MATH 226 - 9.19.2007

MATH 226 - 9.19.2007 - i y-bar) 2- (b + b 1 x i b b 1...

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

View Full Document Right Arrow Icon
MATH 226 – September 19, 2007 Find numerical quantities that describe whether the linear model is good or not o raw variation Σ(y i – y-bar) 2 o variation after fitting the line Σ(y i – ŷ i ) 2 o percent of variation explained by the regression line: R 2 = (Σ(y i – y-bar) 2 - Σ(y i – ŷ i ) 2 )/ Σ(y i – y-bar) 2 Coefficient of determination The sample correlation is defined by o ρ = [Σ((x i – x-bar)(y i – y-bar))]/[sqrt((Σ(y i – y-bar) 2 - Σ(y i – y-bar) 2 )] Fact 1 -1≤ ρ ≤ 1 Fact 2 If y = b 0 + b 1 x then ρ = 1 if b 1 >0 and ρ = -1 if b 1 <0 ρ = [Σ((x i – x-bar)(b 0 + b 1 x i – b 0 – b 1 x-bar))]/[sqrt((Σ(y
Background image of page 1
This is the end of the preview. Sign up to access the rest of the document.

Unformatted text preview: i y-bar) 2- (b + b 1 x i b b 1 x-bar) 2 )] = [b 1 (x i x-bar) 2 ]/[sqrt(b 1 2 )* (x i x-bar) 2 (x i x-bar) 2 ] = b 1 /|b 1 | Fact 3 Correlation between slope and equation for rho o CAUTIONS Correlations never prove causal relations A linear regression line should only be used over the range of the x-values HOMEWORK o QUESTION 1 b prove your claim o QUESTION 4 c use the TI-89, copy the graphs, normal values on p.89 of text y-bar y i ybar y i i...
View Full Document

This note was uploaded on 04/07/2008 for the course MATH 226 taught by Professor Daepp during the Fall '07 term at Bucknell.

Ask a homework question - tutors are online