整合分析中两种假设模型的介绍及实例分析
郑凤英’,彭少麟2
门.山东大学威海分校海洋生物工程系,威海264209:2.中国科学院华南植物研究所广州510650)
【摘要】整合分析(meta一analysis)是刘同一主题下多个独立实验结果进行综合的统计学方法,被认为是到目前为止
最好的数量综合方法。在进行整合分析时,一首选应提出统计假设,根据假设的不同一ijJ将整合分析分为固定效应模型(fixed
effectmodel)和随机效应模型(randomeffectmodel),前者假定有相似的多个研究在同一分组里有一个共同的真实效应
值,由于取样误差,导致在实际效应值的测定中各研究间存在差别;在后者一中,假定各研究间有随机变量,因此,不
共享一个真实效应值。介绍了两种假设模型下整合分析的计算方法,并进行了实例分析。
关键词:整合分析;固定效应模型;随机效应模型
中图分类号:Qz41文献标识码:A文章编号:1008一8873(2004)04一292一03
Comparison of two hypotheses models in meta-analysis and its sample using
ZHENG Feng-ying1, PENG Shao-lin2(1. Shandong University at Weihai, Weihai,264209;
2. South China Institute of Botany, Academia Sinica, Guangzhou 510650, China
Abstract Meta-analysis is a statistical method to summary independent studies under the
same subject. Hypothesis is the dependence of analysis. There are two hypothesis models:
fixed effect size model and the random effect size model. In fixed-model, it is assumed that
there is a real effect size in all studies, and it is sampling error taht leads to the difference
among studies. In random-model, it is assumed taht there is a random variable in all studies,
which and sampling error are sources of the difference among studies. These two models
were introduced and real sample analysis was done. |