s-news
[Top] [All Lists]

Re: FW: group covariates in coxph with a frailty term?

To: lawrence-hunsicker@uiowa.edu
Subject: Re: FW: group covariates in coxph with a frailty term?
From: Terry Therneau <therneau@mayo.edu>
Date: Fri, 22 Aug 2008 11:01:33 -0500 (CDT)
Cc: s-news@lists.biostat.wustl.edu
Reply-to: Terry Therneau <therneau@mayo.edu>
Larry,
  Simple frailties correspond only to the simplest random effects, i.e.,
         random= ~1 | group

in lme notation.  No nesting, only 1 group, no covariates, etc.  

  More general effects are available in coxme.  The attached technical report 
will give you a lot more information.  Unfortunately, it does not yet handle 
the 
following model, which is likely what you want (guessing from the phrasing of 
your question).
  
        random =  ~1+ treatment | institution 
        
It will be in the next revision, which is in progress.  But since it is mostly 
a 
nights and weekends labor, progress will remain very slow until winter returns.

        Terry Therneau
        


<Prev in Thread] Current Thread [Next in Thread>