线性回归模型中的渐近最优性

Asymptotic Optimality in Linear Regression

摘要: 对线性回归中系数的一类估计给出了理论上的最优均方误差.证明了渐近意义下最小二乘估计和lasso估计均不具有最优均方误差性.最后给出了一个具有渐近最小均方误差的回归估计

Abstract: Theoretical optimal Mean Squared Error(MSE)of a class of regression estimators was provided.It was proved that neither the least squares estimator nor the lasso estimator possesses the asymptotic optimality.A new regression estimator which has the asymptotic optimality was also given.