Activity Number:
|
256
|
Type:
|
Contributed
|
Date/Time:
|
Tuesday, August 13, 2002 : 2:00 PM to 3:50 PM
|
Sponsor:
|
Section on Survey Research Methods*
|
Abstract - #301386 |
Title:
|
Multiple Imputation and Statistical Disclosure Control in Microdata
|
Author(s):
|
Fang Liu*+ and Roderick Little
|
Affiliation(s):
|
University of Michigan, Ann Arbor and University of Michigan
|
Address:
|
1420 Washtington Heights, SPH II, M4011, Ann Arbor, Michigan, 48109, USA
|
Keywords:
|
disclosure risk ; statistical disclosure control(SDC) ; multiple imputation(MI) ; general location model ; information loss ; protection
|
Abstract:
|
The fundamental tension in statistcal disclosure control (SDC) in microdata is the trade-off between the protection of individual respondents and the release of enough information for statistical inferences. We consider microdata that include key variables that contain identification information and continuous target variables that include sensitive information. Releasing the original data may expose some individuals in the sample to high risk of disclosure; deleting key variables is a common approach, but this loses information for analysis. This talk describes the use of multiple imputation (MI) (Rubin, 1987) of key variables as an alternative SDC technique between those two extremes. Key variables are treated as categorical, and the general location model is employed to independently impute multiple values for key variables from their posterior predictive distributions. Performance of MI as an SDC technique will be evaluated from the viewpoints of both information loss and protection; simulation results and application of the method to Alameda County Survey data will be presented and issues, such as which cases should be imputed, will also be briefly discussed.
|
- The address information is for the authors that have a + after their name.
- Authors who are presenting talks have a * after their name.
Back to the full JSM 2002 program |