Table of contentsPreserving Confidentiality AND Providing Adequate Data for Statistical Modeling
Overview
R-U Confidentiality Map
NISS Prototype Query System
Confidentiality Concern
Fundamental Abstractions
Why Marginals?
Example 1: 2000 Census
PowerPoint Presentation
Example 2: Risk Factors for Coronary Heart Disease
Example 2: The Data
Example 3: NLTCS
Two-Way Fréchet Bounds
Bounds for Multi-Way Tables
Role of Log-linear Models?
Graphical & Decomposable Log-linear Models
MLEs for Decomposable Log-linear Models
Multi-way Bounds
Multi-Way Bounds (cont.)
Ex. 2: Czech Autoworkers
Bounds for [BF][ABCE][ADE]
Example 2 (cont.)
More on Bounds
Example 2: Release of All 5-way Margins
Example 3: NLTCS
NLTCS Search Results
Perturbation Maintaining Marginal Totals
Perturbation for Protection
Some Ongoing Research
Summary
The End
Stochastic Perturbation Methods
Exact Distribution of Table Given Marginals
Markov Basis “Moves”
Three-way Illustration (k=3)
NISS Table Server: 6-Way Table
|