different linear predictors from coxph and coxme

[Emmanuel Tillard <tillard@cirad.fr>]

Hello S-PLUS user's
I am using S-PLUS 7.0 with Windows XP (Professional developer)
I am trying to compare 2 survival models using the functions coxph (with 
a gaussian frailty term) and coxme
The results obtained from the two models are quite similar (coefficients 
and frailty)
However, i am surprised to see that the linear predictors are different

#an example
fit.coxme <- coxme(Surv(time, status) ~ rx, data=rats, random= ~1|litter)
fit.coxph <- coxph(Surv(time, status) ~ rx + frailty(litter, 
dist="gaussian"), data=rats)
fit.coxme$coef$fixed
fit.coxph$coef

fit.coxme$frail
fit.coxph$frail

fit.coxme$linear.predictor
fit.coxph$linear.predictors

Could somebody explain me why there is a difference in linear predictors 
and how to switch from one type to the other ?
Any help would be very appreciated
Regards

--
Emmanuel Tillard
Veterinaire
CIRAD-EMVT
Unite de recherche 18

UMR868 Elevage des Ruminants en Regions Chaudes (ERRC)
Campus ENSA-INRA
2 place Viala
34060 Montpellier cedex 1

tel:    0499612265 (fixe)
        0633850598 (gsm)
fax:    0467545694
e-mail: tillard@cirad.fr