Mathematical Sciences: Bayesian Inference and Computing

数学科学：贝叶斯推理与计算

• 批准号: 9005858
• 负责人: Joseph Kadane
• 金额: $ 28.71万
• 依托单位: Carnegie-Mellon University
• 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=9005858&HistoricalAwards=false
• 关键词: Mathematical; Sciences; Bayesian; Inference; Computing

Tools will be developed for assessing sensitivity of results to assumptions about the prior probability distribution (which represents knowledge gained from sources other than the data being analyzed). A deeper understanding of the role of parameterization of statistical models will be gained through the development of diagnostics for the detection of a poor approximation by a Normal distribution to another distribution. The focus will be the approximation of the posterior distribution which is the basis for inference in a Bayesian analysis. Appropriate numerical integration strategies will be incorporated in a computing environment for Bayesian analysis. Further work will be carried out on asymptotic Bayesian inference for nonparametric problems in which the assumptions of the statistical model are greatly weakened.

将开发工具，以评估结果对先前概率分布的假设的敏感性（这代表了除了被分析的数据以外获得的知识）。通过开发诊断方法来通过正态分布与另一个分布的正态分布检测差近似，可以获得对统计模型参数化作用的更深入的了解。 重点将是后验分布的近似，这是贝叶斯分析中推断的基础。 适当的数值集成策略将纳入用于贝叶斯分析的计算环境中。将对渐近贝叶斯推断的非参数问题进行进一步的工作，其中统计模型的假设大大削弱了。