Incomplete Categorical Data Design: Non-Randomized Response Techniques for Sensitive Questions in Surveys

Unlike the established randomized response (RR) technique, non-randomized response (NRR) techniques yield reproducible results in survey design and analysis. This book presents new techniques designed to overcome the bias inherent in posing sensitive questions in sociological or behavioral science surveys, without requiring a means of randomization. The authors provide a systematic introduction to NRR techniques that can overcome the limitations of RR techniques, combining the strengths of existing approaches, such as RR models, incomplete data design, expectation-maximization algorithm, data augmentation algorithm, and bootstrap method.