Evaluation of methods for asking sensitive questions

In medical, psychological and sociological research, sensitive questions or extremely personal questions are frequently asked. A questionnaire survey on sensitive questions is complicated because participants often do not answer correctly or refuse to answer completely. In the social science literature, when personal and sensitive questions are asked, it is widely accepted that social desirability bias affects the survey research. There are different methods in the literature aiming to reduce measurement errors such as social desirability bias and to increase the reliability of the participants' responses. These methods address the problem of social desirability bias in order to increase the effectiveness of predictions by using indirect questioning techniques. The sensitive question methods evaluated with our thesis are Randomized Response technique, Nominative technique, Grouped Answer method, Unmatched-Count technique, and Crosswise model. In the first part, these methods are compared with the comprehensive simulation plan. The performances of the methods were evaluated according to the sample size and the current sensitive subject prevalence parameter of the population. In the second part, in order to reduce the small sample size and low prevalence effect for the crosswise model, Uludag correction of the crosswise model was proposed in this thesis study. As a result, the crosswise model performs quite well compared to other methods. In the use of the crosswise model, it is recommended to use the Uludag correction of the crosswise model proposed in this thesis, when working with low prevalence and small samples.