Missing data and prediction: the pattern submodel.

Although the apparent bias in imputations for missing covariates seem small, their total contribution over all individuals can be quite significant.

also in 5,695 other papers