On Tue, 26 Feb 2019, John Paravantis wrote:
Any chances the development team adds zero-inflated Poisson and
negative
binomial regression models (for count data) to gretl?
Negative binomial models are there already, via the "negbin" command (see
the "camtriv.inp" example file); the corresponding GUI element is the same
as for Poisson models, you only need to change the distribution type.
I have seen the zero-inflated Poisson example implemented with code
in the
manual, but was hoping these two count data models could be added natively.
Heh. Time is the scarce resource here. For the moment, you can just
copy-n-paste the example scripts you find at Section 23.7 of the Guide,
but if you have a couple of hours to spare turning those into a function
package shouldn't be difficult ;)
-------------------------------------------------------
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
-------------------------------------------------------