Mathematical Sciences: Location of Outliers in Structured Data

数学科学：结构化数据中异常值的位置

• 批准号: 9010983
• 负责人: Douglas Hawkins
• 金额: $ 6.5万
• 依托单位: University of Minnesota-Twin Cities
• 依托单位国家: 美国
• 项目类别: Continuing Grant
• 财政年份: 1990
• 资助国家: 美国
• 起止时间: 1990-07-01 至 1993-12-31
• 项目状态: 已结题
• 来源: https://www.nsf.gov/awardsearch/showAward?AWD_ID=9010983&HistoricalAwards=false
• 关键词: Mathematical; Sciences; Location; Outliers; Structured

Statistical research will be conducted on finding outliers in data that is based on "elemental sets". This involves fitting many models to the data, each an exact fit to a minimally-sized subset of the original data, and combining the information from these elemental sets in suitable ways to identify multiple outliers reliably and in a single pass. The connection between this method and more traditional methods as well as high breakdown points and acceptable efficiency will be investigated for multivariate data in the multivariate regression model and in other types of structured data. The computational implications and the algebraic connections between elemental sets in regression and the simplex algorithm of linear programming will be explored.

将进行统计研究，以发现离群值 基于“元素集”的数据。这涉及拟合 许多模型的数据，每一个精确拟合到一个最小尺寸的 原始数据的子集，并将来自 这些元素集合以合适的方式识别多个 异常值可靠地在单次通过中。之间的连接 这种方法和传统的方法以及高 将调查故障点和可接受的效率 对于多元回归模型中的多元数据， 其他类型的结构化数据。计算的影响 以及元素集合之间的代数联系， 回归和线性规划的单纯形算法将 被探索。