Phylogenetic Analyses of the Evolution of Continuous Characters

亲缘关系密切的物种往往在许多特征上彼此相似。对跨物种(比较)数据集的统计分析不能假设物种数据点在统计上是独立的或从共同的潜在分布中得出。前者导致假设检验自由度的不确定性；后者可能导致估计问题。几位作者提出了一些分析方法，试图通过将生态或生理兴趣的特征映射到根据独立信息(例如DNA序列)估计的“已知”系统发育图来解决这些问题。计算机模拟技术将被用来研究几种这样的方法的统计特性。首先，分析连续性状相关进化的两种新方法，一种基于自相关模型(Cheverud et al.，1985；Gittleman and Kot，1990)和另一种基于数量遗传学的系谱模型(Lynch，1991)，将其与以前研究的方法(非系统发育、独立对比、最小进化)进行比较，其I类错误率、功率和估计能力。其次，将开发一种用于协方差系统发育分析的模拟方法。第三，将开发计算机程序，以(1)在进化变化的替代模型下模拟数据系统发育，(2)对模拟数据执行协方差分析以获得F统计量的零分布，以及(3)允许基于图形的数据输入和树结构和特征数据的转换。生理学关系的比较分析是将生理学研究与生态学、行为学和生理学联系起来的一种重要的跨学科方法。这项提案侧重于比较分析的统计技术，并将验证和进一步发展对这一方法至关重要的比较方法。

Closely related species often resemble each other for many characteristics. Statistical analyses of cross-species (comparative) data sets cannot assume that species data points are statistically independent or drawn from a common underlying distribution. The former leads to uncertainties about degrees of freedom for hypothesis testing; the latter may lead to problems of estimation. Several authors have proposed analytical methods that attempt to solve these problems by mapping characters of ecological or physiological interest on a "known" phylogeny, estimated from independent information (e.g., DNA sequences). Computer simulation techniques will be used to investigate the statistical properties of several such methods. First, the Type I error rates, power, and estimation abilities of two new methods for analyzing correlated evolution of continuous traits, one based on autocorrelation models (Cheverud et al., 1985; Gittleman and Kot, 1990) and the other on pedigree models from quantitative genetics (Lynch, 1991), will be compared to those previously studied (non-phylogenetic, independent contrasts, minimum evolution). Second, a simulation approach to phylogenetic analysis of covariance will be developed. Third, computer programs will be developed that (1) simulate data phylogenetically under alternative models of evolutionary change, (2) perform analysis of covariance on the simulated data to obtain null distributions of F-statistics, and (3) allow graphically-based data entry and transformation of tree structure and character data. %%% Comparative analyses of phylogentic relationships is an important interdisciplinary approach linking studies in phylogency with ecology, behavior and physiology. This proposal focuses on statistical techniques of comparative analysis and will validate and further develop comparative methods essential to this approach.