Processing math: 100%

Tests on linear model

Description

We consider the regression model y=Xβ+ε with εN(0,σ2Ω)

The GLS estimator of β is ˆβ=(XΩ1X)1XΩ1y and its covariance matrix is σ2(XΩ1X)1=σ2W

The BLUE of σ2 is s2=ˆεΩ1ˆε/(nk)=RSS/(nk), where ˆε=yXˆβ=(IX(XΩ1X)1XΩ1)y

Test on a single coefficient

H0:βi=α, H1:βiα

t=ˆβiαs2wiitT(nk)

Test on multiple coefficients

H0:Rβ=α, H1:Rβα, Rm×k

f=(Rˆβα)(s2RWR)1(Rˆβα)/m=(Rˆβα)(RWR)1(Rˆβα)/mRSS/(nk),fF(m,nk)