Abstract
An often applied procedure in the statistical disclosure control of microdata sets is to prescribe a minimum number of population elements for each category of a combination of identifying variables and to take measures to ensure that there are no categories with a population frequency less than the prescribed minimum. In many cases the population frequencies will be unknown and the disclosure protection procedure can only be applied if a reasonable estimator for these frequencies is available. The usual unbiased direct estimator cannot be
applied because it is based on too few sample observations. Since one of the identifying variables is almost always a regional indicator, it seems natural to consider small area estimators for this problem. In this article a synthetic and a combined estimator are proposed and studied, and expressions for their expected mean squared errors are derived. The proposed estimators are compared by means of an example based on data from the Dutch Labour Force
Survey.
applied because it is based on too few sample observations. Since one of the identifying variables is almost always a regional indicator, it seems natural to consider small area estimators for this problem. In this article a synthetic and a combined estimator are proposed and studied, and expressions for their expected mean squared errors are derived. The proposed estimators are compared by means of an example based on data from the Dutch Labour Force
Survey.
Original language | English |
---|---|
Pages (from-to) | 399-410 |
Journal | Journal of Official Statistics |
Volume | 14 |
Issue number | 4 |
Publication status | Published - 1998 |
Externally published | Yes |