[Date Prev][Date Next][Thread Prev][Thread Next][Date index][Thread index]

Re: st: RE: query about cox model with shared frailty

From   Steven Samuels <[email protected]>
To   [email protected]
Subject   Re: st: RE: query about cox model with shared frailty
Date   Tue, 2 Sep 2008 12:44:46 -0400


If Murali specifies treatment_new as a stratum variable in the Cox model, he will get probably different survival curves (or hazard curves) in each stratum, with possibly different within-stratum variation.

On Sep 2, 2008, at 12:34 PM, Feiveson, Alan H. (JSC-SK311) wrote:

Murali -

The shared frailty model is meant to account for random fluctuations of
the hazard over many groups where the groups are a random sample from
some population. Then multiple obervations pertaining to the same
"group" share this value of the frailty. A typical example is where
"groups" are subjects and each subject is observed multiple times. Then
there is a different (random) value of the frailty for each subject, but
all the observations pertaining to a given subject share the same
frailty. The estimation results then only provide the variance of this
frailty variable over the population of "groups"; i.e. subjects.
Individual values of the frailty are random variables and are not
obtained in the estimation process. In your case, you have only 2
treatments and they are definitely not a random sample from a population
of treatments. Thus treatment_new should be an explanatory variable in
your model, not a frailty variable.

Al Feiveson

-----Original Message-----
From: [email protected]
[mailto:[email protected]] On Behalf Of Murali
Sent: Tuesday, September 02, 2008 10:55 AM
To: [email protected]
Subject: st: query about cox model with shared frailty

I a trying to fit a cox regression model for a dataset consisting
of 2 groups-trainees and non-trainees. Because I suspect within-group
correlation in failure times, the standard cox model is likely to be
misspecified, which I would like to correct for by using the -shared ()-
option,as in below:

stcox ability_score male edu_yrs,shared(treatment_new)

Now my question is the following: since the shared frailty model is
supposed to account for within-group correlation, shouldn't it report
different estimated frailty variances for each group(trainees and
non-trainees)as well as a corresponding set of parameter estimates for
each group, as opposed to what it actually reports- a single set of
parameter estimates and one estimated frailty variance?

Murali Kuchibhotla

* For searches and help try:

* For searches and help try:
*   For searches and help try:

© Copyright 1996–2024 StataCorp LLC   |   Terms of use   |   Privacy   |   Contact us   |   What's new   |   Site index