荟萃分析中寻找样本均值和标准差的十种情况及解决方案。

Ten circumstances and solutions for finding the sample mean and standard deviation for meta-analysis.

作者信息

Chi Kuan-Yu, Li Man-Yun, Chen Chiehfeng, Kang Enoch

机构信息

School of Medicine, College of Medicine, Taipei Medical University, Taipei, Taiwan.

Department of Public Health, School of Medicine, College of Medicine, Taipei Medical University, Taipei, Taiwan.

出版信息

Syst Rev. 2023 Apr 1;12(1):62. doi: 10.1186/s13643-023-02217-1.

Abstract

A common problem in meta-analyses is the unavailability of mean and standard deviation (SD). Unfortunately, only having values of the median, interquartile range (IQR), or range cannot be directly utilized for meta-analysis. Although some estimation and conversion methods have been proposed in the past two decades, there were no published and user-friendly tools developed based on multiple scenarios of missing SD. Therefore, this study aimed to provide a collection of possible circumstances of missing sample means or SD with solutions for teaching and research. A total of 10 common circumstances of missing SD or mean could have available statistics of p value, t value, z score, confidence interval, standard error, median, IQR, and range. Teachers and investigators can use relevant formulas for finding the sample mean and SD according to the available circumstance. Due to the complicated computations, our team provides a free available spreadsheet. With ever-evolving statistical methods, some formulas may be further improved in the future; therefore, it is recommended to involve statisticians in evidence-based practice or systematic reviews.

摘要

荟萃分析中的一个常见问题是均值和标准差(SD)数据不可得。遗憾的是,仅有中位数、四分位间距(IQR)或极差的值无法直接用于荟萃分析。尽管在过去二十年中已经提出了一些估计和转换方法,但尚未开发出基于标准差缺失的多种情况且便于用户使用的已发表工具。因此,本研究旨在提供一系列样本均值或标准差缺失的可能情况及解决方案,以供教学和研究使用。共有10种标准差或均值缺失的常见情况,可能会有p值、t值、z分数、置信区间、标准误差、中位数、四分位间距和极差的可用统计量。教师和研究人员可根据可用情况使用相关公式来计算样本均值和标准差。由于计算过程复杂,我们团队提供了一个免费的电子表格。随着统计方法的不断发展,未来一些公式可能会进一步改进;因此,建议在循证实践或系统评价中让统计学家参与。

文献AI研究员

20分钟写一篇综述,助力文献阅读效率提升50倍

立即体验

用中文搜PubMed

大模型驱动的PubMed中文搜索引擎

马上搜索