R help - Nov 2004 - frailty and time-dependent covariate

Hello,

I'm trying to estimate a cox model with a frailty variable and
time-dependent covariate (below there is the statement I use and the error
message). It's seems to be impossible, because every time I add the
time-dependent covariate the model doesn't converge. Instead, if I estimate
the same model without the time-dependent covariate it's converge. I'd
like knowing if it's a statistical problem due to the model formula or if it
could be a problem related to my data.

Thanks a lot

Emanuela Rossi

> fit_19_1<-coxph(Surv(DATA_INI1,DATA_FIN1,EVENT1)~
V1+V2+alt1+alt2+strata(autocorr1)+cap1+SP+SP2+SP3+SP3:log(DATA_FIN1)+D1500+D3000+D4500+frailty.gaussian(ID),data=SURV1)
Warning messages: 
1: Inner loop failed to coverge for iterations 1 3 in: coxpenal.fit(X, Y,
strats, offset, init = init, control, weights = weights,
2: longer object length
        is not a multiple of shorter object length in: offset + coxfit$fcoef[x[,
fcol]]
3: X matrix deemed to be singular; variable 8 in: coxph(Surv(DATA_INI1,
DATA_FIN1, EVENT1) ~ V1 + V2 + alt1 + alt2 +

	[[alternative HTML version deleted]]

On Wed, 17 Nov 2004, Emanuela Rossi wrote:
> Hello,
>
> I'm trying to estimate a cox model with a frailty variable and 
> time-dependent covariate (below there is the statement I use and the 
> error message). It's seems to be impossible, because every time I add 
> the time-dependent covariate the model doesn't converge. Instead, if I 
> estimate the same model without the time-dependent covariate it's 
> converge. I'd like knowing if it's a statistical problem due to the
> model formula or if it could be a problem related to my data.
The message you quote probably does not indicate a convergence problem, 
since it is only in "iterations 1 3".

Howver, there does seem to be a bug with interactions of any sort in 
frailty models, something connected with reordering terms.  You could try 
defining a variable by hand for your interaction term and fitting the 
model that way.

Incidentally, I am not nearly as enthusiastic about the frailty models as 
many users seem to be.  For most multivariate survival problems I 
wouldn't want to fit a frailty model, and I'm not sure that I would
trust
the penalised likelihood approximation if I did.   The exception would be 
situations where I was actually interested in the variance components, as 
in Dr Therneau's new "kinship" package for linkage analysis with
survival
data.


 	-thomas

>
> Thanks a lot
>
> Emanuela Rossi
>
>
>> fit_19_1<-coxph(Surv(DATA_INI1,DATA_FIN1,EVENT1)~ 
>>
V1+V2+alt1+alt2+strata(autocorr1)+cap1+SP+SP2+SP3+SP3:log(DATA_FIN1)+D1500+D3000+D4500+frailty.gaussian(ID),data=SURV1)
>
> Warning messages:
> 1: Inner loop failed to coverge for iterations 1 3 in: coxpenal.fit(X, Y,
strats, offset, init = init, control, weights = weights,
> 2: longer object length
>        is not a multiple of shorter object length in: offset +
coxfit$fcoef[x[, fcol]]
> 3: X matrix deemed to be singular; variable 8 in: coxph(Surv(DATA_INI1,
DATA_FIN1, EVENT1) ~ V1 + V2 + alt1 + alt2 +
>
> 	[[alternative HTML version deleted]]
>
> ______________________________________________
> R-help at stat.math.ethz.ch mailing list
> https://stat.ethz.ch/mailman/listinfo/r-help
> PLEASE do read the posting guide!
http://www.R-project.org/posting-guide.html
>
Thomas Lumley			Assoc. Professor, Biostatistics
tlumley at u.washington.edu	University of Washington, Seattle