摘要
考虑实际回归问题中存在更多受约束条件的情况,提出了带约束的统一几乎无偏估计类,统一了常见的具有线性约束的回归模型的几乎无偏估计,进一步的研究给出了在均方误差和均方误差矩阵意义下,带约束的统一几乎无偏估计优于一般带约束的最小二乘估计的充分条件和椭球范围.
This paper considers the problem of regression with some restrictions in application, the unified almost unbiased estimator for the vector of parameters in a linear regression model with additional linear restrictions is introduced,which is unified expression for common almost unbiased estimators in linear restricted regression model.Furthermore,sufficient condition and ellipsoid bound are established for the new estimator to be better than the ordinary restricted least squares estimator in terms of mean squares error and mean squares error matrix, respectively.
出处
《系统科学与数学》
CSCD
北大核心
2011年第1期105-113,共9页
Journal of Systems Science and Mathematical Sciences
基金
国家自然科学基金(11001286)
关键词
统一有偏估计
统一几乎无偏估计
均方误差
均方误差矩阵
Unified biased estimator
unified almost unbiased estimator
mean squares error
mean squares error matrix
