Re: st: Average age at first occurrence analysis

[Steven Samuels <sjhsamuels@earthlink.net>]

Andrew,


Andrew:

I'm not familiar with the R package models, and I don't know what  
they do. In Stata, the Weibull model is available in -streg-. The  
Cox  model can be considered a generalization: both are proportional  
hazard models; in the Weibull, there is a parametric form for the  
hazard function; in the Cox model there is not. As Greenland and  
Boshuizen show, the PH assumption is not necessary to an age-at-onset  
analysis. One can compute the difference between portions of any two  
survival curves. If the Weibull assumption does hold, then one can do  
it parametrically.  Obviously the caveats that Wilk and Lash raise  
apply.

The two articles you referred to are both available online.
http://ete-online.com/content/4/1/1    Wilk and Lash
http://ije.oxfordjournals.org/cgi/reprint/26/4/867   Greenland and  
Boshuizen

Best,

Steve

On Feb 11, 2009, at 7:46 PM, Andrew John Brunskill wrote:

> Please would any kind people suggest some stata commands that would  
> allow me to calculate estimated average times , using Weibull type  
> models, with particular emphasis on their use in case only designs.  
> The only reference in the stata list I can find is on the shared  
> frailty model but most of the sources I can find are on the Cox  
> (proportional hazards) model.  There was a very valuable paper by  
> Boshuizen HC and Greenland S Int J Epidemiology 1997 26 4 867  
> âAverage age at first occurrence as an alternative occurrence  
> parameterâ which âencouraged the development of suitable packaged  
> software allowing the fitting of these modelsâ  and an excellent  
> but scary paper by Wilk JB and Lash TL Emerging Themes in  
> Epidemiology 2007 4:1 âRisk factors studies of age-at-onset..:a  
> cautionary taleâ  which shows the methodological and interpretative  
> problems.  Prof Greenland has been kind enough to clarify for me  
> what I need to understand.  What I would love to get as a  
> reasonably numerate medical epidemiologist ( but not someone who  
> models bus waiting times in greek numerals in my head (alas!) )is a  
> good book or source (lecture notes??)  with some worked commands in  
> hopefully stata.  My work is on data on kidney failure and  
> transplant so it might even be helpful  to some other people  
> sometime!   Thanks very much for any replies. Andrew Brunskill  
> andrew@anzdata.org.au
>
> Andrew J Brunskill,
> University of Washington  jaberwok@u.washington.edu
> Fellow in Epidemiology ANZDATA andrew@anzdata.org.au

*
*   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/