[Gretl-devel] Re: tobit's --quiet flag to quite guiet
On Thu, 19 Jan 2023, Marcin Błażejowski wrote:
On 19.01.2023 10:48, Riccardo (Jack) Lucchetti wrote:
> On Thu, 19 Jan 2023, Marcin Błażejowski wrote:
>
>> Since the message in fact comes from 'gretl_bfgs.c', there could be a set
>> option for optimizers for suppressing any warnings/messages.
>
> Does the "set warnings" command help?
Nope.
<hansl>
#set verbose off
set messages off
open mroz.gdt --quiet
list X = educ exper age kidsl6
tobit hours const X --quiet
</hansl>
Not related to the original point, but I thought it was worth sharing
anyway: after some summary investigation, it looks as if the warning
you're getting is definitely a false positive due to poor scaling. If you
rescale the dependent variable appropriately (say, divide it by 100), the
warning message disappears. The threshold we use for printing out the
warning message is definitely too strict.
Perhaps we should look into some auto-scaling strategy for truncated
models (internally, the "tobit" command is just a special case of
"intreg").
-------------------------------------------------------
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
-------------------------------------------------------