A note on multiple imputation for method of moments estimation

• 成果类型: Article
• 署名作者: Yang, S.; Kim, J. K.
• 署名单位: Harvard University; Harvard T.H. Chan School of Public Health; Iowa State University
• 刊物名称: BIOMETRIKA
• ISSN/ISSBN: 0006-3444
• DOI: 10.1093/biomet/asv073
• 发表日期: 2016
• 页码: 244251
• 关键词: inference

摘要

Multiple imputation is widely used for estimation in situations where there are missing data. Rubin (1987) provided an easily applicable formula for multiple imputation variance estimation, but its validity requires the congeniality condition of Meng (1994), which may not be satisfied for method of moments estimation. We give the asymptotic bias of Rubin's variance estimator when method of moments estimation is used in the complete-sample analysis for each imputed dataset. A new variance estimator based on over-imputation is proposed to provide asymptotically valid inference in this case.