Modeling Evasive Response Bias in Randomized Response: Cheater Detection Versus Self-protective No-Saying

Area of Science:

• Statistics
• Survey Methodology
• Social Sciences

Background:

• Randomized response techniques (RRT) aim to reduce bias in sensitive surveys.
• Existing models (cheater detection, self-protective no sayers) partially address evasive responses.
• A need exists for models that integrate different bias types within novel RRT designs.

Purpose of the Study:

• To establish the correspondence between existing RRT bias models.
• To develop and introduce new statistical models for the hybrid "ever/last year" RRT design.
• To account for both self-protective no-saying and cheating behaviors in the "ever/last year" design.

Main Methods:

• Demonstrated the correspondence between the cheater detection and self-protective no sayers models.
• Developed novel statistical models for the "ever/last year" RRT design, including extensions.
• Utilized two empirical surveys on doping use to illustrate the proposed models.

Main Results:

• The proposed models for the "ever/last year" design allow for the inclusion of response bias parameters.
• Models with increased degrees of freedom were developed for more complex "ever/last year" designs.
• The models effectively illustrated bias patterns in doping use surveys.

Conclusions:

• The "ever/last year" design offers a flexible approach to studying sensitive topics.
• The developed models enhance the ability to quantify and correct for response biases in RRT.
• This research provides a foundation for future methodological advancements in sensitive survey research.