Exact Balanced Random Imputation

• Activity Number: 580
• Type: Invited
• Date/Time: Wednesday, August 12, 2015 : 2:00 PM to 3:50 PM
• Sponsor: Survey Research Methods Section
• Author(s): Guillaume Chauvet* and Wilfried Do Paco
• Companies: ENSAI (CREST) and INSEE
• Keywords: balanced imputation ; cube method ; imputation model ; mean-square consistency ; regularized estimator

Abstract

Surveys usually suffer from non-response, which decreases the effective sample size. Some form of random imputation is typically used to handle non-response if we wish to preserve the distribution of the imputed variable. A possible drawback lies in an increased variability due to the imputation variance. Several approaches have been proposed for reducing the imputation variance. One of them consists in using some form of balanced imputation, where donors or residuals are selected at random in such a way that the imputation variance is eliminated (Kalton and Kish, 1981, 1984; Deville, 2006; Chauvet, Deville and Haziza, 2011).

In this work, we propose an implementation of balanced random imputation which enables to fully eliminate the imputation variance for parameters. Also, we study the mean square consistency of the imputed estimator for the total, and the mean square consistency of the imputed estimator for the distribution function. Some simulation results support our findings. This is joint work with Wilfried Do Paco (INSEE).