8:09 a.m.
Le sam. 30 janv. 2021 à 10:51, Riccardo (Jack) Lucchetti
<p002264(a)staff.univpm.it> a écrit :
On Fri, 29 Jan 2021, Artur Bala wrote:
> Hi,
> Probably the topic of "How small is really small?" has been already
> discussed but I'm still a bit confused.
The fact
that many, in the economic profession, have come to the unfortunate habit
of automatically thinking "no stars -> bad, two stars -> good, three stars
-> wow" should not deter us, as authors of a statistical package, from
reporting the statistic in the most precise way possible and refrain from
patronising the user.
Btw, in an OLS estimation (at the end of an iterative WLS process) I ended up with a
somewhat extreme p-value though ‘non-zero’ printout :)
F(4, 45) = 2.71006e+08 with p-value = 1.90538e-165
Best,
Artur
4:53 p.m.
New subject: on near-zero std errors values
9Just some comments on the mathematics underlying this problem. The OLS
X'X (or its equivalent) is symmetric and positive definite. Therefore its
inverse is also symmetric and positive definite. If the variance of one of
your estimated coefficients is zero the inverse can not be positive
definite. Even if that variance is very close to zero you probably will
encounter numerical instability. In either case, I would think that you
should consider the theory underlying the equation that you are
estimating. the F-value in Artur's last email indicates either a
considerable degree of multicollinearity between the dependent variable and
one or more of the independent variables and is thus valuable information
or some other problem in the implementation of the iterative process.
John C Frain
3 Aranleigh Park
Rathfarnham
Dublin 14
Ireland
www.tcd.ie/Economics/staff/frainj/home.html
mailto:frainj@tcd.ie
mailto:frainj@gmail.com
On Mon, 1 Feb 2021 at 08:09, Artur Bala <artur.bala.tn(a)gmail.com> wrote:
Le sam. 30 janv. 2021 à 10:51, Riccardo (Jack) Lucchetti <
p002264(a)staff.univpm.it> a écrit :
> On Fri, 29 Jan 2021, Artur Bala wrote:
> > Hi,
> > Probably the topic of "How small is really small?" has been already
> > discussed but I'm still a bit confused.
>
> The fact
> that many, in the economic profession, have come to the unfortunate habit
> of automatically thinking "no stars -> bad, two stars -> good, three
> stars
> -> wow" should not deter us, as authors of a statistical package, from
> reporting the statistic in the most precise way possible and refrain from
> patronising the user.
>
Btw, in an OLS estimation (at the end of an iterative WLS process) I ended
up with a somewhat extreme p-value though ‘non-zero’ printout :)
F(4, 45) = 2.71006e+08 with p-value = 1.90538e-165
Best,
Artur
_______________________________________________
Gretl-users mailing list -- gretl-users(a)gretlml.univpm.it
To unsubscribe send an email to gretl-users-leave(a)gretlml.univpm.it
Website:
https://gretlml.univpm.it/postorius/lists/gretl-users.gretlml.univpm.it/
5:53 p.m.
New subject: on near-zero std errors values
Artur
Are many or your wights tending to zero in your iterative WLS?
John C Frain
3 Aranleigh Park
Rathfarnham
Dublin 14
Ireland
www.tcd.ie/Economics/staff/frainj/home.html
mailto:frainj@tcd.ie
mailto:frainj@gmail.com
On Mon, 1 Feb 2021 at 08:09, Artur Bala <artur.bala.tn(a)gmail.com> wrote:
Le sam. 30 janv. 2021 à 10:51, Riccardo (Jack) Lucchetti <
p002264(a)staff.univpm.it> a écrit :
> On Fri, 29 Jan 2021, Artur Bala wrote:
> > Hi,
> > Probably the topic of "How small is really small?" has been already
> > discussed but I'm still a bit confused.
>
> The fact
> that many, in the economic profession, have come to the unfortunate habit
> of automatically thinking "no stars -> bad, two stars -> good, three
> stars
> -> wow" should not deter us, as authors of a statistical package, from
> reporting the statistic in the most precise way possible and refrain from
> patronising the user.
>
Btw, in an OLS estimation (at the end of an iterative WLS process) I ended
up with a somewhat extreme p-value though ‘non-zero’ printout :)
F(4, 45) = 2.71006e+08 with p-value = 1.90538e-165
Best,
Artur
_______________________________________________
Gretl-users mailing list -- gretl-users(a)gretlml.univpm.it
To unsubscribe send an email to gretl-users-leave(a)gretlml.univpm.it
Website:
https://gretlml.univpm.it/postorius/lists/gretl-users.gretlml.univpm.it/
6:27 p.m.
New subject: on near-zero std errors values
Le lun. 1 févr. 2021 à 18:54, John C Frain <frainj(a)gmail.com> a écrit :
Artur
Are many or your wights tending to zero in your iterative WLS?
Dear John,
Thank you for your previous comments.
And yes, your suggestion is correct: there's only 6 out of 50 observations
with a non-zero weight. Actually, 5 of them with weight=1 and 1 with a
weight of 8.4e-5.
(Note that the final weights resulting from the iterative WLS are
indireclty used to calculate a pseudo (yvar) which goes through a final
OLS estimation. The F statisrics I reported relates to the OLS model).
Best,
Artur
1508
days inactive
1508
days old
3 comments
2 participants
participants (2)
-
Artur Bala -
John C Frain