Re: st: Cox Regression

[Marcello Pagano <pagano@hsph.harvard.edu>]

I would not think so.  You are basing your decision of non-significance 
on a test whose
validity is based on an assumption.  If the assumption is not valid, 
then neither is the test.
Think of a simple analogy with linear models.  Suppose the relation 
between y and x is a
V-shaped one.  The coefficient of x in a linear model may well test out 
to zero. That would
not mean that the relationship is linear.

m.p.

Timothy.Mak@iop.kcl.ac.uk wrote:


> Hi when performing a Cox regression with or without a time dependent 
> covariate, is a non-significant result in a -stphtest- generally enough 
> evidence for the assumption of proportional hazard? 
> If I have multiple covariates, but I'm only interested in one, is it 
> enough to check only the proportional hazard assumption of the variable 
> I'm interested in? ie using the stphtest, detail, rather than stphtest? 
> Lastly, is it better to use stphtest log, or one of the other option 
> rather than stphtest if hazard increases over time and most of my 
> participants 'died' early during the observation period? 
> Thanks a lot for any answer. 
> Tim 
> *
> *   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/