Re-calculation and Comparison of Sample Weights of the Turkey Demographic and Health Surveys

Abstract

Sample surveys are widely used methods to collect data all over the world and they have been widely conducted on the issues of population growth, demography and health since the last century. Within this framework, the Demographic and Health Survey (DHS) has been organized in more than 90 countries and nationwide population surveys in Turkey under this program has been implemented by Hacettepe University Institute of Population Studies since 1993.

In the DHS Program, the calculation of sampling weights has been changed over time and the reason of this methodological change has not been evaluated. This thesis aims to examine the changes in the weighting approach of each Turkey Demographic and Health Survey, to observe the effects of weight calculation approach changes on the basic selected indicators, to use the same weighting approach across different surveys when comparing the survey results and to make a contribution to the literature about comparing two weighting approaches in a Demographic and Health Survey program framework.

In this thesis, the new sampling weights were obtained and added to the corresponding data sets to produce some statistical tables for selected variables and compared with the current results. There was no pattern observed among surveys in terms of differences in point estimates. Although the widths of confidence intervals were higher in cluster level weights approach, it was possible to ignore these effects because their values were very small. Moreover, the ratio of coefficient of variation and design effect were increased by the cluster level weighting. The cluster level approach gave us higher sampling variances.