期刊文献+

Poisson分布下基于鞍点逼近的风险差置信区间构造

Confidence Intervais Construction for Risk Difference under Poisson Distribution Based on the Saddlepoint Approximation
下载PDF
导出
摘要 在流行病研究中,风险差反映了特定人群中某种暴露因素发展成为特定疾病的风险大小.因此,它的置信区间的构造对于流行病危险因素的推断具有重要的医学意义.然而,传统估计方法在小样本情况下表现不佳.对一些发病周期长、发病率低的疾病,Poisson抽样要比已有的逆抽样等更符合事实.在Poisson采样下利用似然比、得分检验和鞍点逼近方法构造置信区间,通过Monte Carlo模拟和实例对三种区间构造方法进行评价.模拟结果表明:鞍点逼近方法在区间长度和覆盖率上都表现最佳,尤其是在小样本量条件下估计效果较好. In epidemiological studies,the risk difference reflects the development of certain exposure factors in certain populations to be the risk of specific diseases.Therefore,it is of great medical importance to construct the confidence interval for the inference of the risk factors of epidemic disease.However,the traditional estimation method performs poorly in small samples.For some diseases with long periods and low incidence,Poisson sampling is more in line with the fact than the existing inverse sampling.In this paper,the confidence intervals are constructed by means of the likelihood ratio,the score test and the saddle point approximation under the Poisson sampling,and the three methods are evaluated by Carlo Monte simulation and exact cases.The simulation results show that the saddle point approximation method performs best in the interval length and coverage,especially in the small sample size.
作者 何晓申 HE Xiao-shen(School of Statistics, Lanzhou University of Finance and Economics, Lanzhou 730020, Chin)
出处 《兰州文理学院学报(自然科学版)》 2017年第1期28-32,共5页 Journal of Lanzhou University of Arts and Science(Natural Sciences)
关键词 风险差 POISSON分布 得分检验 鞍点逼近 MONTE CARLO模拟 risk difference Poisson distribution score test saddle point approximation Monte Carlo simulation
  • 相关文献

相关作者

内容加载中请稍等...

相关机构

内容加载中请稍等...

相关主题

内容加载中请稍等...

浏览历史

内容加载中请稍等...
;
使用帮助 返回顶部