Multiple imputation of covariates by fully conditional specification: accommodating the substantive model

Abstract

Missing covariate data commonly occur in epidemiological and clinical research, and are often dealt with using multiple imputation (MI). Imputation of partially observed covariates is complicated if the substantive model is non-linear (e.g. Cox proportional hazards model), or contains non-linear (e.g. squared) or interaction terms, and standard software implementations of MI may impute covariates from models that are incompatible with such substantive models. We show how imputation by fully conditional specification, a popular approach for performing MI, can be modified so that covariates are imputed from models which are compatible with the substantive model. We investigate through simulation the performance of this proposal, and compare it to existing approaches. Simulation results suggest our proposal gives consistent estimates for a range of common substantive models, including models which contain non-linear covariate effects or interactions, provided data are missing at random and the assumed imputation models are correctly specified and mutually compatible.

In the original version we defined the concept of congeniality differently from Meng (1994). In this revised version we refrain from using the term congeniality, and instead refer only to compatibility between imputation and substantive models. In the discussion we explain how imputation and substantive model compatibility and congeniality are related

Cited by in corpus (19)

• Multiple imputation for multilevel data with continuous and binary variables
• Multiple imputation of multilevel missing data: An introduction to the R package pan
• Multiple imputation of missing covariate values in multilevel models with random slopes: A cautionary note
• Bayesian correction for covariate measurement error: a frequentist evaluation and comparison with regression calibration
• Multiple imputation in Cox regression when there are time-varying effects of exposures
• Modern Multiple Imputation with Functional Data
• Random Forest Missing Data Algorithms
• JointAI: Joint Analysis and Imputation of Incomplete Data in R
• Assessing treatment effects in observational data with missing confounders: A comparative study of practical doubly-robust and traditional missing data methods
• Multiple imputation of missing covariates when using the Fine-Gray model
• Multiple Imputation Method for High-Dimensional Neuroimaging Data
• Joint distribution properties of Fully Conditional Specification under the normal linear model with normal inverse-gamma priors
• Graphical and numerical diagnostic tools to assess multiple imputation models by posterior predictive checking
• Propensity score analysis with partially observed confounders: how should multiple imputation be used?
• Semiparametric fractional imputation using Gaussian mixture models for handling multivariate missing data
• MIMCA: Multiple imputation for categorical variables with multiple correspondence analysis
• Calibration of prediction rules for life-time outcomes using prognostic Cox regression survival models and multiple imputations to account for missing predictor data with cross-validatory assessment
• Leveraging Two-Phase Data for Improved Prediction of Survival Outcomes with Application to Nasopharyngeal Cancer
• Model Determination for High-Dimensional Longitudinal Data with Missing Observations: An Application to Microfinance Data