Re: st: collinearity in categorical variables

[David Hoaglin <dchoaglin@gmail.com>]

Hi, Mitchell.

Please say more about the information on the internet that suggests
that diagnosis of collinearity is not appropriate for categorical
variables.

For predictors in multiple regression, diagnosis of collinearity
should go into more detail than looking at VIFs.  I have found
-coldiag2- (from SSC, as I recall) useful.  I have not used -perturb-.

In a model, a categorical variable will be represented by a set of
indicator variables, one indicator for each category except the
reference category.  By construction, those indicator variables for
any one categorical variable have zero collinearity.  Some degree of
collinearity might arise from relations between two or more
categorical variables.  Even then, it should seldom be serious enough
to cause problems.

David Hoaglin

On Thu, Apr 25, 2013 at 2:17 PM, Mitchell F. Berman <mfb1@columbia.edu> wrote:
> Stata Users:
>
> We are working on a logistic regression model with both continuous and
> categorical independent variables.
>
> I'm familiar with collin to generate VIF and condition index.  But my
> impression and information on the internet suggests that collin is not
> appropriate for categorical variables.
>
> What would people use to evaluate collinearity (probably not the correct
> term in this case) for categorical variables.
>
> Some people have referenced a Stata module called perturb by John Hendrickx.
> But I don't know that this has been vetted by the Stata community or
> statistical community in general.  And I've not been able to locate a
> functioning download site for perturb.
>
> Thanks in advance.
>
> Mitchell
*
*   For searches and help try:
*   http://www.stata.com/help.cgi?search
*   http://www.stata.com/support/faqs/resources/statalist-faq/
*   http://www.ats.ucla.edu/stat/stata/