Adjustments for Survey Imputed Data Sets to Achieve First- and Second-Order Properties

• Activity Number: 41
• Type: Contributed
• Date/Time: Sunday, August 3, 2014 : 2:00 PM to 3:50 PM
• Sponsor: Survey Research Methods Section
• Author(s): Damiao Da Silva*+ and Lichun Zhang
• Companies: Universidade Federal do Rio Grande do Norte and University of Southampton
• Keywords: Complex surveys ; Missing data ; Survey nonresponse ; Reverse calibration ; Secondary data analysis ; Hot deck imputation

Abstract

For practical convenience reasons, survey datasets are often disseminated to both internal and external secondary users with imputed values in place of the missing observations that occur during the data collection and processing stages. However, the released dataset generally leads to incorrect analyses if standard complete-data methods are applied directly without taking into account the imputation models. To alleviate this problem the survey statistician may, firstly, obtain some key results in any manner that is considered appropriate, and, secondly, calibrate the disseminated data so that these results can be reproduced by relevant standard complete-data procedures. In this paper, we discuss an approach that allows us to control both the first- and second-order properties of the imputed data and can be applied to complex surveys. We illustrate the implementation of the proposed approach with a numerical example.