discussion between users of the gretl econometric software

Riccardo (Jack) Lucchetti | 1 Sep 2007 12:08

Re: System estimation

On Mon, 27 Aug 2007, Riccardo (Jack) Lucchetti wrote:

>> On Fri, 24 Aug 2007, Franck Nadaud wrote:
>> 
>>> I would like to know if it is possible in system estimation to
>>> recover all the coefficients and standards errors, because i
>>> have to compute the AIDS elasticities.

Ok, folks, attached you will find a script and a dataset to replicate the 
AIDS example for SAS that you can find at

http://support.sas.com/rnd/app/examples/ets/aids/index.htm

and

http://support.sas.com/rnd/app/examples/ets/elasticity/index.htm

Please note:

1) you need CVS gretl for this. 1.6.5 will not do. I haven't tested the 
windows snapshot.

2) A few things are hardwired to replicate the above example. With a bit 
of work, though, this script could be turned into a proper function for 
general AIDS systems.

3) I used a different formula for price elasticities than the one in the 
SAS example. Differences are minor.

4) In principle, one may want to compute a standard error for the

(Continue reading)

headers

Franck Nadaud | 3 Sep 2007 16:54

Re: System estimation

Hello all, greetings from Paris !

First, let me hotly thank you for the work around system estimation. Please
apologize for the late response, having been mobilized on many other things.

Effectively Riccardo, I will have to adapt my scripts to compute the
elasticities by the delta method, thanx again. I used to test different
formulas adapted mainly from Green & Alston (1994) paper, and also a WP from
Barnett and Seck:

http://mpra.ub.uni-muenchen.de/417/

In this paper matrix formulas are given, as you said, differences are small,
and it seems that it is mainly the case for agregate data.

I intended to compute elasticities by bootstrap, but dit not have the time to
work along this line.

A this stage I begin a project where we estimate an AIDS on a pseudopanel, so
it is very good news to be able to retreive the estimated coefficient in
system estimation, thanx Allin !

We intend to test in the future different flexible forms, so they may be
included as functions later, Let me know if you are interested in my progress
along.

Cheers all,

sincerely

(Continue reading)

headers

Allin Cottrell | 1 Sep 2007 21:53

Re: System estimation

On Sat, 1 Sep 2007, Riccardo (Jack) Lucchetti wrote:
> 
> Ok, folks, attached you will find a script and a dataset to 
> replicate the AIDS example for SAS that you can find at
> [ URLs given]

Nice work!  

Meanwhile I have fixed up the system estimation code so that the 
accessors $coeff and $vcv produce what I think are the correct 
values (for, respectively, the vectorization of the coefficient 
matrix and the coefficient covariance matrix) for all the system 
estimation methods (from ols to fiml).

Allin.