Multicolinearity in logistic regression using R

Asked Apr 08 '15 at 00:26

Active Jun 11 '17 at 11:48

Viewed 627 times

Upon performing binary logistic regression, I have found VIF, using R programming, as follows:

             GVIF Df  GVIF^(1/(2*Df))
agem     2.213242  3        1.141576
eduM     2.842857  3        1.190216
eduF     2.576725  3        1.170877
ageC     1.315301  1        1.146866
diarrhea 1.031031  1        1.015397
uweight  1.129919  1        1.062977
fever    1.033433  1        1.016579
res      1.341470  1        1.158218
dis      1.440215  6        1.030866
WI       2.610752  4        1.127446
nlc      2.407934  3        1.157730

Based on those results, should I remove agem, eduM, eduF, WI and nlc for multi-collinearity? Or do I need to apply another approach? Could anybody help me?

edited Jun 11 '17 at 11:48

kjetil b halvorsen

63,378
26
142
467

asked Apr 08 '15 at 00:26

JRK

1

Perhaps, you will find [this answer](http://stats.stackexchange.com/a/141613/31372) of mine helpful. – Aleksandr Blekh Apr 08 '15 at 00:45
Three words: elastic net regression – Sycorax Apr 08 '15 at 01:49

Multicolinearity in logistic regression using R

0 Answers0