Summarizing a posterior distribution of trees using agreement subtrees
Summarizing a posterior distribution of trees using agreement subtrees
复制标题
DOI：
10.1080/10635150701485091
复制
发表时间：
2007-01-01
影响因子：
6.5
通讯作者：
Rannala, Bruce
中科院分区：
文献类型：
--
作者：
Cranston, Karen A.;Rannala, Bruce
系统发育的贝叶斯推断在系统发育重建方法中是独一无二的，因为它产生树的后验分布，而不是最佳树的点估计。总结这种分布的最常见方法是报告用每个分区的边际后验概率注释的多数规则共识树。报告单个树会丢弃完整基础分布中包含的信息，并将贝叶斯分析简化为另一种用于查找树的点估计的方法。即使需要对系统发育进行点估计，多数规则共识树也只是一种可能的方法，可能还有其他更适合给定数据集和应用的方法。我们提出了一种总结树分布的方法，该方法基于识别后验分布中经常出现的一致性子树。该方法为类群子集提供完全解析的二叉树，在整个树上具有高边际后验概率，并包括有关分布扩展的附加信息。
Bayesian inference of phylogeny is unique among phylogenetic reconstruction methods in that it produces a posterior distribution of trees rather than a point estimate of the best tree. The most common way to summarize this distribution is to report the majority-rule consensus tree annotated with the marginal posterior probabilities of each partition. Reporting a single tree discards information contained in the full underlying distribution and reduces the Bayesian analysis to simply another method for finding a point estimate of the tree. Even when a point estimate of the phylogeny is desired, the majority- rule consensus tree is only one possible method, and there may be others that are more appropriate for the given data set and application. We present a method for summarizing the distribution of trees that is based on identifying agreement subtrees that are frequently present in the posterior distribution. This method provides fully resolved binary trees for subsets of taxa with high marginal posterior probability on the entire tree and includes additional information about the spread of the distribution.