Estimating the dependence of mixed sensitive response types in randomized response technique

Area of Science:

• Statistics
• Medical Survey Research
• Health Informatics

Background:

• Sensitive questions in healthcare surveys necessitate truthful responses.
• Randomized response technique (RRT) aids truthful data collection.
• Methods for estimating dependence among multiple sensitive responses are limited.

Purpose of the Study:

• To develop a novel method for estimating the dependence structure of multiple, simultaneously collected sensitive responses.
• To address the gap in statistical methods for analyzing complex sensitive data in health surveys.
• To provide a robust estimation technique that does not require the joint distribution of responses.

Main Methods:

• Moment estimation approach applied to sensitive survey data.
• Construction of a covariance matrix for multiple sensitive questions under RRT.
• Calculation of conditional means and partial correlations for continuous sensitive responses.

Main Results:

• The proposed moment estimation method effectively estimates dependence structures with incomplete information from RRT.
• Simulation studies demonstrate the estimator's bias and variance across different sample sizes.
• The method successfully analyzed the dependence among health and pressure survey responses in college students.

Conclusions:

• The developed method provides a viable approach for analyzing the dependence of multiple sensitive responses in healthcare research.
• This technique enhances the utility of RRT by enabling more complex statistical analyses.
• The findings have practical implications for understanding sensitive health behaviors and relationships in population studies.