Polya后验方法在有限总体抽样估计中的模拟研究  

Simulation Study on Polya Posterior Approach to Finite Population Sampling

作　　者：戴明锋[1] 金勇进[2] 孙婕[1] 

机构地区：[1]中国人民大学统计学院,北京100872 [2]中国人民大学应用统计科学研究中心,北京100872

出　　处：《统计与信息论坛》2013年第4期10-13,共4页Journal of Statistics and Information

基　　金：国家社科基金项目<普查数据质量的事后抽查理论及其应用研究>(11BTJ009);中国人民大学研究生科学研究基金项目<基于偏离-份额分析法的我国各地区高技术产业竞争力评价研究>(12XNH159)

摘　　要：polya后验方法作为一种无信息贝叶斯估计方法,在有限总体抽样中,通过观测的样本,构造一系列的模拟总体,然后进行统计推断。通过统计模拟研究了polya后验方法估计的一些特点,并和Bootstrap方法进行比较。模拟结果显示:polya后验方法能够很好地估计总体的均值,随着样本量的增大,估计值与真值的差距越来越小。采用polya后验方法构造的置信区间区间长度较小,能够很好地覆盖真值。Polya posteriori method is a non--informative Bayesian estimation method. It was used in finite population by constructing a series of completed simulated copies of the entire population based on observed sample unites. Then we can make statistical inference. In this paper, some of the features of the Polya sampling method was studied by statistical simulation , In addition, we compare Bootstrap method and Polya posteriori method. The simulation results show： Polya posteriori method can estimate the population mean very well, as the sample size increases, the difference between the estimated value and the true value is getting smaller and smaller. Interval length of confidence intervals constructed using Polya posteriori method is shorter compared with the Bootstrap method. Moreover, the confident interval can cover the ＂truth＂ well.

关 键 词：polya后验 抽样估计 统计模拟 置信区间 

分 类 号：O212.2[理学—概率论与数理统计]