Online Program

Return to main conference page

All Times ET

Program is Subject to Change

Wednesday, June 16
Wed, Jun 16, 1:30 PM - 3:30 PM
TBD
Statistical Confidentiality and Establishment Surveys: Challenges and Solutions to Improving Data Utility

Construction of a Feasible, Complete Table from a Published Table with Suppressions (307975)

Philip Steel, US Census Bureau 
*Bei Wang, Census Bureau 

Keywords: cell suppression, disclosure, bound

One of the objections to cell suppression is that it “puts holes in the data”. This can be a somewhat disingenuous characterization, especially when it is followed by a proposal to produce complete data that distorts rather than suppresses data. When a D appears in a table the reported data is indeed suppressed, however much information about the cell is retained in the table. In particular, the missing cell has bounds and those bounds may be fairly tight. A more legitimate criticism is that the information on the cell still contained in the table has not been presented and no estimate of the missing cell has been made -- that task has been left to a possibly ill equipped user. We describe an extension of cell suppression which produces a complete, additive table, and presents safe values as they are obtained and feasible values where the actual data has been suppressed. The substituted values are marked as such and bounds are provided. We apply this to a variety of tables and test the scalability of the technique. We look at conditioning on prior data. We test some analyses on these data to determine what limitations can be ascertained.