U.S.- Austria Cooperative Research on Asymptotic Bayesian Analysis and Order Selection

该奖项支持耶鲁大学考尔斯经济学研究基金会的Peter Phillips教授与奥地利维也纳技术大学计量经济学与运筹学研究所的Werner plobberger教授合作。他们计划在计量经济学的几个研究方向上合作。首先，他们将对可能非平稳的时间序列进行贝叶斯方法的渐近分析。此外，他们将开发广义贝叶斯模型选择标准，具体参考参数方法中的顺序选择问题。此外，他们打算编写计算机软件来实现他们开发的方法，并将其应用于美国、奥地利和其他几个国家的宏观经济数据集。近年来，科学界对贝叶斯方法的兴趣急剧增长，并影响了统计学和计量经济学的许多不同领域。许多计量经济学家发现贝叶斯方法特别适合于计量经济学中的时间序列分析和推理问题，这两个领域正在与dr。菲利普斯和普莱伯格。他们的一些研究将有助于改进经济时间序列的建模。他们从事的另一个项目在基于数据的模型选择原则下发展假设检验方面具有重要的应用。一些在经济学中非常有趣的假设，比如自回归单位根的存在，在不同的模型中保持不变。这些研究人员计划利用这种不变性，这样他们的“贝叶斯模型”标准就可以用来选择一个系统来对单位根假设进行明确的检验。

This award supports Professor Peter Phillips of the Cowles Foundation for Research in Economics, Yale University, to collaborate with Professor Werner Ploberger of the Institute of Econometrics and Operations Research of the Technical University of Vienna in Austria. They plan to work jointly on several lines of research in econometrics. First, they will carry out asymptotic analysis of Bayesian methods for possibly non-stationary time series. Also, they will develop generalized Bayes model selection criteria with specific reference to problems of order selection in parametric methods. In addition, they intend to write computer software to implement the methods they develop and to apply them to macroeconomic data sets for the USA, Austria, and several other countries. Scientific interest in Bayesian methods has grown sharply over recent years and has affected many different fields of statistics and econometrics. Many econometricians find the Bayesian approach particularly well-suited to problems of time series analysis and inference in econometrics, two areas being explored in the collaboration of Drs. Phillips and Ploberger. Some of their research will contribute to improved modelling of economic time series. The other project they are engaged has an important application in the development of hypothesis tests under data-based model selection principles. Some hypotheses of considerable interest in economics, like that of the presence of an autoregressive unit root, remain invariant across different models. These investigators plan to exploit this invariance so that their "Bayes model" criteria can be used to select a system to conduct an explicit test of the unit root hypothesis.