R help - Aug 2004 - mutlicollinearity and MM-regression

Dear R users,

Usually the variance-inflation factor, which is based on R^2, is used as a
measure for multicollinearity. But, in contrast to OLS regression there is
no robust R^2 available for MM-regressions in R. Do you know if an
equivalent or an alternative nmeasure of multicollinearity is available for
MM-regression in R?


With best regards,
Carsten Colombier

Dr. Carsten Colombier
Economist
Group of Economic Advisers
Swiss Federal Finance Administration
Bundesgasse 3
CH-3003 Bern

phone +41 31 322 63 32
fax +41 31 323 08 33
email: carsten.colombier at efv.admin.ch
www.efv.admin.ch

On Mon, 16 Aug 2004 Carsten.Colombier at efv.admin.ch wrote:
> Dear R users,
> 
> Usually the variance-inflation factor, which is based on R^2, is used as a
> measure for multicollinearity. 
I disagree, strongly, that this is `usual' practice.
> But, in contrast to OLS regression there is
> no robust R^2 available for MM-regressions in R. Do you know if an
> equivalent or an alternative nmeasure of multicollinearity is available for
> MM-regression in R?
?kappa .

-- 
Brian D. Ripley,                  ripley at stats.ox.ac.uk
Professor of Applied Statistics,  http://www.stats.ox.ac.uk/~ripley/
University of Oxford,             Tel:  +44 1865 272861 (self)
1 South Parks Road,                     +44 1865 272866 (PA)
Oxford OX1 3TG, UK                Fax:  +44 1865 272595

--- Carsten.Colombier at efv.admin.ch wrote:
> Dear R users,
> 
> Usually the variance-inflation factor, which is based on R^2, is
> used as a
> measure for multicollinearity. But, in contrast to OLS regression
> there is
> no robust R^2 available for MM-regressions in R. Do you know if an
> equivalent or an alternative nmeasure of multicollinearity is
> available for
> MM-regression in R?
> 
I'm not sure what MM-regression is. But I've just put a general
purpose tool for evaluating collinearity on my website. See
http://www.xs4all.nl/~jhckx/R/perturb/

The perturb programs works by adding small random changes
(perturbations) to selected variables. Categorical variables are
randomly misclassified. This process is repeated a specified number
of times, after which the impact of the perturbations on parameter
stability can be evaluated. It should work with any R-procedure that
has a formula.

The package also contains colldiag, for calculating condition indexes
and variance decomposition proportions. Since this only works on the
independent variables, it should work for your problem as well.

Feedback welcomed. I plan to submit the package to CRAN in a few
days, after I get the help files updated