Jackknifing Weighted Least Squares Estimators
在异方差线性回归模型中，证明了普通刀切法估计加权最小二乘估计量的渐近协方差矩阵是不一致的，提出了一种修正刀切法并证明其一致性，同时讨论了有限样本表现。
SUMMARY In a heteroscedastic linear regression model, the ordinary jackknife estimator of the asymptotic covariance matrix of the weighted least squares estimator is proved to be inconsistent. A modified jackknife procedure is proposed and shown to produce consistent estimators of the asymptotic covariance matrix. Finite sample performances of the jackknife, the modified jackknife and the customary S method are discussed. Some empirical results are also presented.