摘要
在制导工具的系统误差的分离中,主成份估计得到了大量的应用。针对主成份估计进行改进,提出把主成份估计的后pr个主成份不是取消,而是进行一个非均匀的压缩估计。可以证明,改进之后得到的估计在MSE和SSE的意义下均优于主成份估计,从而在预报性能和参数估计的准确度上都得到优良的结果。
We have made heavy use of principal components regression estimation from a large number of highly correlated predictors in the system error separation of guidance facility. Usually the last p-r principal component is discarded, but in this paper we develop a new estimation based on partial shrinkage. The new estimation provide smaller MSE and SSE than principal component esitimation,so it can be looked upon as a good method for both predicting (or control) and the accurate estimation of predicting a response variable.
出处
《导弹与航天运载技术》
北大核心
2005年第3期43-45,共3页
Missiles and Space Vehicles
关键词
复共线性
有偏回归
容许估计
部分压缩
Multicollinearity
Biased regression
Admissible estimation
Partial shrinkage
