Multiple Imputation for Missing Data: Fully Conditional Specification Versus Multivariate Normal Imputation

摘要

Statistical analysis in epidemiologic studies is often hindered by missing data, and multiple imputation is in-ncreasingly being used to handle this problem. In a simulation study, the authors compared 2methods for imputationnthat are widely available in standard software: fully conditional specification (FCS) or ‘‘chained equations’’ andnmultivariate normal imputation (MVNI). The authors created data sets of 1,000 observations to simulate a cohortnstudy, and missing data were induced under 3 missing-data mechanisms. Imputations were performed using FCSn(Royston’s ‘‘ice’’) and MVNI (Schafer’s NORM) in Stata (Stata Corporation, College Station, Texas), with trans-nformations or prediction matching being used to manage nonnormality in the continuous variables. Inferences forna set of regression parameters were compared between these approaches and a complete-case analysis. Asnexpected, both FCS and MVNI were generally less biased than complete-case analysis, and both produced similarnresults despite the presence of binary and ordinal variables that clearly did not follow a normal distribution. Ignoringnskewness in a continuous covariate led to large biases and poor coverage for the corresponding regressionnparameter under both approaches, although inferences for other parameters were largely unaffected. These re-nsults provide reassurance that similar results can be expected from FCS and MVNI in a standard regressionnanalysis involving variously scaled variables.