Optimization Models and Programs for Cell Suppression in Statistical Tables

• Activity Number: 375
• Type: Topic Contributed
• Date/Time: Thursday, August 15, 2002 : 8:30 AM to 10:20 AM
• Sponsor: Section on Government Statistics*
• Author(s): Paul Massell*+
• Affiliation(s): U.S. Census Bureau
• Keywords: cell suppression ; network optimization ; linear programming ; integer programming

Abstract

There has been much international research in recent years on the cell suppression approach to protecting a statistical table. In this paper, the author compares some of the new and old methods from both a theoretical and practical point of view. The basic theoretical difference among the approaches is the type of mathematical programming method that is used to solve the optimization problem involved. However, the usefulness of a cell suppression program often depends on features, in addition to the optimization method used; e.g., 1.) ability to handle linked tables; 2.) allowance for capacities; 3.) guarantee of adequate suppression; 4.) set of expressible cost functions; and 5.) adequate speed for major production runs. The author will compare some of the optimization methods and some of the programs that have been reported in the literature in recent years; e.g., network based programs (Jewett), extended network methods (J. Castro), linear programming based programs (Massell), integer programming based programs (Fischetti, Salazar), hypercube based programs (Giessing), and the Argus system (Hundepool).