st: AFT Models and time origin

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

Megan Andrew wrote:

> *From:* Megan Andrew <mailto:mandrew@ssc.wisc.edu>
> *To:* statalist@hsphsun2.harvard.edu 
> <mailto:statalist@hsphsun2.harvard.edu>
> *Sent:* Tuesday, August 08, 2006 3:29 PM
> *Subject:* AFT Models and time origin
>
> I'm estimating Accelerated Failure Time (AFT) models with a 
> log-logistic distribution to analyze the time to various life course 
> events in the presence of a particular treatment.  Cleves et al. 
> (2004) note that the coefficients in an AFT model are sensitive to 
> when the origin of time is set.  Not surprisingly, setting analysis 
> time to begin at different dates changes the results from my own 
> models.  Though it is clear that setting analysis time to begin at 
> different dates changes model results, it is not exactly clear to me 
> why.  Does anyone have thoughts on the matter or can anyone point me 
> to technical journal articles discussing this characteristic of the 
> AFT model?  I'd like to make a more informed decision as to when I 
> should set the origin of analysis time in my models.
>  
> Thanks,
> Megan Andrew
> University of Wisconsin - Madison
*
*   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/