U.S.-India Collaborative Research: Small-area Estimation Problems

INT 9505197 Lahiri技术描述：该项目为内布拉斯加大学的Partha-sarathi Lahiri提供部分支持，以便与合作者Rahul Mukherjee在加尔各答的印度管理学院进行小面积估计的统计研究。 Lahiri博士还将与加尔各答印度统计研究所的T. J. Rao合作，探索调查抽样的新领域。 大规模国家抽样调查中使用的估计数在国家以下一级的表现不佳。 私家侦探建议探索使用经验贝叶斯和分层贝叶斯方法在开发估计产生可靠的小面积统计。 合作者将调查这种方法，Lahiri已被证明是上级的最佳线性估计的小面积的手段，持有的一般混合模型与随机误差方差分量。 他们将扩展经验贝叶斯和分层贝叶斯方法，以找到离散数据的小区域特性的估计。 适用范围：小面积估计的研究对于在州和县一级以及对于人口的亚群体产生统计数据是重要的。 在印度和美国，各联邦和地方政府需要这些统计数据来制定政策和分配资金。 国外合作者在贝叶斯统计、高阶渐近和调查抽样等领域做了杰出的工作。 慕克吉教授最近在渐近领域的进展将非常有助于证明提案中提到的困难的渐近结果。 这一新的合作活动可望带来许多互利。 ***

INT 9505197 Lahiri Technical Description: This project provides partial support for Partha-sarathi Lahiri, University of Nebraska, to conduct statistical research in small-area estimation at the Indian Institute of Management, Calcutta with collaborator Rahul Mukherjee. Dr. Lahiri will also work with T.J. Rao of the Indian Institute of Statistics, Calcutta to explore new areas in survey sampling. Estimators used in large scale national sample surveys perform poorly at the sub-national level. The P.I. proposes to explore the use of empirical Bayes and hierarchical Bayes methods in developing estimators to generate reliable small area statistics. The collaborators will investigate whether this method which Lahiri has shown to be superior to the best linear estimators for small area means, holds for the general mixed model with random error variance components. They will extend the empirical Bayes and hierarchical Bayes methods to find estimators of small area characteristics for discrete data. Scope: Research in small area estimation is important for generating statistics at the state and county levels and for subgroups of the population. In India and the U.S., such statistics are needed by various federal and local governments for policy making and allocation of funds. The foreign collaborator has done outstanding work in the areas of Bayesian statistics, higher order asymptotics and survey sampling. Professor Mukherjee's recent advances in the area of asymptotics will be very helpful in proving difficult asymptotic results mentioned in the proposal. Many mutual benefits may be expected to result from this new collaborative activity. ***