s-news
[Top] [All Lists]

Summary of answers gamma frailty hazards

To: s-news@lists.biostat.wustl.edu
Subject: Summary of answers gamma frailty hazards
From: "Perperoglou, A. (MStat)" <A.Perperoglou@lumc.nl>
Date: Fri, 9 May 2003 12:14:06 +0200
   I would like to thank all of you who replied. My question was
>Dear S-plus users
>after fitting a gamma frailty model using the coxph command 
>you can ask for the details with coxph.details and then you get 
>an estimate of the hazard increment. 
>Can anyone give me a reference on what exactly this hazard is 
>and how it is estimated?

Terry Therneau replied

>  fit <- coxph(Surv(time, status) ~ age + frailty
>(inst, df=4), data=lung)
>  dt  <- coxph.detail(fit)

>  The coxph.detail function treats the fitted frailty
> coefficients as though
>the were a set of fixed offsets.  So in the above 
>model:

>       risk score for a subject = (age - fit$mean)*fit
>$coef + frailty

>where "frailty" is the appropriate frailty
 >coefficient from the model; the
>first subject is from institution 3 so his value is
 >fit$frail[3]= -.197.

>The hazard increment at a given time is

>       #deaths at that time
        --------------------
>       sum(exp(risk score))

>(for the Breslow method), where the denominator is 
>over those still at risk
>and under observation at that time.

        Terry Therneau

 Also Laurent Drapeau replied that he is out of his office :)

Thanks and best regards

Aris Perperoglou
Department of Medical Statistics
Leiden University Medical Center
P.O.Box 9604,  
2300 RC Leiden, The Netherlands
phone  +31-71-5276829
fax    +31-71-5276799
e-mail a.perperoglou@lumc.nl



<Prev in Thread] Current Thread [Next in Thread>
  • Summary of answers gamma frailty hazards, Perperoglou, A. (MStat) <=