|
Activity Number:
|
518
|
|
Type:
|
Invited
|
|
Date/Time:
|
Thursday, August 10, 2006 : 10:30 AM to 12:20 PM
|
|
Sponsor:
|
Business and Economics Statistics Section
|
| Abstract - #305151 |
|
Title:
|
Combining Synthetic Data and Noise Infusion for Confidentiality Protection of the Quarterly Workforce Indicators
|
|
Author(s):
|
John Abowd*+ and Lars Vilhuber
|
|
Companies:
|
Cornell University and Cornell University
|
|
Address:
|
CISER, Ithaca, NY, 14850,
|
|
Keywords:
|
confidentiality ; noise infusion ; synthetic data ; quarterly workforce indicators ; analytic validity
|
|
Abstract:
|
The Quarterly Workforce Indicators (QWI) statistics currently are protected by a combination of noise infusion and suppression. Dynamically consistent noise infusion is used for all measures. Suppression is used for counts from sparsely populated cells. The Longitudinal Household Employer Dynamics (LHED) Program, which produces the QWIs, has experimented with replacing the suppressed QWIs with synthetic values. The synthetic values are generated by sampling from the posterior predictive distribution of the indicator, given its history and the rules that cause the suppression. The synthesis is done from the underlying confidential QWI data, not from the released data. The use of synthetic data in this application improves the analytic validity of the QWIs without compromising the protection in the noise infusion system.
|
