Approximating the Approximate Slopes of LR, W, and LM Test Statistics
用泰勒级数方法推导LR、W和LM检验统计量近似斜率的近似公式，应用于线性模型中非标量协方差矩阵下的系数约束检验，有助于理解Wald检验过度拒绝的原因。
A Taylor series approach is used to derive approximations to the approximate slope of LR, W, and LM test statistics. These can be useful when the approximate slopes themselves are complicated. The results are applied to the test of linear coefficient restrictions in the linear model with a nonscalar covariance matrix. They can be used to find cases where the Wald test may badly over-reject in this model. The results shed some new light on the over-rejection that has been noted in the special cases of AR(1) disturbances and the SURE model.