Use of Non-Informative Prior in Estimating Population Proportion

張紘炬(Horng-Jinh Chang) ； 郭美貝(Mei-Pei Kuo)

《管理科學研究》 7卷2期 (2011/06) Pp. 63-74

https://doi.org/10.6873/MSR.201106_7(2).0004

摘要

Randomized response techniques are designed for protecting the privacy of respondents and reducing the response bias while eliciting information on sensitive issues. Linear function of sample proportion is commonly used for estimating population proportion. By illustrating with the Warner (1965) model, we investigate the estimation problem of choosing estimators from various perspectives including minimizing the average mean square error and squared average bias of linear estimators. Three practical estimators are proposed, and these are compared with the regular estimator with respect to mean square error criterion. The results also cover to the case of direct response surveys. In particular, one of the proposed estimators may be viewed as a generalization of Böhning and Viwatwongkasem (2005) estimator.

關鍵字

Direct Response ； Privacy Protection ； Randomized Response