Digital Government:

Project Update - June, 2000

1. GEOGRAPHICAL AGGREGATION: Code has been turned over to MCNC for development of a prototype using NASS data. Database design and coding are substantially complete. The user interface has been designed, and is being implemented. Target date for completion of the prototype remains 8/31/00.

2. TABLE AGGREGATION. Code to make tables disclosable by aggregating "adjacent" categories in various dimensions continues to be developed by NISS postdoc Ashish Sanil. Initial versions are being explored.

3. STATISTICAL CONSEQUENCES OF AGGREGATION. A detailed study is in progress, using direct simulation, analytical tools and a Bayesian approach coupled with Markov chain Monte Carlo (MCMC). NISS postdoc Jaeyong Lee and intern Chris Holloman are playing key roles.

4. TABLE SERVERS. Under the leadership of summer intern Karen Brady, a systematic survey and analysis are underway of methods for assessing disclosure risk associated with queries for cross-tabulations from a large contingency table. Special emphasis is being directed to dynamics and scalability, with the goal of selecting methods for the initial version of the table server. Methods under study include cell bounds, simulation, iterative proportional fitting and cell suppression.

5. BOUNDS FOR TABLE ENTRIES. A technical report on research at CMU on scalable methods to compute bounds on entries in large tables will be available on the project Web site (www.niss.org/dg) shortly. Stephen Fienberg and Adrian Dobro are the principals in this effort.

6. RISK VS. INFORMATIVENESS. Research is continuing at CMU/LANL, led by George Duncan and Sallie Keller-McNulty, on problem formulations that accommodate both disclosure risk and informativeness of releasing information.

7. PAPERS. Two papers are under preparation at NISS. The first, targeted for the February, 2001, special issue of IEEE Computer, describes the geographic aggregation system developed for NASS (item 1 above). The second, to appear in a statistics journal, addresses the statistical consequences of aggregation (item 3 above).