Robust Calibrated Optional Randomized Response Techniques for the Estimation of Sensitive Variable Information with Applications to SITs Data
Ahmed Audu
Aminu Muhammed
Yakubu Musa
Mojeed A. Yunusa
Abstract
The use of existing RRTs Models in STIs research offers several advantages, such as increased response accuracy, reduced social desirability bias, and enhanced confidentiality protection. However, the existing RRT models utilize no auxiliary information which can enhance the accuracy and precision of estimate for prevalence of STIs. In addition, the estimators of the existing RRT models are prone to outliers or extreme values being them sample means of the interest variables which in turn can reduce the efficiency and accuracy levels of models. In this research, we proposed new classes of randomized response technique called calibrated three optional randomized response techniques.
These models were created by adjusting current RRT models using calibration techniques. The aim was to enhance C-RRT models to be more efficient, stable,
and robust compared to existing alternatives. The research established theoretical properties, including estimators, variances, privacy levels, and a composite metric for efficiency and privacy, to evaluate the robustness and
applicability of the proposed models. Empirical studies were conducted using simulated data to support the theoretical findings, and demonstrating that the RCTHORRT models exhibited lower variances, higher relative efficiency,
enhanced privacy levels, and a better combined metric of variance and privacy. This indicates the superiority of the C-THORRT models over existing RRT models.
References