ABSTRACT
In this chapter, we first introduce the randomized Warner model to investigate its point estimation, relative efficiency, and degree of privacy protection. Second, to set the comparison of different models on the same footing, we present a so-called crosswise model, which is a non-randomized version of the original Warner model. Two asymptotic confidence intervals and two bootstrap confidence intervals are provided. Asymptotic property of the modified maximum likelihood estimator for the proportion π of the population belonging to the sensitive class is explored. Third, Bayesian methods for the crosswise model are also given. Finally, an induced abortion study and an experimental survey on measuring plagiarism are used to illustrate these methods.
