The Local Power of the Tests of Overidentifying Restrictions
推导了结构方程中过度识别约束的三种检验（似然比、线性化似然比、拉格朗日乘子）的局部功效函数，发现它们在1/T阶上相同且不依赖于k类估计量，并讨论了未校正检验的尺寸和功效性质。
The local power function of the size-corrected likelihood ratio, linearized likelihood ratio, and Lagrange multiplier tests of overidentifying restrictions on a structural equation is the same to the order 1/T. Moreover, this local power function doe s not depend on the k-class estimator used in the calculation of the test statistic. When the author does not use size-corrected tests, a degrees of freedom corrected likelihood ratio test seems to have the best size and power properties. Finally, the implicit null hypothesis of these tests indicates that they can be interpreted as testing the validity of the structural specification of the equation against any other identified structural equation that encompasses the original e quation. Copyright 1988 by Economics Department of the University of Pennsylvania and the Osaka University Institute of Social and Economic Research Association.