[R] stop criterion for stepAIC

[Prof Brian Ripley]

You completely misunderstand: have you even *looked* at the help page?

Description:

     Performs stepwise model selection by exact AIC.

!  If you don't know what AIC is (and it seems so), please research it
(e.g. in MASS the book whose support software you are using without any 
credit).

On Mon, 16 Jun 2003, Thomas Fischer wrote:

> I am using the function stepAIC (library MASS) to run a backward 
> elimination on my linear regression. The new model stepAIC calculates 
> contains coefficients that have a Pr(>|t|) value below 0.1, but I'd 
> like to have only coefficients with 0.001 or below.

Nothing to do with AIC, and very dangerous since stepAIC works with terms, 
not coefficients (and a term may have several coefficients).

> How can I change the stop criterion for stepAIC, so that it is more 
> strict? There is a parameter "steps", but it is only a upper bound of 
> steps.

The only thing you can do is increase `k', for a pseudo-AIC.

-- 
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