1

I have a cross-sectional dataset which I obtained from panel data. All the variables are on macroeconomics with n=75. I want to check my variables for multicollinearity using VIF. I got the following results :

mean VIF: 1.62 mean VIF: 1.55

I do not know what values to use in order to compare VIF. My professor's lecture notes say that if mean VIF>1 then it is evidence for multicollinearity. However, I searched for more information on this issue and I understood that there no appropriate value but it depends of the kind of the research.

How do I know which level is appropriate for my research?

Marquis de Carabas
  • 1,667
  • 19
  • 31
Ant
  • 323
  • 2
  • 13

1 Answers1

1

Personally, I think that using condition indexes and proportion of variance explained is a much better way to diagnose approximate collinearity. That's what my dissertation showed. VIFs don't give as much information and are possibly not as good at the diagnose of how much collinearity is enough to cause problems.

Peter Flom
  • 94,055
  • 35
  • 143
  • 276
  • Thank you for your answer. I read a little bit for condition indexes and I am going to apply it. However, probably there is problem with downloading the "collin" command. I can't download it and I saw that also other users had the same problem. – Ant Mar 08 '15 at 16:09