Multiple Imputation for Missing Data

Two algorithms for producing multiple imputations for missing data are evaluated with simulated data, using a propensity score classifier with the approximate Bayesian bootstrap and a regression-based method employing the data augmentation algorithm.

Abstract

Two algorithms for producing multiple imputations for missing data are evaluated with simulated data. Software using a propensity score classifier with the approximate Bayesian bootstrap produces badly biased estimates of regression coefficients when data on predictor variables are missing at random or missing completely at random. On the other hand, a regression-based method employing the data augmentation algorithm produces estimates with little or no bias.