敏感问题调查设计以及估计

Questionnaire Design And Estimation of the Sensitive Questions

对于多项选择敏感问题抽样调查,为了避开以往使用随机化调查时使用随机化装置带来的不便,提出一种间接问卷调查策略.结合问卷设计对个体隐私的保护度,对辅助信息的选择提出了的建议,并得到了敏感属性比例的极大似然估计和Bayes估计.数据模拟的结果显示,相比于极大似然估计,小样本时Bayes估计便有较高的精度.随着样本量的增加,两种估计结果趋于一致.提出的间接问卷调查策略,相比于以往的随机化调查技术,具有方便、省时等特点,可在实践中应用.

For polytomous sensitive questions survey, an indirect questionnaire design tech- nique is proposed to avoid the inconvenience on traditional randomized response technique. The selection for assisted information is suggested combining the protection of privacy. The Maximum likelihood estimation and Bayesian estimation for the sensitive attribute probabil- ity are derived. Simulation results shown that precision of the Bayesian estimation is better than the Maximum likelihood estimation for small sample, and differences of the both meth- ods become smaller with increase of sample size. The proposed technique exhibited obvious advantages over the traditional randomized response technique, including simple handle, no site limitation and saving time, and it can be applied to practice.