Re: st: Cox model with unobserved heterogeneity

[Maarten Buis <maartenlbuis@gmail.com>]

On Fri, Sep 28, 2012 at 11:12 AM, Nynke  wrote:
> I would like to estimate a mixed proportional hazard model with a flexibel specification of the hazard in Stata, comparable to a Cox model with unobserved heterogeneity. I can estimate a MPH with a parametric form of the hazard and unobserved heterogeneity (streg with frailty) or a Cox model without accounting for unobserved heterogeneity (stcox), but how do I sort of combine these two? And when I have estimated this model, is there a way to check the PH assumption?

I suspect that the model you are looking for is just not identified.
You can estimate a Cox model with group level unobserved
heterogeneity, with the -shared()- option in -stcox-. In that cases it
uses information from how similar observations are within groups. For
individual level unobserved heterogeneity the parametric models use
deviations from theoretical form of the baseline hazard function to
identify the unobserved heterogeneity component. Since Cox regression
has no theoretical form for the baseline hazard function, there is no
information that can be used to estimate the individual level
unobserved heterogeneity component.

Hope this helps,
Maarten

---------------------------------
Maarten L. Buis
WZB
Reichpietschufer 50
10785 Berlin
Germany

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

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