For A Randomized Response Data With Self-preservation Phenomenon Semiparametric Bayesian Regression Analysis

Randomized Response is a common survey method in the investigation of sensitive issues, Because of the intention to protect self-privacy in investigation, respondents still exhibit self-protective behavior in Randomized Response System. This paper analyzes the statistical method of randomized response data in the presence of self-protective behavior. We use Dirichlet process as the prior distribution of link functions. And then we can obtain the marginal post distribution (full conditional distribution) and give the Gibbs sampling method. The simulation results shows that in the median sample volume, our method is better than the parametric method when the link function of parametric method is designed wrong. Finally, we use our method in an example, and compared our method with the parametric method.