Multiple Imputation for Data-Base Construction

D. B. Rubin
DOI: https://doi.org/10.1007/978-3-642-46900-8_53
1988-01-01
Abstract:Multiple imputation is a technique for handling missing values in shared databases. The technique replaces each missing value by a vector of possible values that reflect uncertainty about which value to impute. The resultant multiply-imputed data base consists of the original data set with missing values replaced by pointers to the rows of a supplemental matrix of multiple imputations, each row being the vector of multiple imputations for a missing value. This data base can be analyzed by standard complete-data analysis techniques to obtain valid inferences.
What problem does this paper attempt to address?