摘要
Meta分析中一个较重要的问题是偏倚,它也是导致异质性的原因。当研究存在异质性时,传统Meta分析通常使用基于倒方差法的随机效应模型对结果进行合并。尽管随机效应模型使用基于瞬时估算的量r^2表示研究真实值间偏离程度,以此获得更"保守"的合并结果。然而这种方式并未对偏倚对每项研究结果的影响进行考虑,且存在低估标准误的风险,导致合并结果同样存在偏倚。Doi等提出一种新的加权模型,QE法,能够很好地解决上述问题。本文对QE加权模型及其软件的实现进行详细介绍,并将QE加权模型法与随机效应模型结果以一示例进行对比。
One important problem in meta-analysis is heterogeneity, the result of bias. When inconsistency occurs, traditional work in meta-analysis is employing a random effect model based on inverse variance method to combine the results. Such a method used the moment-based estimator T2 measuring the deviation from true value across studies to obtain a conservative result. It however failed to estimate the influence on each study due to bias and this method may at risk of underestimate the standard error which then may leads to biased summarized estimator. Accordingly, Doi proposed a new weighting procedure, QE method, hopefully be a good solution. In this article, we will introduce the QE method with details on the methodology and software, and then make a comparison between QE and random effect model of the results.
出处
《中国循证医学杂志》
CSCD
2016年第5期612-616,共5页
Chinese Journal of Evidence-based Medicine
基金
国家自然科学基金(编号:30972975)
湖北省教育厅重点项目(编号:D20142102)
