敏感问题抽样调查问卷模型的研究及应用

【作者】 杨克飞；

【导师】 吕恕；

【作者基本信息】 电子科技大学 ， 运筹学与控制论， 2011， 硕士

【摘要】 沃纳1965年提出随机化回答技术(RRT)后,统计学者在此基础上的研究大致分为两类:一是研究样本获取的抽样方法,希望在不同条件下运用一种抽样方式使得调查最优;二是从改进模型自身入手,以期提高调查效率。本文首先讨论RRT和Warner模型及在此基础上改进的Simmons模型和Greenberg模型;其次介绍敏感性问题调查在简单无放回和分层抽样下的应用,并在此基础上将其推广至各层以简单无放回抽样的分层抽样、群规模大小相等与不相等的整群抽样以及二阶段抽样方式中,同时讨论各自的参数估计,给出了参数表达式。随着网络信息的高速发展,敏感性问题在网络调查中的应用已成为一个新课题。由于沃纳提出的RRT在实际调查中需要一个随机化装置,费时耗财,无疑增加了调查的难度,同时传统的随机化装置模型在网上调查中也缺乏现实可操作性,因而在社会网络调查中并未得以推广。鉴于此,本文设计出一系列问卷模型,在仍旧保持随机化回答的原则上除去了随机化回答装置,指出此类模型在网上调查中更占有优势,它方便、实用、经济、操作性强,同时我们给出各种问卷模型的具体设计,参数估计表达式及参数的设定。通过理论公式的对比和数值模拟比较各种问卷模型,提出如何选取问卷模型。最后,将问卷模型应用到各类复杂抽样中并给出参数估计表达式。更多还原

【Abstract】 Since Warner proposed randomized response technique(RRT) in 1965, the research of statistic workers have divided into two categories broadly: one kind is researching the sampling method using a sample of the most excellent under different condition; The others improve the model own in order to improve the survey efficiency. This paper introduces the RRT and the Warner model , Simmons model and Greenberg model first; Followed, introduces the application of the sensitivity sampling survey in simple random sampling without replacement and stratified sampling. We extent it to stratified that each layer using simple random sampling without replacement, and the same-size-group cluster sampling, vary-size-group cluster sampling and two-stage sampling. Also, we discuss their estimation and the parametric expression.With the rapid development of network, the sensitive question survey online becomes a new topic. Since Warner model needs a randomized device in the actual survey. Because of the device requirement high cost, and unrealistic operation, traditional sensitive question survey based on randomized response technique is hardly to be applied in social survey online. Therefore in view of this we design a series of questionnaire models, still maintaining the principle of randomized response, removing the randomized response device(RRD). We propose that those models are dominate in the online survey ,because they are convenient、practical、economic and feasible. Meanwhile, this paper has given the specific design of variety of questionnaire models, parameter estimation and discussed the parameter setting. Comparison of the theoretical formulas and numerical simulations of various questionnaire models, we give a method of how to select the appropriate model. Finally, take these models into application of various types of complex sample and parameter estimation expressions.更多还原