|
Activity Number:
|
418
|
|
Type:
|
Invited
|
|
Date/Time:
|
Thursday, August 12, 2004 : 10:30 AM to 12:20 PM
|
|
Sponsor:
|
National Institute of Statistical Sciences
|
| Abstract - #300030 |
|
Title:
|
Inference-based Measures of Utility for Contingency Tables
|
|
Author(s):
|
Ashish Sanil*+
|
|
Companies:
|
National Institute of Statistical Sciences
|
|
Address:
|
, , AZ, ,
|
|
Keywords:
|
|
|
Abstract:
|
The process of disseminating cross-tabulated data derived from confidential microdata involves a trade-off between protecting confidentiality while releasing as much useful information as possible. Over the years, a number of procedures have been devised that allow disclosure-protected dissemination of tabular data. These methods employ various well-established criteria to ascertain risk, and also various ways of measuring utility of the release based on relatively less rigorous notions of information loss. We propose an approach to measure the utlity of a released table (or a set of tables) based on the notion of "statistical usefulness" of the data to the end user who is interested in statistical analyses of the data (i.e., understanding the structure of associations among attributes of the dataset). This family of utility measures is founded on the extent to which the statistical properties of the released data agree with those of the original dataset. We will illustrate how a risk-utility formulation based on the inference-based measures of utility with examples.
|
- 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 2004 program |
