Berkeley Symposium Economics, Statistics, and Computation for 1998, 1999, 2000 in Berkeley, California

9730461麦克法登近几十年来，随着计算能力的急剧增长，旨在改进统计推断的计算活动和研究也出现了相应的爆炸性增长。由于这一活动广泛而多样，新的计算方法往往在应用研究人员中缓慢渗透，并渗透到教学课程中。伯克利经济学、统计和计算研讨会的续期将为研究的传播、统计学和计量经济学研究人员之间的互动、软件演示和研究生指导提供一个焦点。在解决统计和计量经济学问题时，密集使用计算有望为原本难以解决的问题提供解决方案。例如，在统计学中，Bootstrap和相关抽样技术提供了精确逼近估计量和检验统计量的抽样分布的一般方法。在计量经济学中，模拟方法被用来掩盖高维多项式选择等复杂模型的难解性。这只是众多例子中的两个。稳健的推理方法、图形方法、贝叶斯分析、非线性时间序列模型以及半参数和非参数方法也在很大程度上依赖于新的计算方法。这个项目延续了1993年开始的年度研讨会计划，结合了教学调查和研究报告，以及初级和高级学者的混合。这笔赠款支持1998年、1999年和2000年的研讨会。每次研讨会将持续约5天，将在计量经济学实验室和加州大学伯克利分校经济系的设施中举行。试探性的主题是：非线性和非平稳时间序列模型；统计软件：体系结构、语言、算法；以及贝叶斯和经典计量经济学：计算考虑。另一项重要的次要行政任务是安装和测试与会者提供的软件，并通过计量经济学实验室软件档案馆监督这一软件的文件编制和发布。?？

9730461 McFadden As computational capacity has grown dramatically in recent decades, there has been a corresponding explosion in computational activity and research aimed at improving statistical inferences. Because this activity is widespread and diverse, new computational methods often permeate slowly among applied researchers and into instructional courses. The renewal for the Berkeley Symposium on Economics, Statistics, and Computation will provide a focal point for the dissemination of research, interaction between researchers in statistics and econometrics, demonstration of software, and instruction of graduate students. The intensive use of computation in solving statistical and econometric problems promises to yield solutions to otherwise intractable problems. In statistics, for example, bootstrap and related sampling techniques provide a general method for accurately approximating the sampling distribution of estimators and test statistics. In econometrics, simulation methods are used to cover come the intractability of such complex models as high-dimensional multinominal choice. These are two of many examples. Robust inference methods, graphical methods, Bayesian analyses, nonlinear time series models and semi- and non-parametric methods also draw heavily on new computational approaches. This project continues a program of annual symposia begun in 1993 with a combination of instructional surveys and research presentations, and a mix of junior and senior scholars. This grant supports symposia in each of the years 1998, 1999, and 2000. Each symposium will be about five days in length, and will be held in facilities at the Econometrics Laboratory and the Department of Economics at the University of California at Berkeley. The tentative topics are: Nonlinear & Non-Stationary Time Series Models; Statistical Software: Architecture, Languages, Algorithms; and Bayesian and Classical Econometrics: Computational Considerations. A secondary, but substantial, administrative task would be installation and testing of software provided by conference participants, and supervision of the documentation and release of this software through the Econometrics Laboratory Software Archives. ??