论文信息 - When Are Inferences from Multiple Imputation Valid ?

When Are Inferences from Multiple Imputation Valid ?

Multiple imputation, as described by Rubin, has seen a wide variety of applications. Counterexamples, presented by Fay (1991), and new methods, such as those of J.N.K. Rao and J. Shao, that can asymptotically disagree with the multiple imputation approach, have raised questions about the validity of multiple imputation. This paper identifies critical restrictions on the practical application of multiple imputation. It also discusses alternatives that can provide asymptotically valid inferences for some of the situations in which multiple imputation fails.

R. Fay

[1] D. Rubin,et al. MULTIPLE IMPUTATIONS IN SAMPLE SURVEYS-A PHENOMENOLOGICAL BAYESIAN APPROACH TO NONRESPONSE , 2002 .

[2] J. Shao,et al. Jackknife variance estimation with survey data under hot deck imputation , 1992 .

[3] Donald B. Rubin,et al. Multiple Imputation of Industry and Occupation Codes in Census Public-use Samples Using Bayesian Logistic Regression , 1991 .

[4] Joseph L Schafer,et al. Variance estimation with imputed means , 1991 .

[5] Nathaniel Schenker,et al. Asymptotic results for multiple imputation , 1988 .

[6] D. Rubin,et al. Multiple Imputation for Interval Estimation from Simple Random Samples with Ignorable Nonresponse , 1986 .