中国组织工程研究

中国组织工程研究 ›› 2011, Vol. 15 ›› Issue (4): 733-736.doi: 10.3969/j.issn.1673-8225.2011.04.040

• 骨与关节综述 bone and joint review • 上一篇    下一篇

非比较研究及前后对照研究结局变量Meta分析的效应量选择

沈旭慧，王  珍   

  1. 湖州师范学院医学院，浙江省湖州市 313000
  • 收稿日期:2010-08-13 修回日期:2010-09-25 出版日期:2011-01-22 发布日期:2011-01-22
  • 通讯作者: 王珍，讲师，湖州师范学院医学院临床教研室，浙江省湖州市 313000 wangzhen@hutc.zj.cn
  • 作者简介:沈旭慧，女，1959年生，浙江省湖州市人，汉族，教授；主要从事社区护理与护理教育研究。

Effect size selection for Meta analysis of outcome variables in noncomparative studies and pre-post contrasts studies

Shen Xu-hui, Wang Zhen   

  1. Medical College of Huzhou Teachers College, Huzhou  313000, Zhejiang Province, China
  • Received:2010-08-13 Revised:2010-09-25 Online:2011-01-22 Published:2011-01-22
  • Contact: Wang Zhen, Lecturer, Medical College of Huzhou Teachers College, Huzhou 313000, Zhejiang Province, China wangzhen@hutc.zj. cn
  • About author:Shen Xu-hui, Professor, Medical College of Huzhou Teachers College, Huzhou 313000, Zhejiang Province, China

PDF

970

被引次数

5

摘要：

背景：关于成组设计类型比较研究的Meta分析或系统评价理论与方法学日趋成熟，但对单变量关系，又称非比较性研究或单组研究结局变量以及两变量关系研究，比如前后比较研究变量进行综合分析的文献与方法则相对较少。
目的：探讨针对单变量关系研究集中趋势描述的常用统计量比例或率及均数以及两变量关系研究中前后比较研究的常用统计量均数得分法的Meta分析法。
方法：应用计算机检索中文CNKI全文数据库、维普全文数据库、万方中国博硕数据库、PubMed数据库中非比较性研究以及前后比较研究相关的Meta分析或系统评价方法学文献。
结果与结论：单变量集中趋势描述的几种常见统计量有均数、中位数、众数或比例，但Meta分析最常见的是比例和均数(算术平均数)。实验前后比较研究，从根本上来讲也属于对同一反应人群(单样本)同一变量的测量值的集中趋势研究，与上述讨论的稍有不同之处在于该研究形式是针对同一变量的同一反应人群在两个不同时点测量值集中趋势的比较研究。这种类型的研究在统计学上也称作配对设计的均数比较，针对此类型的研究，有两种效应量指标可以选择，一种是未标化的平均得分，另一种是标化的平均得分。如果纳入Meta分析的各研究结果变量采用相同的测量方法，则可以直接用未标化的平均得分作为效应量，反之，则需用标化的平均得分作为效应量。

关键词: 非比较性研究, 前后对照研究, 结局变量, Meta分析, 效应量选择

Abstract:

BACKGROUND: Meta-analysis or system evaluation theory and method of grouping design pattern comparison are mature. However, there are a few articles or methods for univariate relationship, single group outcome variable and their relationship.
OBJECTIVE: To introduce several effect statistics, such as pooled incidence rate estimates and pooled arithmetic means estimates used for noncomparative studies, mean gain effect sizes estimates used for pre-post contrasts studies.
METHODS: A computer-online search of the full text database of Chinese CNKI, VIP full text database of Chinese Doctoral database, Pubmed database was performed. Relevant meta-analysis or systematic review methodology literature about non-comparative studies and pre-post contrasts studies were included.
RESULTS AND CONCLUSION: Univariate described several common statistics, such as mean, median, mode or ratio, while Meta analysis analyzed ratio and mean (arithmetic mean). Pre-post contrasts studies is to explore the measurement of one variable of one group population (single sample) at two time points. This type of study is also called matched-pairs design mean comparison. Two effective size indexes can be selected in this study: non-standardized mean score and standardized mean score. If one measurement method is used for results variables of Meta analysis, non-standardized mean score can be used, otherwise, standardized mean score should be used as effect size.

中图分类号: