隨機反應法於敏感性資料之調查－－模式檢驗與案例探討

Surveys on Sensitive Issues via Randomized Response Technique-- the Model Reexamination and a Case Study

隨機反應法是調查敏感性議題的有效工具，它的基本邏輯是：如果受訪者的隱私愈受到保護，則對於敏感性議題的調查他們必然會愈願意誠實回答。長久以來，文獻上的發展一直著重於受訪的隱私保護與安全感，以提高調查的合作度。相對地，本文則從統計理論著手，證明Warner模式與「非相關問題」模式中的估計式是優良的估計式，所以它們是可以讓調查者「放心使用」的估計式。再者，本文加入「成功性」的考量以幫助模式中的關鍵參數p（受訪者抽到敏感問題之機率）的設定。本文進而採「非相關問題」模式，針對中部地區校園內婚前性行為比例進行主題調查。從受訪者的隱私保護、調查者的「放心使用」、以及應用範圍的廣泛等三個角度來看，隨機反應法應用於敏感議題調查的空間是較匿名問卷調查為優的。

Randomized response technique is an efficient tool in the investigation of sensitive issues. Its essential logic is based on the protection of privacy. The more the personal privacy is protected, the more the respondent is willing to give honest answers for sensitive issues. For a long time, the development of randomized response technique has focused on the improvement of privacy protection to raise the sense of security and cooperation for investigation. In this paper, based on statistics, we prove that the sensitive behavior estimators appeared in Warner model and unrelated question model are superior; therefore, investigators can apply it with confidence. In addition, we further include the validity consideration of estimators to assist with the determination of model's key parameter p (p is the probability that respondents pick the sensitive issue). In the end, we conduct an investigation for the proportion of pre-martial sexual behavior in campus via unrelated question model. From the viewpoints of privacy protection, confidence in application and application scope, the randomized response technique is superior to anonymous questionnaire.