配对设计中缺乏差值标准差情况下的样本量估计策略  

Sample Size Estimation Strategies in the Absence of Information on the Standard Deviation of Differences in Paired Designs

作　　者：邢星 吴莹 陈平雁 Xing Xing;Wu Ying;Chen Pingyan(Department of Biostatistics,School of Public Health,Southern Medical University 510515,Guangzhou)

机构地区：[1]南方医科大学公共卫生学院生物统计学系,510515

出　　处：《中国卫生统计》2021年第2期181-182,187,共3页Chinese Journal of Health Statistics

基　　金：国家自然科学基金(81673270,81703322)。

摘　　要：目的在临床配对设计试验中,利用定量指标变化值(干预前后差值)进行两组比较时,其样本量的准确估计需要依据两组变化值的均数和标准差。如果参数假设源于外部数据的话,通常会缺少变化值的标准差信息。本研究旨在探索配对设计中缺乏差值标准差情况下的样本量估计策略。方法两组变化值的均数可直接从干预前均数与干预后均数相减得到。经理论推导,假设两组方差齐性,变化值的标准差可利用干预前后的相关系数与原数据共同标准差求得。结果在配对设计中缺少变化值的标准差的情况下,如何准确估计样本量实质上是一个如何准确估计干预前后测量值的相关系数的问题。如果相关系数高估,则导致检验效能偏低,即样本量相对不足;如果相关系数低估,则导致检验效能升高,使得样本量过大而造成浪费。结论缺乏差值标准差的配对设计在进行样本量估计时应借助相关程度来提高样本量估计的有效性。Objective In clinical trials where the changing value of a continuous variable(e.g.before and after an intervention)is used to measure the effect of interest between the two groups,accurate estimation of sample size needs to be based on parameters including the mean and standard deviation of this measurement.If such required information on parameters is derived from external data,the standard deviation of differences is usually missing.Our study aims to develop a sample size estimation strategy for these situations.Methods Mean value of each group can be directly obtained by subtracting the mean value before and after the intervention.By theoretical derivations,under the assumption of variance homogeneity,the standard deviation of differences can be obtained by using the correlation coefficient before and after the intervention and the common standard deviation of the original data.Results In the absence of the standard deviation of the changing value,how to accurately estimate the sample size is essentially a question of how to accurately estimate the correlation coefficient between the variables before and after the intervention.Overestimation of correlation coefficientscan lead to underestimations of required sample size and make tests inefficient,and vice versa,an unnecessary waste of sample size and resources.Conclusion The matched-pair design which lacks the standard deviation of the difference should improve the ralidity of the sample size estimation by means of correlation estimation.

关 键 词：样本量 变化值 标准差 检验效能 

分 类 号：R195.1[医药卫生—卫生统计学]