Multiple imputation (MI) is actually somewhat of a misnomer. The phrase is best understood as the name for a post-imputation variance estimation tool that involves repetitions of the imputation process. The father of multiple imputation, Donald Rubin, originally envisioned MI as a tool for the preparation of public use files (PUFs). He advocated that data publishers use MI in order to simplify and improve the analyses conducted by PUF consumers. So far, few data publishers have adopted MI. More usage of MI has been found in highly multivariate analyses with complex missing data structures, such as in the scoring of standardized tests with adaptive item sampling. In that literature, the multiple imputations are most often referred to as plausible values.

Motivation

MI is most commonly used in ...

  • Loading...
locked icon

Sign in to access this content

Get a 30 day FREE TRIAL

  • Watch videos from a variety of sources bringing classroom topics to life
  • Read modern, diverse business cases
  • Explore hundreds of books and reference titles