Re: st: poisson vs cox

[Enzo Coviello <enzo.coviello@alice.it>]

Hi Raoul,

using age as time axis the effect of age at diagnosis should not be 
clearly estimated by the Cox model.
So, any effect of this covariate in the derived Poisson model could 
be a residual effect caused by the aggregated nature of data.

Changing time axis and introducing the appropriate regressor for it 
in Poisson model could be a useful check of this hypothesis.

Best
Enzo


> Hi,
>
> I am running a poisson regression, but if I run a cox regression and
> use the same covariates, then some of the rate ratios clearly differ
> between the models. I always thought that a cox regression should give
> approx. the same results as a poisson regression. What could be the
> reason for this?
>
> In more detail:
>
> I first stset with age as time:
> .stset dox, fail(fail) enter(doe) origin(dob) id(id)
>
> then I used stsplit to create age and calender period bands and used
> poisson regression with additional covariates (sex, age at diagnosis,
> treatment etc)..
>
> then, as a check and out of curiousity I ran a cox regression model,
> but the rate ratios for treatment and in particular for age at
> diagnosis differ between the models.
> Thanks for your help.
>
> Raoul
> *
> *   For searches and help try:
> *   http://www.stata.com/support/faqs/res/findit.html
> *   http://www.stata.com/support/statalist/faq
> *   http://www.ats.ucla.edu/stat/stata/
*
*   For searches and help try:
*   http://www.stata.com/support/faqs/res/findit.html
*   http://www.stata.com/support/statalist/faq
*   http://www.ats.ucla.edu/stat/stata/