Professional Documents
Culture Documents
STAT/MATH 571A
Professor Lingling An
16
Outline
• Measures of multicollinearity
• Remedies
16-1
Informal diagnostics-
multicollinearity
16-2
Measures of multicollinearity
16-3
Variance Inflation Factor
s
• for b∗ = ( ss1y b1, ..., p−1
sy bp−1 ) it can be shown
that σ 2{b∗} = (σ ∗)2rXX −1
16-4
Variance Inflation Factor
−1
• VIFk is the the kth diagonal element of rXX
2=
• special case: if only two X variables, R1
2 = r2
R2 12
VIFk
P
VIF =
p−1
• Equivalent to VIF
• Tolerance(TOL) = 1/VIF
16-6
Example Page 256
• X2 is thigh circumference
• X3 is midarm circumference
16-7
Output
Analysis of Variance
Sum of Mean
Source DF Squares Square F Value Pr > F
Model 3 396.98461 132.32820 21.52 <.0001
Error 16 98.40489 6.15031
Corrected Total 19 495.38950
Parameter Estimates
Parameter Standard
Variable DF Estimate Error t Value Pr > |t|
Intercept 1 117.08469 99.78240 1.17 0.2578
skinfold 1 4.33409 3.01551 1.44 0.1699
thigh 1 -2.85685 2.58202 -1.11 0.2849
midarm 1 -2.18606 1.59550 -1.37 0.1896
16-8
SAS code for VIF and TOL
Parameter Standard
Variable DF Estimate Error t Value Pr > |t| Tolerance
Parameter Estimates
Variance
Variable DF Inflation
Intercept 1 0
skinfold 1 708.84291
thigh 1 564.34339
midarm 1 104.60601
16-9
Some remedial measures
bR = (rXX + cI)−1rY X
16-11
Choice of c
16-12
SAS Commands
options nocenter ls=72;
goptions colors=(’none’);
data a1;
infile ’C:\datasets\Ch07ta01.txt’;
input skinfold thigh midarm fat;
16-13
Ridge Trace
16-15
Obs _RIDGE_ skinfold thigh midarm
16-16
Obs _RIDGE_ _RMSE_ Intercept skinfold thigh mida
16-17
Conclusion
Notes:
16-18
• Major drawback: ordinary inference proce-
dures don’t work so well.
• ridge.sas
16-19