A New Method of Projection-Based Inference in GMM with Weakly Identified Nuisance Parameters
A New Method of Projection-Based Inference in GMM with Weakly Identified Nuisance Parameters
复制标题
弱识别干扰参数的 GMM 中基于投影的推理新方法
DOI：
--
复制
发表时间：
2008
期刊：
影响因子：
--
通讯作者：
E. Zivot
中科院分区：
文献类型：
--
作者：
Saraswata Chaudhuri;E. Zivot
对参数子集的基于投影的测试是有用的，因为当难以估计讨厌的参数或它们的识别状态有问题时，它们不会过度拒绝真实的参数值。然而，他们也经常被批评过于保守。我们克服了这种保守性，引入了一个新的基于投影的测试，是比传统的基于投影的测试更强大。当所有的参数都被识别时，新的测试甚至渐近等价于相关的基于插件的测试。扩展到弱识别参数的模型表明，新的测试是不占主导地位的相关插件为基础的测试。
Projection-based tests for subsets of parameters are useful because they do not over-reject the true parameter values when either it is difficult to estimate the nuisance parameters or their identification status is questionable. However, they are also often criticized for being overly conservative. We overcome this conservativeness by introducing a new projection-based test that is more powerful than the traditional projection-based tests. The new test is even asymptotically equivalent to the related plug-in-based tests when all the parameters are identified. Extension to models with weakly identified parameters shows that the new test is not dominated by the related plug-in-based tests.