4

I am running a logistic regression in R and am attempting to determine if multicollinearity is a problem with my model.
When I run vif() on my final model, I get GVIF and GVIF^1/(2*Df) columns. From what I have read GVIF^1/(2*Df) is what I should use to assess muticollinearity, but I have been unable to determine what values I should use as a cut-off point.

Any help would be greatly appreciated.

kjetil b halvorsen
  • 63,378
  • 26
  • 142
  • 467
user1753987
  • 173
  • 1
  • 7
  • 2
    Before modelling, you should look at correlations between your (assumed) independent variables. I think you should rephrase your question to "how to identify multicolinearity" and ask on crossvalidated.com, instead. – Brandon Bertelsen Oct 17 '12 at 17:27
  • The same issue is discussed here with some more conclusions: http://stats.stackexchange.com/questions/70679/which-variance-inflation-factor-should-i-be-using-textgvif-or-textgvif/96584#96584 – MsGISRocker May 06 '14 at 12:05

0 Answers0