笔记:多元回归OLS估计量有效性证明
OLS估计量 β ^ = C Y \hat\beta=CY β ^ = C Y 假设 β ∗ \beta^* β ∗ 是其他方法得到的关于无偏估计量: β ∗ = C ∗ Y \beta^*=C^*Y β ∗ = C ∗ Y 其中 C ∗ = C + D = ( X ′ X ) − 1 X ′ + D C^*=C+D=(X'X)^{-1}X'+D C ∗ = C + D = ( X ′ X ) − 1 X ′ + D , D D D 为固定矩阵。于是 β ∗ = C ∗ Y = C ∗ X β + C ∗ μ E ( β ∗ ) = C ∗ X β + C ∗ E ( μ ) = C ∗ X β \beta^*=C^*Y=C^*X\beta+C^*\mu \\E(\beta^*)=C^*X\beta+C^*E(\mu)=C^*X\beta β ∗ = C ∗ Y = C ∗ X β + C ∗ μ E ( β ∗ ) = C ∗ X β + C ∗ E ( μ ) = C ∗ X β 根据无偏性要求 C ∗ X = I C^*X=I C ∗ X = I 。而 C ∗ X = ( X ′ X ) − 1 X ′ X + D X C^*X=(X'X)^{-1}X'X+DX C ∗ X = ( X ′ X ) − 1 X ′ X + D X 所以需要 D X = 0 DX=0 D X