Meta分析模型权重变异比较及其与I2的相关性.docVIP

Meta分析模型权重变异比较及其与I2的相关性 石修权,刘 丹,刘 俊 （遵义医学院公共卫生学院流行病与卫生统计学教研室，贵州省遵义市563003的与的与Meta分析的学习和结果的合理解释均具有重要意义。 ：MetaR-05, R181.3+2 The comparsion of variance of weights in meta-analysis models and its correlation with I-square SHI Xiuquan, LIU Dan, LIU Jun. Department of Epidemiology and Health Statistics, School of Public Health, Zunyi Medical College (Zunyi Guizhou 563003, China) Abstract: Objective To explore the difference of variance of weights() in two models and its correlation with I-square of heterogeneity(I2) in meta-analyses. Methods Weights and I2 were extracted from meta-analyses which were published in recent three years, then were computed. The difference of was compared in two models and the correlation was investigated between and I2. Results in random-effect model was less than fixed-effect model (t=2.739, P=0.015); and the correlation coefficient between and I2 was -0.505 and P-value was 0.039. Conclusion in random-effect model was significant lower than fixed-effect mode and a negative correlation between and I2 was existed. Moreover, it plays an important role to understand how to assign weight in two models, even in the study and suitable explanation of the results in meta-analyses if the reasons of difference and correlation between I2 and were thoroughly searched. Key words: Meta-analysis; Model; I-square; Weight; Correlation Supported by the Foundation of Zunyi Medical College(XG200907). Corresponding author: SHI Xiuquan, E-mail: xqshi@zmc.edu.cn. Meta分析是对多项原始研究结果的，科学性体现对每个单项研究赋予一个恰当的权重研究间异质性严重程度异质性I2。赋予异质性越大，研究权重的变异度？本研究探讨权重方差与异质性I2的相关性。×100%换算求得I2。然后计算权重的变异度(方差值)。不同模型下权重方差的差异比较采用t检验，权重方差与固定效应模型 随机效应模型 第一作者 权重方差 异质性I2(%) 权重方差 异质性I2(%) 谢圣芳0.102 26.4 丁鑫0.002 95.4 王京京0.121 0.0 蔡俊英0.074 55.0 史亚祥0.057 41.3 孙捷豪0.051 63.7 魏明0.032 0.0 李娟娥0.019 48.2 廖明娟0.072 71.6 段凯男0.067 80.0 Zhou Y-17 0.205 10.7 余欣慧0.013 61.5 Zhou Y-27 0.059 46.3 Lin KY-115 0.013 91.0 李嘉-10.137 42.0 Lin KY-215 0.014 90.0 李嘉-2 0.032 0.0 合计n1=9 0