Functional Censored Quantile Regression for Investigating Heterogeneous Effects in Survival Data

用于研究生存数据异质效应的函数删失分位数回归

基本信息

批准号：
10164703
负责人：
Qi Zheng
金额：
$ 7.8万
依托单位：
UNIVERSITY OF LOUISVILLE
依托单位国家：
美国
项目类别：
财政年份：
2020
资助国家：
美国
起止时间：
2020-05-15 至 2022-01-31
项目状态：
已结题

来源：
https://reporter.nih.gov/project-details/10164703
关键词：
Address Awareness Big Data Biological Markers Biomedical Research Brain Injuries Cancer Patient Characteristics Collaborations Collection Community Health Computer software Data Data Analyses Dependence Diagnostic Disease Evaluation Event Gaussian model Goals Hematological Disease Hemodialysis Hemoglobin Heterogeneity Image Life Literature Long-Term Survivors Magnetic Resonance Imaging Malignant Neoplasms Measurement Measures Medical Mental Health Methods Modeling Modernization Nonparametric Statistics Outcome Patients Physicians Procedures Prognosis Property Public Health Public Health Applications Research Reporting Research Research Personnel Research Project Grants Residual state Seminal Severity of illness Specific qualifier value Statistical Methods Statistical Models Structure Technology Testing Time Visit Work base cancer recurrence data acquisition discrete time flexibility follow-up innovation insight interest prognostic value response success theories tool vector

项目摘要

Project Summary/Abstract Functional biomarkers that are in the form of curves, images, and objects, are often collected in biomedical studies nowadays, due to the rapid advances in data acquisition technology. In recent research, there has been growing awareness that the underlying association between the functional biomarkers and outcomes may be prone to considerable heterogeneity. Those heterogeneous associations often shed insight on scientiﬁc discoveries and entail signiﬁcant implications, but tend to be overlooked by many existing functional data analysis procedures with a narrow focus on the response mean. Our proposal aims at offering researchers alternative tools to explore the comprehensive information of the relationship between functional biomarkers and survival time. In contrast to statistical models that presume constant effects of covariates, quantile regression (QR) ac- commodates varying effects and may reveal more detailed dependence structure of outcomes on covariates. Regretfully, QR for functional data has barely been studied. The objective of this proposal is to make the QR framework applicable for investigating the regression heterogeneity in functional survival data, to develop reliable and efﬁcient estimation approaches, and to obtain sharp inference on the effects of functional biomarkers. We will propose a “local” functional censored QR (FCQR) method to evaluate the impacts of functional biomarkers on the survival time at a single or multiple pre-speciﬁed quantile levels and develop a related signiﬁcance test for testing the impact of functional biomarkers (Aim 1). Then we will develop a “global” FCQR method to investigate the varying effects of functional biomarkers on the survival time over a region of quantile levels, which will provide researchers with a comprehensive picture about the covariates-response association. In addition, two inference procedures, including a bootstrap resampling method for estimating the standard errors and the martingale-based model diagnostics, will be developed (Aim 2). Moreover, we will extend the “local” FCQR method to longitudi- nal measurements of functional biomarkers for dynamic prediction of residual life (Aim 3). Also, we will develop statistical software that efﬁciently implements the proposed methods. The innovation of our proposal is at least three-fold. Firstly, it will produce reliable and efﬁcient FCQR tools that facilitate the identiﬁcation and evaluation of new valuable functional biomarkers. Secondly, the successful completion of my proposal can signiﬁcantly advance the theory of QR and semi/non-parametric statistics, and further, broaden their applications in lots of biomedical research. Thirdly, our proposed methods will also serve as a ﬂexible platform for examining the heterogeneity in functional data. They can be readily extended to other public health applications.

项目摘要/摘要以曲线，图像和物体形式的功能性生物标志物经常在双性恋研究中收集如今，由于数据采集技术的迅速发展。对功能生物标志物与结果之间的基本关联可能很容易易于大勇气相当多的异质性。需要显着含义，但要被许多现有的功能数据分析程序忽略对呼吸平均值的关注狭窄。功能生物标志物与生存时间之间关系的全面信息。与假定消耗协变量综合效应的统计模型相反，分位数回归（QR）ac- 商品的变化效果，可能会揭示更详细的结果对协变量的依赖性结构。遗憾的是，用于功能数据的QR几乎没有研究。适用于调查功能生存数据中回归异质性的框架，以开发可靠和有效的估计方法，并获得了对功能生物标志物的影响的敏锐推断将提出一种“局部”功能性QR（FCQR）方法来评估功能生物标志物的影响在单个或多个预先指定的分位水平上的生存时间，并开发了显着性测试测试功能性生物标志物的影响（AIM 1）将开发一种“全局” FCQR方法功能生物标志物对分位水平区域的生存时间的不同影响，这将主要是有关协变量 - 响应协会的全面图片的研究人员。程序，包括一个估算标准错误的自举重采样方法和基于Martingale的错误模型诊断将开发（AIM 2）。功能生物标志物的NAL测量用于剩余寿命的动态预测（AIM 3）。有效地实现支撑方法的统计软件。我们的提议的创新至少是三倍。这有助于识别和评估新的有价值的官能生物标志物。我的建议的压缩可以显着提高QR和半/非参数统计理论，以及此外，在许多生物医学研究中扩大了他们的应用。作为功能数据中的异质性的灵活平台。公共卫生应用。