[R] coxph error

[Thomas Lumley]

On Tue, 10 Feb 2004, Fredrick Schumacher wrote:

> R list:
>
>
>
> I am using a 'for' loop to run a number of different models (stratified
> by different variables) with coxph.   The data becomes sparse when some
> strata are used causing the model to become unstable.  The following
> error occurs and the analysis is terminated.
>
>
>
> >Error in fitter(X, Y, strats, offset, init, control, weights = weights,
> :
>
>       (converted from warning) Loglik converged before variable  1 ;
> beta may be infinite.
>
>
>
> Is there any way to have R skip the models without convergence and apply
> 'NA' to the results, then go on to the other models?

This is usually a warning, not an error, and for good reason since it does
not reliably indicate lack of convergence (poor sensitivity and poor
specificity). You have deliberately turned it into an error with the
setting of options(warn).

If you really want it to be an error but not fatal then use try() to
capture it.


	-thomas