Data required for the surveillance of the population of small areas and the implementation and evaluation of health preventive programmes are usually obtained from surveys conducted within each relevant small area. The substantial cost of local surveys has encouraged the search for other methods of obtaining the required information. One alternative consists of using small area estimators. Despite extensive applications of these procedures in diverse fields, guidelines concerning their use for the prediction of health variables are still lacking. In an effort to explore the applicability of small area estimators to the prediction of health parameters of Quebec's health areas, we conducted two empirical evaluations of these methods. Using data from Canadian surveys, estimates of health variables were produced for several Quebec's areas according to different techniques of small area estimation. The estimates were compared to a "standard" for each area and health variable, on the basis of average mean square error percents and Spearman correlations. Synthetic, regression-sample, and empirical Bayes estimators were evaluated. We observed that the more variable a health characteristic was among areas, the more difficult it was to predict accurately. While no small area estimator performed uniformly well for all the variables considered, the linear regression-sample estimators were generally at advantage according to the different criteria of evaluation. In the studied context, no gain was obtained by using more sophisticated procedures like the empirical Bayes estimators.
| Identifer | oai:union.ndltd.org:LACETR/oai:collectionscanada.gc.ca:QMM.28702 |
| Date | January 1994 |
| Creators | Cardin, Sylvie. |
| Contributors | Boivin, Jean-Francois (advisor) |
| Publisher | McGill University |
| Source Sets | Library and Archives Canada ETDs Repository / Centre d'archives des thèses électroniques de Bibliothèque et Archives Canada |
| Language | English |
| Detected Language | English |
| Type | Electronic Thesis or Dissertation |
| Format | application/pdf |
| Coverage | Doctor of Philosophy (Department of Epidemiology and Biostatistics.) |
| Rights | All items in eScholarship@McGill are protected by copyright with all rights reserved unless otherwise indicated. |
| Relation | alephsysno: 001453338, proquestno: NN05683, Theses scanned by UMI/ProQuest. |
Page generated in 0.0018 seconds