摘要
在panel模型中,如果方差分量已知,对回归系数的检验,存在一致最优功效检验.但往往方差分量未知,这时采用的方法就是用它们的估计代替它们.不同的方差分量的估计,就得到不同的检验统计量.这些检验统计量的分布都是未知的,在小样本情况下,很难控制它们的检验水平.本文采用广义p值的方法,给出了一种精确的检验.模拟结果显示,这种检验能很好的控制检验水平,并且有更高的检验功效.同时,本文利用广义置信域的方法给出了回归系数的广义置信球.
In panel model, there is uniformly most powerful test for testing regression coefficient when variance components are known. But the variance component are often unknown, so we usually substitute them with their estimates. Different test statistics come from different estimating methods, and their test sizes are unknown because of their unknown distributions. In this paper, we use the method of generalized p value to construct an exact test. Simulation shows that the test of generalized p value is more powerful than other tests with its testing size approximating nominal level. In the paper, we also construct a generalized confidence sphere with the concept of generalized confidence region.
出处
《应用数学学报》
CSCD
北大核心
2008年第2期367-373,共7页
Acta Mathematicae Applicatae Sinica
基金
北京市属市管高等学校人才强教计划资助项目．
