Dear listers-
SPLUS2000R3 + R1.5 WINNT4
This is not yet a summary to my previous post about predicting with a
frailty term (to which Geir Egil Eide has commented) and I preface this
with the caveat that the question is related to survReg/coxph and frailty,
but is not really a mechanics question. Read on if you'd like.
I've asked a variant of this before, but here goes again.
Imagine a data set with 6 hospitals and 50 patients monitored in each
hospital. Each hospital was randomly assigned to a 2 treatment x 3 diet
factorial design such that each of the 100 patients were receiving the same
treatment and diet combination in the hosipital and clearly there is one
hospital for each treat*diet combination. Moreover, there is the chance
(lets assume) for a significant number of ties.
Would an apporpriate/acceptable model be
survReg(Surv(time,cens)~treat*diet+frailty(hospital)?
or
coxph(Surv(time,cens)~treat*diet+frailty(hospital),method="efron")
In reality, I have a data set with 168 hospital equivalents and from
50-2500 patients/hospital. There are 2x3x2x2x2x4=192 treatment
combinations each with a unique number of patients... the treatments, in
full factorial, thus exceed the number of "hospitals", meaning that some of
the cells in the model are empty (some hospitals, and all of the patients
in that hospital, have been lost during the study). Moreover, there are
alot of ties within each "hospital".
Philisophically, hospital should be a random or frailty term... but as the
number of treatment combinations = or exceeds the number of "hospitals", is
it still appropriate? And the ties....
Any advice would be appreciated (including give up ;)
cheers
andrew
----------------------------------------------------------
Dr. Andrew Beckerman
Institute of Biological Science
University of Stirling
Stirling FK9 4LA
+44 (0)1786 then w-467808 f-464994
(APB is not responsible for anything below this)
--
The University of Stirling is a university established in Scotland by
charter at Stirling, FK9 4LA. Privileged/Confidential Information may
be contained in this message. If you are not the addressee indicated
in this message (or responsible for delivery of the message to such
person), you may not disclose, copy or deliver this message to anyone
and any action taken or omitted to be taken in reliance on it, is
prohibited and may be unlawful. In such case, you should destroy this
message and kindly notify the sender by reply email. Please advise
immediately if you or your employer do not consent to Internet email
for messages of this kind. Opinions, conclusions and other
information in this message that do not relate to the official
business of the University of Stirling shall be understood as neither
given nor endorsed by it.
|