单样本均数样本量估算的案例与影响因素  

Case and affecting factors of one-sample mean sample size estimation

作　　者：冯玉婷 曹蕊 柴倩云[1] 罗慜婧 高一城 刘芷含 陶立元[2,3] 费宇彤[1] FENG Yuting;CAO Rui;CHAI Qianyun;LUO Minjing;GAO Yicheng;LIU Zhihan;TAO Liyuan;FEI Yutong(Center for Evidence-based Medicine,Bejing University of Chinese Medicine,Beijing 100029,China;Research Center of Clinical Epidemiology,Peking University Third Hospital,Beijing 100191,China;Medical Examination Centre,Peking University Third Hospital,Beijing 100191,China)

机构地区：[1]北京中医药大学循证医学中心,北京100029 [2]北京大学第三医院临床流行病研究中心,北京100191 [3]北京大学第三医院体检中心,北京100191

出　　处：《中华中医药杂志》2024年第7期3812-3817,共6页China Journal of Traditional Chinese Medicine and Pharmacy

基　　金：国家自然科学基金项目(No.82074282)。

摘　　要：目的:探讨单样本均数在差异性及非差异性检验中的样本量变化规律。方法:首先通过举例说明不同检验方法之间样本估算的差异,其次探索Cohen's d值、均数差(试验值-参考值,δ)、标准差(σ)、界值(△)以及它们之间比值大小对样本量的影响,同时探索单样本均数检验中非参数校正对样本量估算结果的影响。结果:当Cohen's d值从0~1取值时样本量逐渐下降,当其越接近1时样本量越小,同时其下降速度也在变小。当Cohen's d值接近0.1时,样本量约1000例;当Cohen's d值接近0.4时,样本量仅有约50例;但当Cohen's d值≥1时,样本量≤10。与非参数校正前的结果比较,Doubleexponential法校正后样本量下降32%左右,Logistic法校正后下降7%左右,Normal法校正后的样本量增加6%左右,不同非参数校正方法间样本量相差较大。在非劣效性检验中,样本量随着[σ/△]或[δ/△]绝对值的增大而增大;在优效性检验中,样本量随着[σ/△]绝对值的增大而增大,随着[δ/△]绝对值的增大而减小;在等效性检验中,样本量随着[σ/△]的增大而增大,随着[δ/△]的增大而减小。结论:明确不同检验统计方法中参数变化对样本量的影响,可帮助中医药领域研究者明确样本量变化规律,从而根据研究目的及研究实际校正样本量参数值,选择正确的方法计算得出合理恰当的样本量。Objective:To explore the rule of sample size change in one-sample means in difference and non-difference tests.Methods:The differences in sample estimation between different tests were firstly illustrated by giving examples,followed by exploring the effects of Cohen's d value,mean difference(test value-reference value,δ),standard deviation(σ),and margin(△)and the magnitude of the ratio between them on sample size,and also exploring the effects of nonparametric correction in the one-sample mean test on the results of sample size estimation.Results:The data simulation results show a gradual decrease in sample size when Cohen's d value were taken from O~l,and the sample size becomed smaller as it gets closer to 1,while its rate of decline also becomed smaller.When the Cohen's d value was close to 0.1,the sample size was about 1000 cases;when the Cohen's d value was close to 0.4,the sample size was only about 50 cases;but when the Cohen's d value was greater than or equal to 1,the sample size was less than or equal to 1o.Compared to the results before non-parametric correction,the sample size decreased about 32%after correction by the Double exponential method,decreased about 7%after correction by the Logistic method,and increased about 6%after correction by the Normal method,with a large difference in sample size between different nonparametric correction methods.In the non-inferiority test,the sample size increased with the absolute value of[σ/△]or[δ/△];in the superiority test,the sample size increased with the absolute value of[σ/]and decreased with the absolute value of[δ/△];in the equivalence test,the sample size increased with the absolute value of[σ/△]and decreased with the absolute value of[δ/△].Conclusion:Clarifying the effect of parameter changes in different test statistics on sample size can help researchers clarify the pattern of sample size changes,so as to correct the values of sample size parameters according to the research purpose and research reality,and choose the correct metho

关 键 词：单样本 均数 样本量估算 效应量 差别性检验 非差异性检验 因素 

分 类 号：R195.1[医药卫生—卫生统计学]