R help - May 2003 - stepAIC

Hi,

I tried to use stepAIC for a cox proportional model
(30 independent variables) built by coxph. I found
that no matter whether I used direction="forward" or
direction="backward" as an option in stepAIC, I
alweays got exactly the same output, i.e. it seems
only backward selection was performed (eliminate 1
variable at a time). why did it happen like this? or
for cox model, stepAIC can only perform backward
selection?

thanks

zander

stepAIC is support software for MASS4.  There is an example in the book of 
it doing what you claim it cannot do.  So do look it up and see what your 
mistake was.

On Tue, 20 May 2003, array chip wrote:
> I tried to use stepAIC for a cox proportional model
> (30 independent variables) built by coxph. I found
> that no matter whether I used direction="forward" or
> direction="backward" as an option in stepAIC, I
> alweays got exactly the same output, i.e. it seems
> only backward selection was performed (eliminate 1
> variable at a time). why did it happen like this? or
> for cox model, stepAIC can only perform backward
> selection?
-- 
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