Re: st: stratified Cox proportional hazards model and AIC

[Maarten Buis <maartenlbuis@gmail.com>]

On Tue, Nov 1, 2011 at 5:23 AM, Brigham Whitman wrote:
> I am using the stcox command to perform a Cox proportional hazards
> model, and using the log likelihood in the output, I am computing AICc
> values to determine the best model.  If anybody has ever done this, I
> am trying to find out if I can compare the AICc value for a model that
> is stratified (in this case by sex) with the AICc value of a different
> model that does not stratify by sex.  Can I do this?  Or do I need to
> compare stratified models together as one suite of models, and
> non-stratified models as a different suite of models?

I don't think you can. The problem would be determining the difference
in the number of parameters between the two models. The stratified
model uses in your case two non-parametric estimates of the baseline
hazard (one for women and one for men) and the non-stratified model
constraints the baseline hazards to have the same general shape and
allows them to differ by one fixed proportion.

Hope this helps,
Maarten

--------------------------
Maarten L. Buis
Institut fuer Soziologie
Universitaet Tuebingen
Wilhelmstrasse 36
72074 Tuebingen
Germany


http://www.maartenbuis.nl
--------------------------

*
*   For searches and help try:
*   http://www.stata.com/help.cgi?search
*   http://www.stata.com/support/statalist/faq
*   http://www.ats.ucla.edu/stat/stata/