R help - Jul 2011 - coxme for random effects only model

Dear all,

 

I have encountered the following problem where coxme seems to allow
model with only random effect in R 2.11.1 but not in R 2.13.0. Following
is the error message using rat example data. Any comment on this is
appreciated. 

 

In R2.13
> library(coxme)
> rat1 <- coxme(Surv(time, status) ~ rx + (1|litter), rats)
> rat0 <- coxme(Surv(time, status) ~ (1|litter), rats)
Error in coxme.fit(X, Y, strats, offset, init, control, weights weights,  : 

  No starting estimate was successful

 

 

In R 2.11.1
> library(coxme)
> rat1 <- coxme(Surv(time, status) ~ rx + (1|litter), rats)
> rat0 <- coxme(Surv(time, status) ~ (1|litter), rats)
> rat0
Cox mixed-effects model fit by maximum likelihood

  Data: rats

  events, n = 40, 150

  Iterations= 22 113 

                    NULL Integrated Penalized

Log-likelihood -185.6556  -184.8205 -177.8707

 

                  Chisq    df       p   AIC    BIC

Integrated loglik  1.67  1.00 0.19623 -0.33  -2.02

Penalized loglik 15.57 12.03 0.21372 -8.50 -28.82

 

Model:  Surv(time, status) ~ (1 | litter) 

 

Random effects

Group  Variable  Std Dev   Variance 

 litter Intercept 0.6441629 0.4149458

 

Ruby Chang


	[[alternative HTML version deleted]]

---begin inclusion --
I have encountered the following problem where coxme seems to allow
model with only random effect in R 2.11.1 but not in R 2.13.0. Following
is the error message using rat example data. Any comment on this is
appreciated. 
--- end inclusion

What version of coxme is being used?  (Use the sessionInfo() command per
the posting guide).
  The coxme call you wrote works for me, but then I'm not running the
same version of coxme as anyone else ;-)  
  The code for initial values of the iteration has recently been
updated, driven by an example data set that was very fussy.  I had to
implement a multiple-value grid search for starting estimates.

Terry Therneau
author of coxme