A case study on the use of multiple imputation

Multiple imputation is a relatively new technique for dealing with missing values on items from survey data. Rather than deleting observations for which a value is missing, or assigning a single value to incomplete observations, one replaces each missing item with 2 or more values. Inferences then can be made with the complete data set. An application of multiple imputation is presented using the 1987-1988 National Survey of Families and Households. Several binary indicators of whether the respondent's elderly mother/mother-in-law is married are imputed. Descriptive statistics are then presented for the sample of adult children with an unmarried mother or mother-in-law.