346 – Disclosure, Confidentiality, Privacy
Measuring Risk in Tables Where the Study Variable May Be Negative
Ann-Marie Flygare
Statistics Sweden
Ingegerd Jansson
Statistics Sweden
Tiina Orusild
Statistics Sweden
Disclosure control of tables where the study variable, and thus the cell values, may take negative values poses a particular problem. The common sensitivity rules like the dominance rule and the p% rule do not apply. To solve this, it has been suggested to transform cell values (e.g. add a constant or take absolute values) in order to make all values positive and facilitate the use of the common sensitivity rules. With this approach, it is assumed that the risk scenario for variables that may take negative values is similar to variables that only take positive values. We use an empirical study to illustrate how the common sensitivity rules perform in different situations and we initiate a discussion of the sensitivity of data that may take negative values and discuss the need for a different approach to determining the sensitive values. Ideas for modified measures of risk are presented.