Semiparametric and Empirical Process Methods in Oncology

项目简介：拟建项目的长期目标是为肿瘤临床、流行病学和基础科学研究开发灵活的半参数分析工具。半参数模型在医学研究中可能非常有用，因为它们既包括通常易于解释的参数成分，也包括在存在生物不确定性时允许更大灵活性的非参数成分。虽然这些模型在科学上非常引人注目，但这些模型的用户友好推理技术的有限可用性严重限制了它们在实践中的使用。造成这种限制的部分原因是模型灵活性的增加增加了推理的难度。解决这一差距在推理方法是提出的研究的中心目标。这一目标将通过实现以下四个目标来实现：(1)开发和评估更灵活和有效的统计分析方法，用于正确审查的时间到事件数据的转换模型；(2)开发和评估区间审查癌症研究中灵活的半参数风险因素评估工具；(3)为癌症研究的半参数模型创建计算效率推断工具；(4)开发灵活的半参数方法，用于分析用于癌症研究的超高维筛选数据。虽然这些目标的应用领域似乎多种多样，但所有目标都涉及半参数模型推理和经验过程方法的结合。因此，这些目标的统计基础是高度相关的。相关性：此外，这些目标将为不同的癌症研究人员提供科学意义、灵活和用户友好的数据分析工具的显著改进集合。

DESCRIPTION (provided by applicant): Project Summary: The long-term goal of the proposed project is to develop flexible semi-parametric analysis tools for clinical, epidemiological and basic science studies in oncology. Semi-parametric models are potentially very useful in medical research because they involve both a parametric component which is usually easy to interpret and a nonparametric component which permits greater flexibility in the presence of biologic uncertainty. While these models are highly compelling scientifically, the limited availability of user-friendly inferential techniques for such models has severely restricted their use in practice. Part of the reason for this limitation is that the increased flexibility of the models adds an order of magnitude to the difficulty of the inference. Addressing this gap in inferential methodology is the central goal of the proposed research. This goal will be accomplished through achieving the following four aims: (1) Develop and evaluate more flexible and effective statistical analysis methods for transformation models in right censored time-to-event data; (2) Develop and evaluate tools for flexible semi-parametric risk-factor assessment in interval censored cancer studies; (3) Create tools for computationally efficient inference in semi-parametric models for cancer research; and (4) Develop flexible, semi-parametric methods for the analysis of extremely high dimensional screening data for cancer studies. While the application areas of these aims seems diverse, all of the aims involve a combination of semi-parametric model inference and empirical process methods. Thus the statistical underpinnings of the aims are highly interrelated. Relevance: Moreover, these aims will provide diverse cancer researchers with a significantly improved collection of scientifically meaningful, flexible, and user- friendly data analysis tools.