[Gretl-devel] RE probit and perfect prediction
Hi all,
a user just sent me a report on a problem with the RE probit estimator:
if one tries a specification that would lead to perfect prediction in the
pooled specification, the command just aborts instead of dropping the
offending regressor from the list, as it probably should.
The following script exemplifies the problem and suggests a possible
solution:
<hansl>
nulldata 24
set seed 123
setobs 3 1:1 --stacked-time-series
series x = normal()
series d = uniform() > 0.8
series y = (1 + x + d + normal()) > 0
list X = const x d
# this won't do
# probit y X --random
# this will
probit y X --quiet
list X = $xlist
probit y X --random
</hansl>
It's very unlikely that I'll have time to work on this over the weekend,
but will try to do so on Monday. If you have alternative ideas on how to
handle this, suggestions are welcome!
-------------------------------------------------------
Riccardo (Jack) Lucchetti
Dipartimento di Scienze Economiche e Sociali (DiSES)
Università Politecnica delle Marche
(formerly known as Università di Ancona)
r.lucchetti(a)univpm.it
http://www2.econ.univpm.it/servizi/hpp/lucchetti
-------------------------------------------------------