Research On The Sensitive Question Survey Techniques

With the continuous development and progress of society, various problems andcontradictions in the society present a diversified trend. To solve these problems, the firstthing to do is to have a comprehensive, accurate understanding of these issues. However, itoften appears some sensitive questions in the investigation of these social problems, suchas a tax evasion behavior, drug experiences, a bribery behavior, a cheating behavior.Because of these problems involve the privacy of respondents or private interests, thus letrespondents have the worry and give false answers even refuse to answer. However,studies on these sensitive questions are very necessary, because these problems directlyaffect the development and stability of society. In order to improve the security of sensitivequestions survey and reduce the feeling of insecurity among respondents, which improvethe accuracy of the investigation result, many domestic and foreign scholars began to studyon techniques on sensitive question surveies.This paper introduces techniques on sensitive question surveies systematically,including the randomized response model and the non-randomized response model ofsensitive questions. In the randomized response model, this paper introduces Warner’smodel, the unrelated question randomized response model, the two stage randomizedresponse model and the randomized response model for multiple sensitive attributes. Onthe basis of the unrelated question randomized response model and the two stagerandomized response model, this paper proposes the two stage unrelated questionrandomized response model. On the basis of the unrelated question randomized responsemodel and the randomized response model for multiple sensitive attributes, this paperproposes the unrelated question randomized response model for multiple sensitiveattributes. In the non-randomized response model, this paper introduces the triangle model,the cross model and the parallel model and proposes the method to compare twopopulations in the non-randomized response model based on the parallel model. To thetriangle model, combined with the stratified sampling method, this paper proposes thetriangle model under stratified sampling. Finally, the non-randomized response modelwhich is rarely applied in practice is applied in the survey for achievements of teachingreform. This paper calculates the proportion of students who skip classes and tests theapplication effect of the triangle model under stratified sampling.