奇异G-M模子参数估计的相对效率.pdfVIP

* ˆ cov(β ) F e( β)= ˆ cov(β) F G-M 2001 Gauss-Markoff 2 Y=X β + ε E( ε )=0, cov( ε )= σ Σ (1) Y n ×1 , X n ×p , r(x) ≤p , β p ×1 ε 2 n ×1 ,σ , Σ n × n r=p, Σ[3][9] (LSE)BLUE rp, Σ, 1Rao.C.R 1991 β BLUE . * + + + + β =(X T X) X T Y A A Moore-penrse T= Σ+X UX , U r(T)=r( ΣMX) , * + + + + + + 2 cov( β )=(X T X) X T ΣT X(X T X) ⋅σ 2 + ˆ U=kI,k0. β LSE β=(X X) X Y. + 2 ˆ cov( β )=(X X)