Meta分析中异质性方差区间估计方法及改进

The Estimating Methods and Improvement for Heterogeneity Variance in Meta Analysis

Meta分析中的异质性程度可以用异质性方差度量。为了估计异质性方差,人们研究并给出了各种估计方法,如矩估计、最大似然估计、经验贝叶斯估计等。在点估计的基础上,还进一步研究了异质性方差的区间估计,从而更加准确和有效地度量异质性大小。构建置信区间的方法也有很多,如似然估计、WALD型置信区间、基于Q统计量的置信区间等。在介绍已有的几种异质性方差区间估计方法的基础上,给出了假设τ2近似服从正态分布的情况下,异质性方差的区间估计方法,导出了异质性方差置信区间的解析表达式。仿真计算及实例分析表明该方法是稳健、可靠的。

The heterogeneity in Meta analysis could be measured by variance. To estimate the heterogeneity va- riance, some methods such as moment estimation, maximum likelihood estimation, empirical Bayes estimation were given. On the basis of point estimation, the interval estimation of heterogeneity variance to measure the heterogenei- ty mote accurately and effectively is could researched. There are many methods for building the confidence intervals for heterogeneity variance, such as the likelihood estimation, WALD-type confidence intervals and Q-profile confi- dence intervals, etc. On the basis of introducing the estimation methods for heteroveneity variance, the heterogenei- ty variance subjects are assumed to the normal distribution approximately. Then the estimating method and analyti- cal expression for confidence intervals of heterogeneity variance are given. The simulation and case study demon- strate the new method is moderate and credible.