Resampling Methods for Quantitative Genetic Analysis

进化生物学和农业科学中许多问题的解决依赖于对被研究种群中存在的遗传变异和协变水平的准确估计。这些所谓的数量遗传参数的估计和测试传统上是基于限制性假设和一套有限的实验程序。这里提出的工作旨在通过利用最近开发的重新抽样方法(主要是刀切和自举)作为一种更一般的方法来提供对数量遗传参数的统计假设检验的手段，从而消除许多这些限制。将开发和测试用于简化参数估计和执行重采样分析的软件。关于使用重采样技术的一个主要问题是，必须为使用重采样技术的每个应用程序重新验证它们。因此，将花费大量时间来测试新设计的重抽样程序的统计特性。将使用上面开发的重采样方法来分析已知分布的数据集的蒙特卡罗模拟，并对照预期结果测试这些方法的误差估计和功率特性的准确性。一旦开发和验证，该软件将极大地简化定量遗传数据的分析，并使许多生物学家对他们得出的结论更有信心。虽然数量遗传学在农业系统中已经使用了几十年，但进化生物学家直到最近15年或20年才开始使用数量遗传学方法。在此期间出现的核心问题之一是性状间遗传变异和协方差模式的进化变化的性质。这一问题通常是关于遗传方差-协方差矩阵是否随时间保持不变的问题。本提案中将开发的软件将允许提出关于遗传协方差结构进化的更广泛的一组问题。通过利用方差-协方差分析方法的最新发展，使用这里开发的方法的生物学家将能够处理关于来自两个或更多种群的遗传方差-协方差矩阵之间的关系的整个假设层次。一旦这些假设得到解决，导致两个种群分化的净选择模式就可以重建和测试。总而言之，将开发一套连贯的方法和程序，将那些从事数量遗传学的人从乏味的分析中解放出来，并允许他们追求许多新的和令人兴奋的研究途径。

The solution to many questions in evolutionary biology and agricultural science depend on the accurate estimation of the level of genetic variation and covariation present in the populations being studied. Estimation and testing of the these so-called quantitative genetic parameters has been traditionally based on restrictive assumptions and a limited set of experimental procedures. The work proposed here aims to eliminate many of these restrictions by utilizing recently developed resampling methods (primarily the jackknife and bootstrap) as a means of providing a more general approach to the statistical hypothesis testing of quantitative genetic parameters. Software for streamlining parameter estimation and for performing the resampling analysis will be developed and tested. A primary concern regarding the use of resampling techniques is that they must be revalidated for each application in which they are used. A great deal of time will therefore be devoted to testing the statistical properties of the newly devised resampling procedures. Monte carlo simulations of data sets of known distribution will be analyzed using the resampling methods developed above, and the error estimation and power properties of these methods tested for accuracy against the expected results. Once developed and validated, this software will greatly simplify the analysis of quantitative genetic data, and provide many biologists with greater confidence in the conclusions they reach. Although quantitative genetics has been used for many decades in agricultural systems, it is only in the last fifteen or twenty years that evolutionary biologists have begun to make use of quantitative genetic methodology. One of the central questions that has arisen during that time is nature of the evolutionary change in the pattern of genetic variance and covariance among traits. This line of questio ning is usually developed in terms of asking whether the genetic variance-covariance matrix remains unchanged through time. Software to be developed in this proposal will allow a much broader set of questions regarding the evolution of genetic covariance structure to be asked. By utilizing recent developments in the analysis of variance-covariance methods, biologists employing the methods developed here will be able to address an entire hierarchy of hypotheses regarding the relationship between genetic variance-covariance matrices from two or more populations. Once these hypotheses are addressed, the net pattern of selection that led to the divergence of the two populations can be reconstructed and tested. In all, a coherent set of methods and procedures will be developed that will free those doing quantitative genetics from the tedium of analysis and allow them to pursue many new and exciting avenues of research.