Normal Linear Model Notes

# Assume p q so model under h0 is reduced compared to

This preview shows page 1. Sign up to view the full content.

This is the end of the preview. Sign up to access the rest of the document.

Unformatted text preview: Model with q − 1 covariates. H1 : β = β0 = (β1 , β2 , . . . , βp ). Model with p − 1 covariates. Assume p > q so model under H0 is reduced compared to model for H1 . Fit model under H0 (H1 ) and compute sum of squares S0 (S1 ) Since model under H1 has more parameters S0 > S1 , Test H0 with F= (S0 − S1 ) S1 (D0 − D1 ) D1 / = / ∼ F (p−q , N −p) (p − q ) (N − p) (p − q ) (N − p) . Reject H0 if Fobs > F(1−α) (p − q , N − p) UNM Residuals ˆ ˆ Based on ei = Yi − Yi = Yi − xiT b = obs − ﬁtted ˆ Standardized residuals: ri = σ(1−ei )1/2 ; i = 1, 2, . . . , N ˆ hii where hi i is i − th element on the diagonal of H = X (X T X )−1 X T . Plots: Normal probability plots, independence, homoscedasticity, constant variance, etc. High leverage or inﬂuential: hii > 2(p/N ), DFITSi = ri (hii /(1 − hii )1/2 . Cook’s distance: Di = (1/p)(DFITS )2 . Large values means ”inﬂuential”. 1 Di = (b − b(i ) )T X T X (b − b(i ) ) p and b(i ) is the estimate of β without observation i . UNM...
View Full Document

## This note was uploaded on 01/27/2014 for the course STAT 574 taught by Professor Gabrielhuerta during the Fall '13 term at New Mexico.

Ask a homework question - tutors are online