Meta分析中异质性方差区间估计方法及改进

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。本文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 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 estimating methods and improvement for heterogeneity variance in Meta analysis

The heterogeneity in Meta analysis could be measured by variance. To estimate the heterogeneity variance, some methods such as moment estimation,maximum likelihood estimation, empirical Bayes estimation were given. On the basis of point estimation, we could research the interval estimation of heterogeneity variance to measure the heterogeneity mote accurately and effectively. There are many methods for building the confidence intervals for heterogeneity variance, such as the likelihood estimation, WALD-type confidence intervals and Q-profile confidence intervals, etc. On the basis of introducing the estimation methods for heteroveneity variance, this paper assumes the heterogeneity variance subjects to the normal distribution approximately. Then the estimating method and analytical expression for confidence intervals of heterogeneity variance were given.The simulation and case study demonstrate the new method is moderate and credible.