Skip to main content

Table 4 Summary of the missing data methods investigated

From: Comparison of techniques for handling missing covariate data within prognostic modelling studies: a simulation study

Method Label Method Description Library used within R statistical software Number of iterations
CC Complete case analysis: Analyses only cases with complete data for all covariates   -
SI Single imputation performed using PMM 'pmm' function in 'mice' 20
MI-NORM Multiple imputation (MI) using data augmentation approach [31] with a multivariate normal assumption for all variables 'norm' [41] 100
MI-MIX MI using data augmentation approach using a general location model 'mix' [42] 100
MI-MIX-no truncating MI using data augmentation approach using a general location model, but imputed values are not truncated to within plausible range 'mix' [42] 100
MI-MICE MI using regression switching imputation [9]. Linear model are used for continuous covariates and logistic model for binary covariates and dummy variables for categorical covariates 'mice' [43] 20
MI-MICE-PMM MI using MICE with PMM 'pmm' function in 'mice' [43] 20
MI-MICE-PMM-no transformation MI using MICE with PMM without transforming the incomplete covariates 'pmm' function in 'mice' [43] 20
MI-Aregimpute MI using flexible additive imputation models [20] with PMM 'aregImpute' function in 'Hmisc' [44] 1
  1. Key: PMM = predictive mean matching; MI = multiple imputation