Bookmark and Share

Notice: On April 23, 2014, Statalist moved from an email list to a forum, based at

[Date Prev][Date Next][Thread Prev][Thread Next][Date Index][Thread Index]

Re: st: Variable Inflation Factor (VIF) for non-linear models

From   Nick Cox <>
To   "" <>
Subject   Re: st: Variable Inflation Factor (VIF) for non-linear models
Date   Sun, 27 Oct 2013 13:44:19 +0000

The VIFs (plural) depend on the predictors alone and are defined for a
regression model.

-xtpoisson- is a regression model, wide sense, but the machinery is
different. I don't think there is a definition of VIF that is modified
for such models: we are no longer doing any kind of least squares, in

Nothing stops you doing a multiple regression (with _any_ response) to
get VIFs. The folklore I am aware of implies that multicollinearity
shown up that way could be a severe problem for any kind of model,
regardless of the link function  or the exact machinery. On the whole,
multicollinearity seems less of a problem to me than people fear,
provided you show some sensitivity to not including predictors that
have very similar meaning and to the consequences of powers and
products as predictors.

On 27 October 2013 10:42, James Bernard <> wrote:
> Hi all,
> I would like to know if after a count model (xtpoisson command) Stata
> can give me VIF of the predictors?
> I looked for it in the post-estimation section of -xtpoisson-, but
> could not find any thing. Is it a theoretical issues that VIF can not
> be calculated for count models? Or, is it a software issues?
> Is there any substitute for it in log-linear models? I don't want to
> use correlations to justify absence of multicollinearity.
> Thanks,
> James
> *
> *   For searches and help try:
> *
> *
> *
*   For searches and help try:

© Copyright 1996–2018 StataCorp LLC   |   Terms of use   |   Privacy   |   Contact us   |   Site index