New Developments on Quantile Regression Analysis of Censored Data: Theory, Methodology and Computation

截尾数据分位数回归分析的新进展：理论、方法和计算

• 批准号: 1308960
• 负责人: Lan Wang
• 金额: $ 12万
• 依托单位: University of Minnesota-Twin Cities
• 依托单位国家: 美国
• 项目类别: Standard Grant
• 财政年份: 2013
• 资助国家: 美国
• 起止时间: 2013-09-01 至 2016-08-31
• 项目状态: 已结题
• 来源: https://www.nsf.gov/awardsearch/showAward?AWD_ID=1308960&HistoricalAwards=false
• 关键词: New; Developments; Quantile; Regression; Analysis

Quantile regression has recently emerged as a valuable semiparametric alternative to the popular Cox model for analyzing censored data. It directly models survival time; thus is easy to interpret. More importantly, it relaxes the proportional hazards constraint associated with the Cox model and is particularly powerful for heterogeneous data. Despite the remarkable recent progress, several important and challenging statistical problems remain unsolved. For example, there exists limited literature on censored quantile regression when the sample arises from an observational study and is not representative of the target population; when the censored data come from genomics studies involving high-dimensional covariates; or when random effects are present due to the incorporation of latent variables. Motivated by these challenging problems, this project will develop novel methodology, theory and algorithms, which have the potential to significantly advance the applications of censored quantile regression. The PI will rigorously study the theoretical properties of the proposed new procedures and investigate their applications in practical data analysis. Censored data arise in diverse fields such as economics, engineering, medicine, psychology and sociology. The new methodology and theory are expected to make important contributions to the current body of knowledge on statistical analysis of survival data. In particular, the proposed research will make timely contributions to high-dimensional data analysis with censored responses, which has important applications in modern genomics and is still a relatively unexplored research area. The PI will develop useful software packages and make them freely available to the research community. The research results will be incorporated in different levels of statistical courses. The PI will also incorporate her research activity with graduate education. Students from minority groups will be especially encouraged to participate in the proposed projects.

分位数回归最近已成为流行的 Cox 模型的一种有价值的半参数替代品，用于分析审查数据。它直接模拟生存时间；因此很容易解释。更重要的是，它放宽了与 Cox 模型相关的比例风险约束，并且对于异构数据特别强大。 尽管最近取得了显着进展，但一些重要且具有挑战性的统计问题仍未得到解决。 例如，当样本来自观察性研究并且不能代表目标人群时，关于截尾分位数回归的文献有限；当审查数据来自涉及高维协变量的基因组学研究时；或者当由于潜在变量的结合而出现随机效应时。在这些具有挑战性的问题的推动下，该项目将开发新颖的方法、理论和算法，这些方法、理论和算法有可能显着推进截尾分位数回归的应用。 PI 将严格研究所提出的新程序的理论特性，并研究其在实际数据分析中的应用。 审查数据出现在经济学、工程、医学、心理学和社会学等不同领域。新的方法和理论预计将为当前生存数据统计分析的知识体系做出重要贡献。 特别是，所提出的研究将为具有审查响应的高维数据分析做出及时的贡献，这在现代基因组学中具有重要的应用，并且仍然是一个相对未经探索的研究领域。 PI 将开发有用的软件包并将其免费提供给研究界。研究成果将纳入不同级别的统计课程。 PI 还将把她的研究活动与研究生教育结合起来。 将特别鼓励少数群体的学生参与拟议的项目。