Brian B asked about the 'strata' argument in survreg():
Statsci changed the name of the routine to survReg, you may be invoking the
older routine.
We had a debate about this. In making the survreg() changes from my
survival4 to survival5 release, the new routine is not upwardly compatable
with the old. Some arguments changed, and the output structure most definitely
did. Should we:
a. Keep the old name. In 2 years no one will remember that you used to
have to say "dist='normal', link='log'" instead of the new "dist='lognormal'".
b. Create a new name survReg and class, so that old code works without
change, but current users will confuse the routines.
Neither is perfect. They voted for (b). (I mildly preferred a.)
----
2. Are there any functions for fitting a Cox model to interval censored data?
No. Perhaps someday.
3. Drawing a survival curve for interval censored data?
Yes, if one assumes a parametric model.
> dfit <- survreg(Surv(time1, time2, status, type='interval') ~1,
data=donnell)
> xx <- predict(dfit, newdata=donnell[1,], type='quantile',
p= (1:29)/30 )
> plot(xx, 1- (1:29)/30, xlab='Time', ylab='Survival')
(The donnell data can be found in the test suite of survival5. It
was sent to me by a user (Donnell) as an example where the
survreg() doesn't converge. survReg and censorReg do fine)
For an interval-censored KM: no.
Terry Therneau
-----------------------------------------------------------------------
This message was distributed by s-news@wubios.wustl.edu. To unsubscribe
send e-mail to s-news-request@wubios.wustl.edu with the BODY of the
message: unsubscribe s-news
|