Online Program Home
My Program

Abstract Details

Activity Number: 527
Type: Topic Contributed
Date/Time: Wednesday, August 3, 2016 : 10:30 AM to 12:20 PM
Sponsor: Government Statistics Section
Abstract #321006 View Presentation
Title: Estimating Regression Parameters from a Sensitive Variable with Noise Multiplication
Author(s): Marlow Lemons* and Bimal Sinha and Aref Dajani
Companies: U.S. Census Bureau and University of Maryland Baltimore County and U.S. Census Bureau
Keywords: noise multiplication ; disclosure avoidance ; regression
Abstract:

Publically-released data users often experience the challenge of computing regression estimates that are similar to those they would have attained were they able to access the internal data. When a variable is binned for being too sensitive for release, the task becomes more critical. This study investigates how releasing frequency data for a sensitive variable and unbinned data for a non-sensitive covariate affect the likelihood of estimating regression parameters that do not substantially differ from those obtained from the internal data. Simulations will determine whether the likelihood is impacted by the number of bins, the underlying distribution of the sensitive variable, and whether noise multiplication is introduced. An application to economic survey data follows.


Authors who are presenting talks have a * after their name.

Back to the full JSM 2016 program

 
 
Copyright © American Statistical Association