A Generalized Regression Estimation of the Item Sum Technique in Sensitive Surveys

Abstract

Survey researchers often find it difficult to collect reliable data of human populations, yet the validity of any research depends mainly on the accuracy of self-reported behavior especially when the respondents are to reflect about sensitive issues or highly personal matter. It is therefore important to develop methods of improving interviewees responses in any survey. The Item Sum Technique (IST) is the most recent indirect questioning method and it is a variant of the Item Count Technique (ICT) which can be used only for qualitative responses. The aim of this study is to estimate the sensitive characteristic when using the IST especially if two or more sensitive questions are investigated. It also focuses on the theoretical framework which includes the introduction of a classical method called the Generalized Regression model (GREG) using the IST. The efficiency of the GREG method was ascertained in comparison to the Calibration estimator by an extensive simulation study. Results from the statistical analysis indicates that the GREG estimator competes well with the calibration method and can further be used for a small sample size or data that is not normally distributed.