摘要
一、引言考虑多重线性回归模型Y=Xβ+ε,(1)其中,Y=(y_1,…y_n_)′为 n×p 观察矩阵,X=(x_1,…,x_n)′为 n×(k+1)列满秩设计矩阵,β=(β_0,β_1,…β_k)′为(k+1)×p 未知参数矩阵,ε=(ε_1,…ε_n)′为 n×p 随机误差矩阵,ε_1…,ε_n 相互独立.
In this paper,we discuss the relationship between the collective influences of multiplecases and the influence of each case on regression analysis in multivariate linear regression,that is,the factorization of the collective influences of cases.We also discuss the magnitudeof influence of case i on the fitted value at case j,that is,the cross influence of data;and theupper bound of the cross influence is obtained.Furthermore,we reveal the role of off-diago-nal elements of the hat matrix in the factorization and the cross influence of data.
出处
《系统科学与数学》
CSCD
北大核心
1993年第3期229-232,共4页
Journal of Systems Science and Mathematical Sciences
基金
山西省自然科学基金资助课题
