survReg, frailties, and predictions

[Brian Beckage <bb2@duke.edu>]

Hi,

I'm trying to predict survival times using a fitted survReg object that 
includes a gaussian frailty, e.g.
fit.final3<-survReg(Herb.surv~Species*Understory+Protection+Overstory*Understory+
        frailty(Plot,distribution="gaussian",sparse=F),data = Herb.factor, 
dist = "loglogistic")
I can successfully use predict() with the existing data, e.g.

Y<-predict(object=fit.final3)

> Y
   [1] 1519.68175 2159.55649  635.31659
   [4]   57.54016  902.94895  696.96931 ... etc.

but if I use new data, e.g.

> Herb.factor.pred[1:5,]
   Species Understory Overstory  Protection  Plot
 1    Acru   NonRhodo       Gap              U     Ln1
 2    Acru      Rhodo          Gap              U     Ln1
 3    Acru   NonRhodo       Gap              P     Ln1
 4    Acru      Rhodo          Gap              P     Ln1
 5    Acru   NonRhodo      Canopy          U     Ln1


I encounter the following error

> Y <- predict(object = fit.final3, newdata = Herb.factor.pred)

Error in model.matrix.default(T..: No data for variable "frailty(Plot, 
distribution = "gaussian", sparse = F)"
What am I doing incorrectly?

Secondly, is there a way make predictions using the parameters estimated 
with the frailty term but without including the frailty term in the 
predictions (e.g. with frailties equal to 0)?
I'm using Splus 2000 r3.

Thank you,
Brian Beckage