【作者】 刘鹏；

【导师】 高歌；

【作者基本信息】 苏州大学 ， 公共卫生， 2012， 硕士

【摘要】 目的：在敏感问题的抽样调查中，调查指标涉及个人隐私或者其真实结果不被社会和公众所认同时，被调查者往往会出于抗拒或者隐瞒的心理不愿意给出真实答案或者给出错误答案。随机化应答技术（Randomized Response Technique,RRT）的出现有效解决了这一问题。其最初在1965年由沃纳（Warner）提出，后由Simmons等人进行了改进，出现了多种针对不同类型问题的模型。但要保证敏感问题调查结果的真实性与可靠性还需与科学的抽样方法相结合。在本课题组进行研究之前，国内外对于敏感问题的调查研究局限于简单随机抽样，或调查后仅就单次调查结果进行讨论。本文选定了数量特征敏感问题的无关联、加法、乘法模型与分层抽样方法进行组合，推导出其相关的统计公式，旨在解决敏感问题的复杂抽样问题。并将其应用于北京MSM人群调查当中。并且实现了用EXCEL软件对调查方法和统计公式的蒙特卡洛模拟以评价其信度与效度。为进行敏感问题大规模抽样调查提供简便、科学、可靠的调查方法、统计公式及评价方法。方法：在对数量特征问题分层抽样调查的公式推导过程中运用了数理统计学的基本理论和方法。并对北京市城区、近郊、远郊的MSM人群应用数量特征敏感问题随机应答加法模型进行分层随机抽样调查。对调查结果运用EXCEL软件中相关函数、公式和宏进行了模拟调查和计算分析以评价其有效性及可靠性。结果：推导出了分层抽样时数量特征敏感问题无关联问题模型、加法模型、乘法模型的总体均值的估计量及其方差的计算公式。在对本文调查方法进行实例应用过程中，调查得出北京市MSM人群首次发生男男性行为的平均年龄为22.07岁；每人每月发生男男性行为的不同性伴数的均值为2.74个；每人每月发生男男性行为次数的均值为4.59次。对基于EXCEL实现的蒙特卡洛模拟抽样调查结果进行了150次模拟样本与模拟总体差异的Z检验，其中142次模拟抽样的结果与模拟总体的差异无统计学意义(P<0.05)。结论：1.本文所研究的分层随机抽样下数量特征敏感问题RRT模型的调查方法及其统计公式有较高的信度和效度，在对北京市MSM人群进行调查的过程中应用成功。2.调查结果提示北京市MSM人群男男性行为存在多性伴的现象，多性伴是MSM人群中艾滋病及性病患病率高于一般人群的原因之一，需引起相关卫生部门的重视。3.基于EXCEL软件所实现的蒙特卡洛模拟有简便、上手容易、适合非计算机专业人员使用等优点。更多还原

【Abstract】 Objective：In the survey of sensitive question, when it comes to sensitive problem,respondents often refuse to answer or providing false answer in order to protect theirsecrets. The randomized response technique (RRT) solve this difficult problem. It wascreated by Warner in1965.Then it was improved by Simmons and other people, theycreated multiple kinds of model to solve different kinds of questions. To ensure thevalidity and reliability of survey, RRT must be used with scientific sampling method.Before the study of our project team, researches of randomized response technique(RRT) for sensitive questions were developed and studied with the restriction that thesample was selected through simple random sampling. the discussion just base on thesingle survey. In this paper, we combine the stratified random sampling with doubleirrelevant problem model、additive model、multiply model, derived relevant formulas tosolve complex sampling on sensitive question survey. We also applied these in thesurvey of MSM population in Beijing. What’s more, we use the Excel to apply MonteCarlo simulate in order to assess validity and reliability of our method. In this way, weprovide scientific and effective design on sensitive questions under the condition ofstratified sampling from large population.Method:When deriving relevant formula, the basic theories and methods of mathematicalstatistics were used. During the survey of MSM population of Beijing urban, peri-urban,sub urban,the additive model of RRT was used.In the process of simulation, functions,formulas and macros of Excel were used to calculate the result and assess the validityand reliability of our method.Result:The formula were deduced for the estimator of the population proportion and its estimated variance for double irrelevant problem model、additive model、multiplymodel used for sensitive questions survey in stratified sampling. The survey results ofBeijing MSM population are as follows: The average age of first MSM age was22.07.The average number of different sexual partners of MSM was2.74person every monthfor per person. It was4.59times of the number of MSM on average monthly. In the150times Z test of the result of Monte Carlo simulation based on the Excelsoftware,142times test have no statistical significance.Conclusion:1. The Stratified Random Sampling technology innovation for QuantitativeSensitive Question have meaning. The application in the MSM populationof Beijing is successful.2. The survey show that multi sexual partners phenomenon does exist amongthe MSM population of Beijing. Multi sexual partners is one of the reasonthat cause the AIDS and venereal disease rates among MSM higher than inthe general population. This phenomenon should cause the attention of therelated department of health.3. The Monte Carlo simulation based on the Excel software is a simple methodto learn and use, especially for the non-computer professionals.更多还原