Mathematical Sciences: Produce a World Wide Web Program of the Conference on Statistical ChallengeÐ and Possible Approaches in the Analysis of Massive Data Sets

提案：DMS 95-22317 PI：John亚历山大机构：美国国家科学院标题：产生一个世界范围的网络计划的会议统计挑战和可能的方法在分析大量数据集 摘要 关于统计和海量数据集的讲习班， 国家研究理事会应用和理论统计委员会， 重点是如何有效地处理需要统计分析的大量数据。在许多实际情况下，研究人员 面临着标准统计技术不足以处理当前数据集规模爆炸的问题，在这些数据集中，观测数据的数量很容易达到千兆字节甚至兆兆字节的范围。 许多最流行的方法所依据的统计假设，如数据的“同质性”，在这些情况下可能是错误的，从而使普通统计工具无效，甚至危险。 一个相应的问题是在当前许多问题的维数的增长。 同样，标准的方法，在不同的时代，甚至在许多情况下早于计算机，开发出来的，是不够强大的这些挑战。 关于统计和海量数据集的研讨会探讨了海量数据集带来的挑战，并在以下背景下进行了探讨： 各种真实的世界和当前紧迫的应用领域，包括大气，生物和地球科学，商业和市场营销，生态学，工程和技术。以这些地区为背景， 关于共同需求的具体的多学科研究问题， 包括相关的统计学原理、计算方面、数据分析和大量数据集的可视化。 研讨会上科学家们进行了热烈的讨论。 这些讨论不容易以传统的出版物形式记录下来。 与此相反，活动是通过制作讲习班录像带，然后通过万维网传播的方式记录和传播的。 这种新的传播方式带来的好处是 及时向广大听众提供实际讨论的情况。

Proposal: DMS 95-22317 PI: John Alexander Institution: National Academy of Sciences Title: Produce a World Wide Wed Program of the Conference on Statistical Challenges and Possible Approaches in the Analysis of Massive Data Sets ABSTRACT The workshop on Statistics and Massive Data Sets, hosted by the Committee on Applied and Theoretical Statistics of the National Research Council, focuses on how to deal effectively with massive amounts of data that need statistical analyses. In numerous practical contexts, researchers are confronted with the inadequacy of standard statistical techniques to handle the explosion in the size of current data sets, where the number of observations can easily run into the giga- and even terabyte ranges. Statistical assumptions underlying many of the most popular methods, such as "homogeneity" of the data, may simply be wrong in these cases, thus rendering common statistical tools ineffective and even dangerous. A corresponding problem is the growth in the number of dimensions in many current problems. Again standard approaches, which were developed in a different era, even predating the computer in many cases, are not powerful enough for these challenges. The workshop on Statistics and Massive Datasets explores the challenges presented by massive data sets, and does so in the context of a variety of real world and currently pressing application areas, including the atmospheric, biological and geo sciences, business and marketing, ecology, and engineering and technology. With these areas as a backdrop, specific over-arching multi-disciplinary research issues on common needs, including relevant statistical principles, computational aspects, data analysis, and visualization of massive data sets are identified. The workshop involves highly animated discussions among scientists. These discussions are not easily captured in a traditional published proceedings format. In contrast, the activities are capture d and distributed via the preparation of a videotape of the workshop followed by a World Wide Web (WWW) dissemination. This new dissemination method brings the benefit of the actual discussions to a wide audience in a timely fashion.