相关差别指标的区间估计方法比较  

A Comparative Study on the Interval Estimation for Relative Difference

作　　者：苏宇楠[1] 魏传华[1] 苏强[2] 

机构地区：[1]中央民族大学理学院,北京100081 [2]首都医科大学附属北京世纪坛医院检验科,北京100038

出　　处：《统计与信息论坛》2017年第3期3-9,共7页Journal of Statistics and Information

基　　金：国家社会科学基金项目<收入差距的代际传递机制及在民族地区的实证研究>(16CTJ011);国家统计局重点研究基金项目<基于时空统计模型的京津冀区域协同发展的统计监测指标体系研究>(2014LZ16);北京高等学校"青年英才计划"项目<复杂数据下半参数模型的研究>(YETP1316);中央民族大学校级科研奖励项目<区间估计的流行病学模拟与应用研究>(2016LXY22);首都医科大学附属北京世纪坛医院院级青年课题<前列腺增生与前列腺癌鉴别诊断标志物的研究>(2014-q09)

摘　　要：在公共疾病控制领域,重大稀有疾病的发病率非常低,符合逆抽样特征,量化分析重大稀有疾病的发病率并对其特点进行分析,为了研究在带有群内相关条件下的整群抽样问题,通过β-二项分布抽样对比流行病学中相关差别指标的六种渐近置信区间的构造方法,综合考虑实际覆盖率与区间长度对各种方法的优劣及适用情况并对比分析。研究表明,Wald型置信区间与对数变换的置信区间对发病率的估计表现因参数而定,而Bootstrap类方法不稳定。本研究找出了不同区间估计方法的适用场合,认为应合理看待置信区间这种评估方法在流行病学中的实际应用。In the field of public disease control, the incidence of major rare diseases is very low, in accordance with the inverse sampling characteristics, in order to quantify the incidence of major rare diseases and analysis of its characteristics. This paper mainly studies on the related conditions within the group with cluster sampling, by/^binomial distribution, under inner-correlated cluster sampling conditions by comparing the differences related to the epidemiology of this indicator in the six asymptotic confidence intervals constructed. The advantages and disadvantages of various methods and their application conditions are analyzed by considering the actual coverage ratio and the interval length. The results show that the estimation performance of the confidence interval of the Wald type confidence interval and the logarithmic transformation is dependent on the parameters, and the Bootstrap method is not stable. In this paper, the application of different interval estimation methods is found, and the practical application of confidence interval estimation method in epidemiology is reasonable.

关 键 词：β-二项分布 相关差别 整群抽样 流行病学 

分 类 号：O212[理学—概率论与数理统计] R195.1[理学—数学]