|
Abstract:
|
Cell Suppression is used to avoid disclosure in Economic Census tabulations. It is a complicated process due to the complexity of the data structure, the magnitude of data size, and interactions between non-overlapping data releases. The main issues with any cell suppression methodology are undersuppression and oversuppression. Ideally we want data fully protected without compromising data quality, i.e. to publishas much as possible. The recently remodeled cell suppression program used in the 2012 Economic Census can eliminate undersuppression, providing full protection. This program also reduces oversuppression over previous applications, thus improving data quality. The modernized cell suppression uses linear programming (LP) which attemps to minimize n total cell value suppressed. It is a compromised objective, since our objectives are to control the cell suppression pattern by either minimizing (1) the total suppressed cell value; (ii) the number of suppressed cells; or (iii) a balance in between (i) and (ii). The first two objectives may be easily achieved by introducing a binary variable into the LP model. This leads to integer programming (IP) which is NP-hard an
|
