IV估计框架下模型设定检验问题的讨论

统计研究 ›› 2012, Vol. 29 ›› Issue (2): 80-87.

IV估计框架下模型设定检验问题的讨论

王美今等

出版日期:2012-02-15 发布日期:2012-02-10

Discussion about Model Specification Testing under Instrumental Variable Estimation

Wang Meijin et al.

Online:2012-02-15 Published:2012-02-10

摘要/Abstract

摘要： IV估计框架下各种统计量的良好性质依赖于相应的模型设定，如果这些模型设定未能得到数据的支持，其统计推断结论将是不可靠的。如判定计量经济模型是否存在内生性的Hausman检验，实证研究中同一问题的检验结果可能大相径庭。如何通过合理的模型设定检验程序来获得模型参数科学、可靠的估计结果和检验结论呢？本文讨论了工具变量估计框架下的各种模型设定检验问题，明确了各个检验统计量的适用条件及其逻辑联系，给出了工具变量估计框架下模型设定检验的一般步骤。

关键词: IV估计, 模型设定检验

Abstract: The good properties of various statistics under the instrumental variable estimation framework depend on the corresponding model assumptions. If these assumptions are not supported by the data, the statistical inference conclusions will be unreliable. For example, Hausman test is a classical procedure for testing endogeneity. However, in empirical research the results of this test may be in opposite directions. How to carry out the model specification testing in the presence of possible endogenous variable problem in empirical research? In this paper, we discuss a variety of model specification issues, and declare the conditions of the various test statistics and the precedence relationship explicitly. On this basis, we put up a general testing procedure for model specification.

Key words: Instrumental Variable Estimation, Model Specification Testing

王美今等. IV估计框架下模型设定检验问题的讨论[J]. 统计研究, 2012, 29(2): 80-87.

Wang Meijin et al.. Discussion about Model Specification Testing under Instrumental Variable Estimation[J]. , 2012, 29(2): 80-87.