Çankaya Üniversitesi Fen-Edebiyat Fakültesi, Journal of Arts and Sciences Say›: 6 / Aral›k 2006 Methods Used in Reduction of Bias Arising

Various types of survey errors, especially nonresponse errors, may seriously deteriorate data quality. Nonresponse has more one reason to worry about its harmful effects on the survey estimates. So, nonresponse and methods dealing with nonresponse have increasingly become a standard part of survey sampling. A nonresponse occurs in a survey when, for any reason, a selected unit does not respond. The usual methods of estimation in the presence of nonresponse give biased results. Because of this special estimation techniques are required to deal with the problem. Imputation and reweighting are two standard methods provided by the literature for treating nonresponse. In the recent years, scientist became increased to concern with the calibration approach to reweigthing method in the presence of nonresponse. The calibration approach generates the final weights which are as close as possible to specified design weights, while respecting known auxiliary population totals or unbiased estimates of these totals. This calibration procedure requires the formulation of a suitable auxiliary vector, through a selection from a possible larger set of auxiliary variables. In this study standard methods for the reduction of bias and errors arising from nonresponse are explained. The calibration approach is examined as theoretically and simulation is performed by macro generated in C++ programming language to study how alternative specifications of the auxiliary vector affect the quality of estimators derived by the calibration technique. In the application, the calibration estimators and quality measures such as relative bias, variance are computed and interpreted for the population total