Tests on linear model
Description
We consider the regression model y=Xβ+ε with ε∼N(0,σ2Ω)
The GLS estimator of β is ˆβ=(X′Ω−1X)−1X′Ω−1y and its covariance matrix is σ2(X′Ω−1X)−1=σ2W
The BLUE of σ2 is s2=ˆεΩ−1ˆε/(n−k)=RSS/(n−k), where ˆε=y−Xˆβ=(I−X(X′Ω−1X)−1X′Ω−1)y
Test on a single coefficient
H0:βi=α, H1:βi≠α
t=ˆβi−α√s2wiit∼T(n−k)Test on multiple coefficients
H0:Rβ=α, H1:Rβ≠α, R∼m×k
f=(Rˆβ−α)′(s2RWR′)−1(Rˆβ−α)/m=(Rˆβ−α)′(RWR′)−1(Rˆβ−α)/mRSS/(n−k),f∼F(m,n−k)