Mathematical Sciences: Conference on Multiple Decision Theory and Related Topics

DMS 95-00061，“多重决策理论及相关主题会议”PI：James Berger摘要：1995年6月8日至12日，将在普渡大学举行一次多重决策理论及相关主题会议。会议将包括两个部分，一个是关于多重决策理论的传统的邀请和贡献演讲的会议，另一个是关于最近的多重模型选择方法的密集研讨会，称为“内在贝叶斯因素”。会议的传统部分将包括由该领域领导人邀请的27次演讲，其中包括旨在评估该领域研究现状和评估未来需求的演讲。研讨会阶段将有大约20名活跃的研究人员参与内在贝叶斯因子的开发，这是一种非常新的、很有前途的方法，几乎适用于任何模型选择问题，无论是嵌套的还是非嵌套的，不需要渐近性。研讨会阶段将采用非传统框架，鼓励讨论未解决的问题，而不是仅仅强调介绍最近的研究。多重决策理论涉及有两个或多个种群的情况，希望选择“较好的”种群，对种群进行排序，或做出其他涉及种群比较的决策。这类情况在统计应用的几乎所有领域中的各种应用问题中都是常见的。近年来，该领域开展了大量活动，有该领域领导人参加的关于这一主题的会议非常及时。会议的讲习班阶段将聚集一小群研究人员，研究在多种统计模型中进行选择的最佳方法。模型选择通常是统计分析中最重要的阶段，将要讨论的新方法在自动选择最佳模型的能力方面非常有希望。努力程度：通过纳入关于内在贝叶斯因素的密集研讨会，会议的努力程度正在提高。增加这个研讨会的可能性是因为人们意识到许多活跃的内在贝叶斯因素研究人员将出席多重决策理论会议；因此，将会议延长两天以纳入这个研讨会似乎在科学和财务上都是明智的。预算说明：来自NSF的10,000美元请求不会改变。不过，约5000美元将用于参加密集讲习班的人员的费用。(当然，这些人也将参加会议的早期部分。)增加努力所产生的额外费用将由普渡大学承担。

Re: DMS 95-00061, "Conference on Multiple Decision Theory and Related Topics" PI: James Berger Abstract: During June 8-12, 1995, a conference on Multiple Decision Theory and related topics will be held at Purdue University. The conference will consist of two parts, a traditional conference of invited and contributed talks in multiple decision theory, and an intensive workshop on the recent approach to multiple model selection called "intrinsic Bayes factors." The traditional part of the conference will consist of 27 invited talks by leaders in the field, including talks aimed at assessing the current status of research in the area and assessing future needs. The workshop phase will involve approximately 20 active researchers in development of intrinsic Bayes factors, a very recent and promising methodology which is applicable to virtually any model selection problem, nested or nonnested, without requiring asymptotics. The workshop phase will adopt a non-traditional framework that encourages discussion of unsolved problems, rather than solely emphasizing presentation of recent research. Multiple decision theory is concerned with situations in which there are two or more populations, and it is desired to select the "better" populations, rank order the populations, or make other decisions involving the comparison of populations. This type of scenario is common to a huge variety of applied problems in virtually all areas of statistical application. The area has seen a great deal of activity in recent years, and a conference on the subject involving the leaders in the field is very timely. The workshop phase of the conference will bring together a smaller group of researchers to study optimal ways to select among multiple statistical models. Model selection is typically the most important phase of a statistical analysis, and the new methodology that is to be discussed is extremely promising in terms of its ability to automatically select the best models. Level of Effort: The level of effort in the conference is being increased, by inclusion of the intensive workshop on intrinsic Bayes factors. The possibility of adding this workshop arose when it was realized that many of the active researchers on intrinsic Bayes factors would be at the Multiple Decision Theory conference; extending the conference for two days to incorporate this workshop thus seemed scientifically and financially wise. Budget Explanation: The request for $10,000 from NSF is not being changed. Approximately $5000 will, however, be used for the expenses of those attending the intensive workshop. (These individuals will, of course, also be participating in the earlier portion of the conference.) Additional expenses incurred by the increased level of effort will be met by Purdue University.