[ECON20141711]

数量经济学研讨会

     
 
  报告题目：Second Order Properties of Empirical Likelihood Ratio Tests for General Parameter Hypothesis Testing Problems  
  报告人：马骏 博士  
  报告时间：2014年09月30日下午12:30-13:30  
  报告地点：明德主楼729  
 
 
  报告摘要： This paper focuses on parameter hypothesis testing problems in the inferential framework of moment restriction models using empirical likelihood ratio tests. We consider possibly overidentified models and tests of general nonlinear restrictions on the parameter under the null hypothesis. We show that empirical likelihood ratio tests for the family of testing problems defined by linear parametric restrictions are Bartlett correctable. This is an important generalization of the theorems on Bartlett correctability of empirical likelihood ratio tests by Chen and Cui (2006, Biometrika, 93, 215-220) and Chen and Cui (2007, Journal of Econometrics, 141, 492-516) because it allows for much more interesting parameter testing problems and is potentially useful in many econometric applications. Based on this finding, we propose empirical likelihood ratio tests with second order refinement via Bartlett correction or adjusted empirical likelihood. Our second theoretical finding is that if the parametric restriction by the null hypothesis is nonlinear, the second order refinement results established for linear testing problems cannot be extended to these cases in general. In the Monte Carlo experiment, we focus on parameter testing problems for which Bartlett correctability of empirical likelihood ratio tests have not been established in the literature and show that our new testing methods have good finite sample properties.  
 
 
  报告人简介：  
  马骏，英属哥伦比亚（UBC）大学经济学博士。他的主要研究方向为计量经济学，主要研究兴趣为调整的经验似然（adjusted empirical likelihood）以及经验似然的高阶渐近性质等。  
 
 
      数量经济教研室               运筹学与数量经济研究所               365体育官方唯一入口               2014年09月      
 
      为了加强与国内外高水平数量经济学学者的交流和合作，更好地促进数量经济学与院内其他学科之间的合作交流，数量经济学教研室将举办数量经济学研讨会。研讨会侧重于理论计量经济学、应用计量经济学及数理经济学的最新研究成果，欢迎各位学者参加或报告。如果您有研究成果想要报告，请与数量经济学教研室章永辉博士联系（yonghui.zhang@hotmail.com）。