Mathematical Sciences: Adaptive Statistical Procedures

数学科学：自适应统计过程

• 批准号: 8803259
• 负责人: Andrew Rukhin
• 金额: $ 5.9万
• 依托单位: University of Massachusetts Amherst
• 依托单位国家: 美国
• 项目类别: Standard Grant
• 财政年份: 1988
• 资助国家: 美国
• 起止时间: 1988-06-15 至 1990-11-30
• 项目状态: 已结题
• 来源: https://www.nsf.gov/awardsearch/showAward?AWD_ID=8803259&HistoricalAwards=false
• 关键词: Mathematical; Sciences; Adaptive; Statistical; Procedures

Rukhin will work on adaptive statistical procedures which are relevant when probabilities of large deviations are used as a performance criterion. In a hypothesis testing situation, Rukhin plans to obtain conditions for the existence of an adaptive test, i.e., a test which is fully asymptotically Bahadur efficient for any fixed value of an unknown nuisance parameter and such that the test is independent of this parameter value. The form of the adaptive test statistics will be derived, and their behavior for moderate sample sizes will be explored. Similar questions will be investigated in the problem of estimating a vector parameter in the presence of an infinite-dimensional nuisanceparameter. Recurrent classification procedures will be studied and improved reliability quantile estimators will be obtained. This work in the field of statistics focuses on solving problems that will adapt hypotheses testing procedures to information as it accummulates. Success on these problems will contribute to the knowledge of how to optimize the performance of scientific experimentation.

Rukhin将致力于自适应统计程序， 当大偏差的概率被用作 性能标准 在假设检验的情况下，Rukhin 计划获得自适应测试存在的条件， 也就是说，完全渐近Bahadur有效的测试， 未知干扰参数的任何固定值，并且 该测试与该参数值无关。 的形式 自适应测试统计量将被导出，以及它们的行为， 将探索中等样本量。 类似的问题将 在估计向量参数的问题中被研究 in the presence存在of an infinite-dimensional无限维nuisanceparameter参数. 将研究和改进经常性分类程序 将获得可靠性分位数估计值。 统计领域的这项工作侧重于解决 问题，将适应假设检验程序， 信息，因为它是。 在这些问题上的成功将 有助于了解如何优化性能， 科学实验