两阶段RDD方法无回答的影响及其改进

Influence and Improvement of Non-Response in Two-Stage RDD Method

计算机辅助电话调查应用日益普遍,在住户调查中为了提高住户的比例,发展了Mitofsky-Waks-berg两阶段RDD方法。文章分析了在无回答率较高,且初级单元间无回答率差异较大时引起的偏倚,分析偏倚的特点,并且提出一种基于随机组方法的抽样设计,旨在解决计算机辅助调查中的无回答问题及地区间无回答率差异,并且能够得到性能良好的估计量。

The application of computer assisted telephone survey is becoming more and more common. In order to improve the proportion of households in household survey, the Mitofsky-Waksberg two-phase RDD method has been developed. This paper analyzes the bias caused by the high non-response rate and the large difference in the non-response rate among the primary sampling units, as well as the characteristics of the bias. And then the paper proposes a sampling design based on the random group method which aims to solve the non-response in computer assisted telephone survey and the difference in the non-response rate among the regions and also be able to get good estimators.