On Fri, 27 Jun 2008, Gordon Hughes wrote:
Another query about mle. As before, I have been experimenting with
stochastic frontier models and I have found that mle behaves entirely
differently with two formulations that seem to be equivalent.
Both report identical log-likelihoods & gradients at iteration 1,
version 2 blows up at iteration 2 while version 1 continues properly.
Technically, the reason for the blow-up in version 2 is that it uses a
step length of 1, whereas version 1 uses a step length of 1e-7. I cannot
understand the reason for the different behaviour but in this case it is
Now THIS is weird.
In this period Allin is away and I won't be able to work on the
source for at least a few days.
It would be extremely helpful if you could collect as much evidence as you
can on cases like this and post it to the list (or privately to us if you
prefer) together with the appropriate datasets if necessary.
Thanks a lot for your work!
Riccardo (Jack) Lucchetti
Dipartimento di Economia
Università Politecnica delle Marche