully augmented weighted estimators for Cox regression with missing covariates

AU - Qi, Lihong

AU - Wang, Ying Fang

AU - He, Yulei

PY - 2010/11

Y1 - 2010/11

N2 - Several approaches exist for handling missing covariates in the Cox proportional hazards model. The multiple imputation (MI) is relatively easy to implement with various software available and results in consistent estimates if the imputation model is correct. On the other hand, the fully augmented weighted estimators (FAWEs) recover a substantial proportion of the efficiency and have the doubly robust property. In this paper, we compare the FAWEs and the MI through a comprehensive simulation study. For the MI, we consider the multiple imputation by chained equation and focus on two imputation methods: Bayesian linear regression imputation and predictive mean matching. Simulation results show that the imputation methods can be rather sensitive to model misspecification and may have large bias when the censoring time depends on the missing covariates. In contrast, the FAWEs allow the censoring time to depend on the missing covariates and are remarkably robust as long as getting either the conditional expectations or the selection probability correct due to the doubly robust property. The comparison suggests that the FAWEs show the potential for being a competitive and attractive tool for tackling the analysis of survival data with missing covariates.

AB - Several approaches exist for handling missing covariates in the Cox proportional hazards model. The multiple imputation (MI) is relatively easy to implement with various software available and results in consistent estimates if the imputation model is correct. On the other hand, the fully augmented weighted estimators (FAWEs) recover a substantial proportion of the efficiency and have the doubly robust property. In this paper, we compare the FAWEs and the MI through a comprehensive simulation study. For the MI, we consider the multiple imputation by chained equation and focus on two imputation methods: Bayesian linear regression imputation and predictive mean matching. Simulation results show that the imputation methods can be rather sensitive to model misspecification and may have large bias when the censoring time depends on the missing covariates. In contrast, the FAWEs allow the censoring time to depend on the missing covariates and are remarkably robust as long as getting either the conditional expectations or the selection probability correct due to the doubly robust property. The comparison suggests that the FAWEs show the potential for being a competitive and attractive tool for tackling the analysis of survival data with missing covariates.

KW - Accelerated failure time model

KW - Augmented inverse probability weighted estimators

KW - Doubly robust property

KW - Missing data

KW - Proportional hazards model

KW - Survival analysis

UR - http://www.scopus.com/inward/record.url?scp=77958605331&partnerID=8YFLogxK

UR - http://www.scopus.com/inward/citedby.url?scp=77958605331&partnerID=8YFLogxK

U2 - 10.1002/sim.4016

DO - 10.1002/sim.4016

M3 - Article

C2 - 20806403

AN - SCOPUS:77958605331

VL - 29

SP - 2592

EP - 2604

JO - Statistics in Medicine

JF - Statistics in Medicine

SN - 0277-6715

IS - 25

ER -