Stability of Numerical Classifications

在这项以成就为基础的更新建议中，首席调查员索卡尔计划继续研究重建生物体系统发育(祖先-后代关系)及其相关分类的各种数值方法。Sokal过去的工作试图建立三种不同方法对分类群之间关系进行排序的准确性和稳定性(瓦格纳简约树法、最大似然估计法和他自己创建的UPGMA聚类法)。在该项目期间，将针对进化特征变化的不同假设构建计算机模拟。6到8个真实数据集将与模拟数据进行比较，并将使用列出的三种方法分析这两个数据集的稳定性和准确性。在过去的十年里，系统学见证了数据分析方法的激增，索卡尔过去关于这些方法在模拟数据上的相对优点的工作得到了广泛的受众。然而，实践中的系统主义者还不能先验地预测哪种方法最适合手头的真实数据。所提出的基于当前字符变化模型的模拟，然后与真实数据集进行比较，将解决这个问题。如果成功，它将填补真实数据和模拟数据之间的空白，从而开辟一个富有成效的研究领域。

In this Accomplishment Based Renewal proposal, Principal Investigator Sokal plans continued study of the various numerical approaches to reconstructing phylogenies (ancestor-descendant relationships) and their associated classifications of organisms. Sokal's past work has sought to establish the accuracy and stability of three different methods for ordering relationships among taxa (Wagner parsimony trees, maximum likelihood estimation, and his own creation, UPGMA, a clustering technique). During the period of this project, computer simulations will be constructed for different hypotheses of evolutionary character change. Between six and eight real datasets will be compared to the simulated data, and both sets will be analyzed for stability and accuracy using the three methods listed. Systematics has witnessed the proliferation of approaches to data analysis in the past decade, and Sokal's past work on the relative merits of these approaches on simulated data has had a wide audience. Yet the practicing systematist is not yet able to predict a priori which approach is the most suitable for the real data at hand. The proposed simulations based on current models of character change, followed by comparisons with real data sets, will address this problem. If successful, it will fill the gap between real and simulated data and will thereby open an area of fruitful research.