It's a bug in summary.aareg which no one found until now.
What's wrong:
If dfbeta=TRUE then there is a second estimate of variance calculated, labeled
as
test.var2. If maxtime is set, then both estimates of variance need to be
recalculated by
the summary routine. An incorrect if-then-else flow led it to look for
test.var2 when it
wasn't relevant. My test cases with maxtime also happened to have
dfbeta=TRUE.
Short term solution: set dfbeta=TRUE. A bit more computation time though.
Long term: I'll fix it, and a new version of survival will one day appear.
(With 200+
packages that depend on survival, new releases now require a lot of checking.
No
overnight fixes).
Terry T
On 10/05/2012 05:00 AM, r-help-request at r-project.org
wrote:> Hi all,
>
> I've ventured into the world of nonparametric survival and I would like
to use the "maxtime" option for printing and plotting my aareg fit.
>
> However, my fit does not have "test.var2" and this stops the
print and plot when adding a maxtime.
>
> My code is as follows:
>
> Response<-Surv(Time,Event)
> Model<-aareg(Response~Factor1*Factor2)
> Model2<-aareg(Response~Factor1+Factor2) #Just did this to see if the
interaction term had anything to do with it
>
> Model, print(Model), summary(Model), and plot(Model) seem to work fine, but
as soon as I try summary/print/plot(Model, maxtime=400) it tells me that
test.var2 is not found and when I look at summary(Model), there is indeed a
NULL under test.var2.
>
> Anyone know why it doesn't include test.var2? Is this a compatibility
problem? I'm using R version 2.13 (I know it's quite old, but updating
is a pain when you don't have admin rights to your computer) and just
updated the survival package (no warning messages).
>
> Any input would be much appreciated.
>
> Cheers,
>
> Freya
