|
Activity Number:
|
468
|
|
Type:
|
Topic Contributed
|
|
Date/Time:
|
Thursday, August 7, 2008 : 8:30 AM to 10:20 AM
|
|
Sponsor:
|
Section on Government Statistics
|
| Abstract - #301018 |
|
Title:
|
Producing Partially Synthetic Data to Avoid Disclosure
|
|
Author(s):
|
Sam Hawala*+
|
|
Companies:
|
U.S. Census Bureau
|
|
Address:
|
4700 Silver Hill Road, Washington D.C., MD, 20233,
|
|
Keywords:
|
Disclosure ; Synthetic Data
|
|
Abstract:
|
This paper discusses the combined use of model fitting to data and matching using the model predicted values to reproduce the aggregate behavior and the main features of a data set. We approximate the data using semiparametric regression models combining a simple additive structure with the flexibility of the nonparametric approach. Then we use the model predicted values to simulate new data to preserve confidentiality. In the case of highly sensitive data this method provides the necessary protection from disclosure.
|
