[R] difference between coxph and cph

Thomas Lumley tlumley at u.washington.edu
Wed Apr 21 16:42:11 CEST 2004

On Wed, 21 Apr 2004, [iso-8859-1] Göran Broström wrote:
> More seriously, the difference may well be of numerical character,
> different convergence criteria, "unbalanced" data, etc. It is really
> impossible to say without knowing what your data are (and without looking
> into the code of coxph and cph).

The Cox partial loglikelihood is concave, so it really doesn't take much
numerical care to get the right maximum (unlike some of the parameteric
survival models) except on really ugly data sets.

This one requires reading either the output or the documentation, rather
than the code.  Reading the documentation is especially important for the
Design package -- the whole point of the package is to provide a different
interface to regression modelling.


More information about the R-help mailing list