[Date Prev][Date Next][Thread Prev][Thread Next][Date index][Thread index]

Re: st: Multicollinearity and logit

From   Richard Williams <>
To, <>
Subject   Re: st: Multicollinearity and logit
Date   Tue, 18 Mar 2008 18:30:57 -0500

At 07:37 AM 3/18/2008, Herve STOLOWY wrote:
Dear Statalisters:

I have a question concerning multicollinearity in a logit regression.

How could I check multicollinearity? I tried several things.

- Correlation matrix: several independent variables are correlated.

- Logit regression followed by -vif, uncentered-.  I get high VIFs
(maximum = 10), making me think about a high correlation.

- OLS regression of the same model (not my primary model, but just to
see what happens) followed by -vif-: I get very low VIFs (maximum = 2).

- -collin- (type findit collin) with the independent variables: I get
very low VIFs (maximum = 2).

What is better? I am puzzled with the -vif, uncentered- after the logit
which returns very high VIFs.
I'm surprised that -vif- works after logit; it is not a documented post-estimation command for logit. Given that it does work, I am surprised that it only works with the -uncentered- option. I wonder if this is a bug and if the results mean anything.

I always tell people that you check multicollinearity in logistic regression pretty much the same way you check it in OLS regression. Multic is a problem with the X variables, not Y, and does not depend on the link function. So, the steps you describe above are fine, except I am dubious of -vif, uncentered-.

Richard Williams, Notre Dame Dept of Sociology
OFFICE: (574)631-6668, (574)631-6463
HOME: (574)289-5227
EMAIL: Richard.A.Williams.5@ND.Edu

* For searches and help try:

© Copyright 1996–2019 StataCorp LLC   |   Terms of use   |   Privacy   |   Contact us   |   What's new   |   Site index