基于精确度考虑的二分类数据样本量估计  

Sample Size Calculation that Control for Precision of Estimate in a Single Binomial Population

作　　者：戎芬[1] 贾彬[2] 林燧恒[2] 

机构地区：[1]上海中医药大学预防医学教研室,上海201203 [2]复旦大学公共卫生学院卫生统计教研室

出　　处：《数理医药学杂志》2015年第6期799-802,共4页Journal of Mathematical Medicine

基　　金：上海市教委重点学科(第5期)建设资助项目(J50301)

摘　　要：目的:对传统的二分类数据样本量估计方法进行调整,以达到同时兼顾假设检验效能和精确度的要求。方法:采用"两步法"将效应估计的显著性和精确度结合起来估计样本含量。在传统方法计算所得样本量的基础上,逐步增加样本量,并同时计算可信区间宽度低于期望宽度的概率,当此概率超过预先设定的水平时即可停止,即直到可信区间宽度达到要求的宽度为止。结果:增大后的样本量可以较好的兼顾显著性和精确度的要求。结论:为达到一定的精确度要求,对传统样本含量估计过程进行调整是必要的。Objective：Adjust the conventional sample size estimation for binomial data by considering hy- pothesis testing together with precision control. Methods： Apply a two-stage iterative algorithm to increase the sample size and calculate the probability of the actual CI width less than the expected width until this probability is larger than the specific level so that a desired confidence interval width estimated by the improved method could be attained. Results： When the sample size is increased at some extent beginning from the value estimated by the conventional method, both power and precision can be guaranteed. Conclusion： Traditional sample size formulas should be adjusted in order to account for the precision of the estimator.

关 键 词：二分类数据 样本量估计 可信区间宽度 

分 类 号：R195.1[医药卫生—卫生统计学]