元分析中三种统计异质性估计方法的比较
Comparison of Three Estimators of Statistical Heterogeneity in Meta-analysis
-
摘要: 元分析中异质性的大小对模型选择和调节分析有着重要的参考价值，只有统计性质好的估计方法才能保证效应量的估计精度。通常有3种方法或途径可以计算异质性的大小，即矩估计、极大似然估计（M L ）和限制性极大似然估计（REM L ）。通过模拟研究发现，当研究数量超过40时，3种估计方法的均方误趋于一致；但在低于40的处理条件中，多数限制性极大似然估计的均方误介于矩估计和极大似然估计两者之间，更接近于统计量C‐R下限值，故推荐限制性极大似然估计作为异质性的首要估计方法。Abstract: It is very important for model choice and moderator analysis for the quantization of heterogeneity in meta‐analysis ,because the good statistical estimator can ensure the precision of the estimates of the pop‐ulation effect size .T here are three approaches to estimate the heterogeneity ,including moment estimation , maximum likelihood estimation and restricted maximum likelihood estimation .The simulation results indi‐cate that there is no difference in mean squared error using three estimators ,w hen the study size is more than 40 .While the size is less than 40 ,most of outcomes using REML are situated between moment esti‐mation and ML .This shows that REML is more closed to the Cramer‐Rao lower bounds .It is concluded that the REM L is the first estimator for the quantization of heterogeneity .
-
-
计量
- 文章访问数: 1058
- HTML全文浏览数: 719
- PDF下载数: 1
- 施引文献: 0
下载: