[R] glm and lrm disagree with zero table cells

[Thomas Lumley]

On Thu, 24 Oct 2002, Eric Rescorla wrote:

> I've noticed that glm and lrm give extremely different results if you
> attempt to fit a saturated model to a dataset with zero cells. Consider,
> for instance the data from, Agresti's Death Penalty example [0].

The MLEs are infinite, so there are bound to be problems.  Both methods
will stop short of infinity, when the coefficients or likelihood stop
changing, but where they stop will depend on the exact tolerance and
stopping rule.

They aren't actually that different in terms of fitted values.  For
example, the zero cell has a fitted probability of 0.00025 and 0.0005 in
the two models.


	-thomas



-.-.-.-.-.-.-.-.-.-.-.-.-.-.-.-.-.-.-.-.-.-.-.-.-.-.-.-.-.-.-.-.-.-.-.-.-.-.-.-
r-help mailing list -- Read http://www.ci.tuwien.ac.at/~hornik/R/R-FAQ.html
Send "info", "help", or "[un]subscribe"
(in the "body", not the subject !)  To: r-help-request at stat.math.ethz.ch
_._._._._._._._._._._._._._._._._._._._._._._._._._._._._._._._._._._._._._._._