Missing-Data Methods for Generalized Linear Models

摘要: Missing data is a major issue in many applied problems, especially the biomedical sciences. We review four common approaches for inference generalized linear models (GLMs) with missing covariate data: maximum likelihood (ML), multiple imputation (MI), fully Bayesian (FB), and weighted estimating equations (WEEs). There considerable interest how these methodologies are related, properties of each approach, advantages disadvantages methodology, computational implementation. examine that at random nonignorable missing. For ML, we focus on techniques using EM algorithm, particular, discuss by method weights related procedures as discussed Ibrahim. MI, developed Rubin. FB, considered Ibrahim et al. WEE, Robins use real dataset detailed simulation study to compare methods.