Scientific Workstations for Research in Computationally- Intensive Econometric Methods

过去十年，随着计算机能力的爆炸性增长，计量经济学开始转向使用计算机密集型方法，以揭示更多数据，并强化得出的推论。在半参数和非参数估计、参数和非参数模型中重采样方法和交叉验证的使用、推理的模拟方法、动态随机规划模型的估计和贝叶斯方法等领域的创新尤为强烈。将这些方法转化为应用软件和经验计量经济学一直参差不齐，部分原因是计算能力尚未普遍获得。该项目的主要目标是加快对计算密集型方法和应用的研究，并鼓励更快地测试、记录和传播算法。它计划为一个从事计算密集型计量经济学工作的研究人员联盟提供Unix工作站，以及一个通用文件服务器，以方便访问软件和数据。除了这份核心提案中描述的伯克利的设施外，还有其他几个机构的学者提出的子提案，他们也在同一领域工作。伯克利文件服务器将不仅为这些学者，而且也为本提案中未包括的其他人提供访问专门软件和数据集的机会。这项技术是为伯克利研究项目和联盟计算设施计划的，非常适合计算需求。它有望在理论计量经济学建模方法方面取得重大进展。

The explosion in computer capacity in the past decade has been accompanied by a reorientation of econometrics toward using computer-intensive methods to reveal more about data and sharpen the inferences drawn. Innovations have been particularly intense in the areas of semiparametric and nonparametric estimation, use of resampling methods and cross-validation in parametric and nonparametric models, simulation methods for inference, estimation of dynamic stochastic programming models, and Bayesian methods. Translation of these methods into applications software and empirical econometrics has been uneven, partly because computational capacity has not been generally available. The primary objective of this project is to accelerate research on computationally intensive methods and applications, and to encourage more rapid testing , documentation, and dissemination of algorithms. It plans to provide UNIX workstations for a consortium of researchers working on computationally intensive econometrics, and a common file server to facilitate access to software and data. In addition to the facilities at Berkeley, described in this core proposal, there are subproposals from scholars at several other institutions who are working in the same area. The Berkeley file server will provide access not only for these scholars, but also for others not included in this proposal, to specialized software and data sets. The technology which is planned for both the Berkeley research programs and the consortium computing facility is well suited to the computing demands. It offers the promise of significant advances in theoretical econometric modelling methods.