[R] Correlated frailty model

[Terry Therneau]

>  I am trying to a estimate a correlated frailty model. My dataset is 
> made up of 40000 observations. I would like to know if it is too big or 
> I have done some mistakes in the following code.

  I have used larger data sets successfully.  I need to know a little bit
more about our problem.

> coxme(Surv(yearspan)~ south+mod, data=usa,random= ~ 1|group)

  What is the dimensionality of the covariates, e.g., are they factors with 
hundreds of levels?  How many unique levels are in the group variable?
The error message is surprising, as it appears to be coming from  conversion of
a sparse matrix to dense form.  The vector of frequencies for group might also
be useful.

	Terry Therneau