| |
Notes on sensitivity measures and protection levels
Anneke Loeve
Statistics Netherlands
Summary.
At Statistics Netherlands, a heuristic approach for cell suppression in
non-negative hierarchical tables, HiTaS, is developed. HiTaS itself makes
use of ARGUS, a cell suppression method for the protection of
non-hierarchical tables. HiTaS protects a hierarchical table in a top-down
approach by finding protection patterns for non-hierarchical sub-tables
using ARGUS. ARGUS is based on a method by Fischetti and Salazar. To use
this method, the cells which have to be protected are assigned the status
primary unsafe. For these cells,
non-negative protection levels can be chosen. These protection levels define
a protection interval around the cell value. If a cell is primary unsafe, it
will be contained in the final protection pattern, along with secondary
suppressed cells such that the value of the first cell cannot be calculated
to lie within the protection interval.
For the determination of the primary unsafe cells, some kind of safety rules
are used. Often a combination of a linear sensitivity measure and a number
rule will be employed. It would be nice if the choice of the protection
levels is related to the rules used to determine the safety of the cells. In
this note, some safety rules and their relation with the choice of
protection levels are discussed.
Keywords: HiTaS, ARGUS, hierarchical tables, cell suppression, (linear)
sensitivity measures, protection levels.
Keywords:
HiTaS, ARGUS, hierarchical tables, cell suppression, (linear)
sensitivity measures, protection levels.
Address for correspondence:
Anneke Loeve Statistics Netherlands P.O. Box 4000 2270 JM Voorburg The Netherlands Phone: +31-70-3375528
|