CORE--BIOSTATISTICS FACILITY

观察者研究提供了必要的和令人信服的评估 医学成像技术。因此，大量这样的研究 都包括在这个计划项目中。一些基于计算机的模型 还将进行评估。核心人员将提供 领导研究设计，并将进行所需的统计 分析。研究计划将直接通过优化来改进 设计和分析，并间接地由科学家解放时间。堆芯 人员还将进行所有研究数据的管理。 我们还将通过开发更多的方法来支持任何研究计划 比目前可用的效率更高。以前的许多分析 多年来，人们一直使用非线性模型作为第一步。必然性 小样本和相关观测不允许使用可用的 直接分析方法，因为它们依赖于大样本 参数，并仅适用于有限的设计。因此，非线性模型 将参数估计作为传统线性模型的输入 方法以便提供可辩护的推论。它的局限性 现有的方法还没有得到很好的宣传，也没有得到很好的理解 医学成像研究人员。因此，我们将延长可用时间。 方法，并分享这些进展。 与整个医学成像社区合作。 统计核心活动有四个组成部分，每个组成部分都有自己的组成部分 明确的目标。1、设计。贯彻设计原则 由Muller，Barton和Benigus[1984]和Kleinbaum，Kper推荐 和Muller[1988]在所有的研究中。2、分析。进行所有统计 项目研究的数据分析。3、研究数据管理。 存档和管理统计分析产生的所有数据 为项目研究制定学习计划。4、方法。发展精准的小 一类带重复项的非线性模型的样本假设检验 医学成像研究所需的措施。

Observer studies provide a necessary and compelling evaluation of any medical imaging technique. Consequently a large number of such studies are included in this program project. A number of computer based model evaluations will also be conducted. The core personnel will provide leadership in study design and will conduct the required statistical analyses. The research program will be improved directly by optimal design and analysis, and indirectly by the scientist time freed. Core personnel will also conduct all research data management. We shall also support any program of research by developing methods more efficient than those presently available. Many analyses in previous years have used nonlinear models as the first step. The necessarily small samples and correlated observations disallow using available methods for direct analysis because they depend on large sample arguments and only apply to limited designs. Hence nonlinear model parameter estimates were used as input to traditional linear model methods in order to provide defensible inference. The limitations of existing methods have not been well publicized to, nor well understood by, medical imaging researchers. Hence we shall extend available methods to the more general case mentioned, and share those advances with the entire medical imaging community. The statistical core activity has four components, each with its own specific aims. 1, Design. Implement the principles of design recommended by Muller, Barton, and Benigus [1984] and Kleinbaum, Kupper and Muller [1988] in all studies. 2, Analysis. Conduct all statistical data analyses for project studies. 3, Research Data Management. Archive and manage all data resulting from statistical analysis and study planning for project studies. 4, Methods. Develop accurate small sample hypothesis tests for a class of nonlinear models with repeated measures needed for medical imaging studies.