[Gretl-users] Re: help
Dear Sir,
Thank you very much, for quick and exact answer.
Yours sincerely
L. Pfajfar
-----Original Message-----
From: Riccardo (Jack) Lucchetti [mailto:r.lucchetti@univpm.it]
Sent: Wednesday, November 27, 2019 2:31 PM
To: Gretl list
Subject: [Gretl-users] Re: help
On Wed, 27 Nov 2019, Lovrenc Pfajfar wrote:
My problem is: how exactly is calculated
> the chi-squre in Gretl output of tobit model. (is it a likelihood
> ratio test and what exact formula is behind it).
It's not an LR test, it's a Wald test for equality to 0 of all
coefficients except for the constant term. For example:
<hansl>
nulldata 100
set seed 76767
x1 = normal()
x2 = normal()
ystar = 1 + x1 + x2 + normal()
y = ystar > 0 ? ystar : 0
tobit y const x1 x2
beta = $coeff[2:]
V = $vcv[2:,2:]
scalar W = qform(beta', invpd(V))
printf "Wald test = %g (p-value = %g)\n", W, pvalue(x, rows(beta), W)
</hansl>
-------------------------------------------------------
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
-------------------------------------------------------