Peter-Paul de Wolf

HiTaS: A Heuristic Approach to Cell Suppression in Hierarchical Tables

This paper describes a heuristic approach to find suppression patterns in tables that exhibit a hierarchical structure in at least one of the explanatory variables. The hierarchical structure implies that there exist (many) sub-totals, i.e., that (many) sub-tables can be constructed. These sub-tables should be protected in such a way that they cannot be used to undo the protection of any of the other tables. The proposed heuristic approach has a top-down structure: when a table of high level (sub-)totals is suppressed, its interior turns into the marginals of possibly several tables on a lower level. These lower level tables are then protected while keeping the marginals fixed.

Risk, utility and PRAM

PRAM (Post Randomization Method) is a disclosure control method for microdata, introduced in 1997. Unfortunately, PRAM has not yet been applied extensively by statistical agencies in protecting their microdata. This is partly due to the fact that little knowledge is available on the effect of PRAM on disclosure control as well as on the loss of information it induces. In this paper, we will try to make up for this lack of knowledge, by supplying some empirical information on the behaviour of PRAM. To be able to achieve this, some basic measures for loss of information and disclosure risk will be introduced. PRAM will be applied to one specific microdata file of over 6 million records, using several models in applying the procedure.

Adjusting the τ -argus modular approach to deal with linked tables

The software package @t-argus offers a very efficient algorithm for secondary cell suppression known as either HiTaS or the modular approach. The method is well suited for the protection of up to 3-dimensional hierarchical tables. In practice, statistical agencies release multiple tabulations based on the same dataset. Usually these tables are linked through certain linear constraints. In such a case cell suppressions must obviously be coordinated between tables. In this paper we investigate into the possibilities for an extension of the modular approach to deal with linked tables.

How to Make the T -ARGUS Modular Method Applicable to Linked Tables

The software package T-ARGUS offers a very efficient algorithm for secondary cell suppression known as either HiTaS or the Modular approach. The method is well suited for the protection of up to 3-dimensional hierarchical tables. In practice, statistical agencies release multiple tabulations based on the same dataset. Usually these tables are linked through certain linear constraints. In such a case cell suppressions must obviously be coordinated between tables. In this paper we investigate into the possibilities for an extension of the modular approach to deal with linked tables.

Spatial Smoothing and Statistical Disclosure Control

Cartographic maps have many practical uses and can be an attractive alternative for disseminating detailed frequency tables. However, a detailed map may disclose private data of individual units of a population. We will describe some smoothing algorithms to display spatial distribution patterns. In certain situations, the disclosure risk of a spatial distribution pattern, can be formulated in terms of a frequency table disclosure problem. In this paper we will explore the effects of spatial smoothing related to statistical disclosure control.

Safely Plotting Continuous Variables on a Map

A plotted spatial distribution of a variable is an interesting type of statistical output favored by many users. Examples include the spatial distribution of people that make use of child care, of the amount of electricity used by businesses or of the exhaust of certain gasses by industry. However, a spatial distribution plot may be exploited to link information to a single unit of interest. Traditional disclosure control methods and disclosure risk measures can not readily be applied to this type of maps. In previous papers [5, 6] we discussed plotting the distribution of a dichotomous variable on a cartographic map. In the present paper we focus on plotting a continuous variable and derive a suitable risk measure, that not only detects unsafe areas, but also contains a recipe to repair them. We apply the risk measure to the spatial distribution of the energy consumption of enterprises to test and describe its properties.

p% should dominate

Both the (n,k)-dominance rule as well as the p%-rule are well known and often used sensitivity measures in determining which cells are unsafe to publish in tabular output. The p%-rule has some theoretical advantages over the dominance rule, hence it is generally advised to use that rule instead of the latter one. In this paper we investigate the relation between the (n,k)-dominance rule and the p%-rule. We propose a dynamic rule to determine a value p*(k) that yields, approximately, the same number of unsafe cells as a corresponding (n,k)-dominance rule.

Three ways to deal with a set of linked SBS tables using τ-ARGUS

In Council Regulation no. 2701/98 of the European Committee, a framework is given on an extensive set of tables concerning economic statistics. Some of these tables are linked to each other. Until recently, there existed no practical solution to a consistent protection of that set of tables, save for a rather naive one. In this paper we will show the new way specific sets of linked tables can be protected using the τ-argus software and compare this with two other approaches.

Reducing the Set of Tables τ -ARGUS Considers in a Hierarchical Setting

A heuristic for disclosure control in hierarchical tables was introduced in [5]. In that heuristic, the complete set of all possible subtables is being protected in a sequential way. In this article, we will show that it is possible to reduce the set of subtables, to a set that contains subtables with the same dimension as the complete hierarchical table only. I.e., in case of a three dimensional hierarchical table, only three dimensional subtables need to be checked, not all two or one dimensional subtables.