A Simpler Standard Error Formula for Multiple Imputation Based on a Maximum Likelihood Estimate
Year: 2012
Type: preprint
Abstract: Maximum likelihood (ML) imputation is a form of multiple imputation (MI) that draws random imputations conditionally on a maximum likelihood estimate. ML imputation is simpler and more efficient than the more popular Bayesian approach to MI which draws imputations from the posterior predictive distribution. A disadvantage of ML imputation has been the difficulty of estimating the standard error of the ML imputation point estimate. This paper derives a simple new standard error formula, then tests it through simulation. The formula performs well if the fraction of missing information is small to moderate, but has a negative bias if the fraction of missing information is large.
Source: arXiv (Cornell University)
Authors: von Hippel, Thomas S. H. Paul
Language: English
Subfield: Statistics and Probability
Field: Mathematics
Domain: Physical Sciences
Open Access status: green
OpenAlex ID: W1653326011