回归系数最小二乘估计的Bootstrap估计

Bootstrap Approximation of the LS Estimate of the Regression Parameters

本文对线性模型的回归系数β的线性函数α~Tβ的最小二乘估计α~T(?)建立了一种新的bootstrap逼近,给出了逼近的相合性定理,得到了o(n^(-1/2))的逼近速度。

In this paper, a new bootstrap approximation of the LS estimate of a linear function αTβ of the regression parameters (1 in a linear model is established. It is proved that the approximation is consistent and asymptotically more accurate than approximation by the limiting normal distribution.