论文部分内容阅读
目的探索Meta分析的贝叶斯分层模型分析和后验计算的网格抽样模拟。方法应用贝叶斯分层模型对一组激素预防新生儿肺透明膜病的临床试验数据进行Meta分析,采用网格抽样的方法获得参数的后验样本和后验估计量。结果激素效应的经验贝叶斯估计值为-0.490(95%可信区间为-0.814~-0.461),其完全贝叶斯估计值为-0.551(95%可信区间为-0.878~-0.379),结果都表明临床上使用激素能够明显降低新生儿肺透明膜病的发生。结论相对传统Meta分析策略,贝叶斯分层模型更加灵活。
Objective To explore the Bayesian stratification model analysis and post-computation grid sampling simulation of Meta analysis. Methods A Bayesian stratified model was used to analyze the clinical data of a group of hormones for the prevention of neonatal hyaline membrane disease. The posterior sampling and posterior estimates of the parameters were obtained by using the grid sampling method. The empirical Bayesian estimate of the hormone effect was -0.490 (95% confidence interval -0.814 -0.461) and its complete Bayes estimate was -0.551 (95% confidence interval -0.878 -0.379) The results show that the clinical use of hormones can significantly reduce the incidence of neonatal hyaline membrane disease. Conclusion Bayesian stratification model is more flexible than traditional Meta-analysis strategy.