Methods Of Investigation Of Sensitive Issues

Sampling survey is widely used to investigate all kinds of questions, includingsociety, economy. Sensitive questions are in terms of confidentiality, which peopleusually do not wish to give correct information, for example, proneness to tax evasion,addiction to drugs, ambling habits and so on. When open or direct surveys are aboutsensitive matters, the respondents usually refuse to answer or give wrong answer, inorder to protect own privacy or other intention. It is necessary to use RandomizedResponse Technique (i.e. RRT) for getting reliable sample data on sensitive matters.RRT, namely respondents answer the sensitive questions with preassignedprobability p using randomized device in survey, and estimating an unknown proportionof people(Ï€_A) possessing a sensitive characteristic in a given community, comparing todirect answer, RRT can protect respondents' privacy efficiently, making them answertruly. Warner(1965) developed Warner model that is called randomized responsetechnique, Warner's randomized response technique is designed to eliminate evasiveanswer bias and keep respondents' confidentiality, many variants of Warner model havebeen presented, such as Simmons model, Mangat model,Kuk model and so on.Sensitive questions are divided into two categories: one is qualitative sensitive questions;the other is quantitative sensitive questions. Investigating qualitative sensitive questionsis estimating the population proportion of people possessing a sensitive trait in a givenpopulation. Investigating quantitative sensitive questions is estimating the average orsum of sensitive data.Basic randomized response models are introduced in the first chapter. Somemethods on qualitative sensitive questions are discussed in the second chapter, Bayesianmethods are used to value the parameters, discussing two different Bayesian methods,and then getting the approximation by Maple. Unrelated question technique is a methodof investigating qualitative sensitive questions, and an improved method is proposedwhich is more efficient. In the survey, two sensitive questions or multinomial sensitivequestions are discussed. Some methods on quantitative sensitive questions are discussedin the third chapter, the random amputation model is a very simple method which is used to practice, which turns qualitative sensitive questions into quantitative sensitivequestions, avoiding direct answer, it's accepted easily by respondents, an advancedrandom amputation model is presented which is more accuracy and improvesrespondents' cooperation. Multiplication model is also a method which is used toinvestigate quantitative sensitive questions, and an advanced randomized method ispresented whose estimator is more precise. Additive model is a preferably method, anadvanced method is applied to stratified sampling in the paper, which is more efficientand very helpful in large sample survey. Mangat model(1994) is analyzed in thesituation of incomplete true response in the fourth chapter, and the random amputationmodel is compared to its advanced model.