Symmetric stratified truth detection models

Abstract

When collecting sensitive information, many respondents are not willing to tell the truth about their attribute. For these respondents to give truthful information, they need to be assured that their identity will not be known. This can be achieved if data is collected in a randomization manner. Randomized response techniques (RRTs) aim to reduce untruthfulness in the assessment of sensitive attributes but differ regarding privacy protection. The less protection a method offers, the more likely respondents lie about their true characteristics. In this paper, we have shown that the symmetric stratified approach is relatively better than asymmetric stratified approach. We have shown that there is more compliance in the symmetric stratified models than in the asymmetric stratified models and therefore we recommend the use of symmetric stratified technique as a sampling method in collecting data on sensitive characteristics.