Abstract
The increasing demand for small area data in fine detail can be satisfied through the use of special estimation techniques. However, as the size of samples often cannot be increased because of cost constraints, these estimation techniques can only be applied, if the variables studied comply with certain conditions. Synthetic or SPREE techniques offer a possibility to increase the reliability of estimates of small area characteristics.
Subject-matter statisticians sometimes express skepticism as to the-results obtained with such techniques. To overcome this difficulty, a simulation experiment has been carried out to prove the usefulness of these techniques. Ten variables were selected from the 1980 population census. Their exact values were known. Four different synthetic estimators were tested and compared with the simple direct estimator. The ten variables were studied at both a fairly aggregated level (minor domains) and in great detail (mini domains). The simple direct estimator with the synthetic or composite synthetic estimators shows an improvement of average reliability of 40 to 50 per cent.
This simulation experiment also proved the ease with which the techniques can be applied. This aspect is worth mentioning, as it came as a surprise to many statisticians.
Get full access to this article
View all access options for this article.