[R] Conditional gap time frailty cox model for recurrent events

[Therneau, Terry M., Ph.D.]

You can ignore the message below.  The maximizing routine buried within the frailty() 
command buried with coxph() has a maximizer that is not the brightest.  It sometimes gets 
lost but then finds its way again.  The message is from one of those.  It likely took a 
not-so-good update step, and took a couple of iterations to recover.

In coxpenal.fit(X, Y, strats, offset, init = init, control, weights = weights,  :
    Inner loop failed to coverge for iterations 3 4

To be fair the maximizing problem for a mixed effects Cox model is not easy.  In the coxme 
code I have spent much more time on the details of this.

Terry Therneau

------------------------------

On 09/06/2016 05:00 AM, r-help-request at r-project.org wrote:
> Dear Elisabetta,
>
> I have no direct answer to your question, but a suggestion: Use the
> 'coxme' function (in the package with the same name). In the help page
> for 'frailty' (survival) you will find: "The coxme package has
> superseded this method. It is faster, more stable, and more flexible."
>
> Hth, G?ran
>
> On 2016-09-05 11:42, Elisabetta Petracci wrote:
>> >Dear users,
>> >
>> >I am fitting a conditional gap time frailty cox model weighting
>> >observations by means of inverse probability time dependent weigths.
>> >Attached find the self-explaining dataset.
>> >
>> >I have used the following sintax:
>> >
>> >coxph(Surv(gaptstart,gaptstop,status)~treat+strata(nrecord01)+frailty(id,distribution="gamma",method="em"),
>> >data=dataNOTDrr,weights=dataNOTDrr$weight)
>> >
>> >
>> >And I get the following warning:
>> >
>> >Warning message:
>> >In coxpenal.fit(X, Y, strats, offset, init = init, control, weights =
>> >weights,  :
>> >   Inner loop failed to coverge for iterations 3 4
>> >
>> >
>> >I have tried to:
>> >- leave out the weights but I get the error anyway
>> >- to randomly select a subset of patients and I don't get the error. This
>> >seems to suggest that the problem is with some observations.
>> >
>> >Any suggestion?
>> >
>> >Many thanks,
>> >
>> >Elisabetta
>> >