function problems
by Sven Schreiber

Hi,
I'm currently trying to use the cool function-package features of gretl.
In the process I'm having some problems (with the cvs version from
yesterday).
Illustration:
1. start gretl
2. load example data2-3.gdt
3. execute the following function definition (without the email
linebreaks, of course):
function alphatest2 (int order, int rank, list endo, int consttrend,
bool seasonals)
vecm order rank endo
end function
4. Then execute the following script:
genr myorder = 2
genr myrank = 1
list myvar = unemp infl
genr myconsttrend = 3
genr myboolseasonals = 0
alphatest2 (myorder, myrank, myvar, myconsttrend, myboolseasonals)
... which sends gretl in an infinite loop it seems, so it has to be killed.
Experimentation showed that a function with the command:
vecm order 1 endo
works well.
And something totally unrelated: In the command reference index (text
format), when I click on genr, I get matrix.
thanks,
Sven
17 years

Multiple Linear Regression
by Martin Sykora

I would be very gratefull if somebody would be so nice to point
out to me which file in the C source-code is responsible for
multiple linear regression computation (the coefficients) and
the ADF (Augmented Dickey-Fullet Test). I am completely lost
in the source code and accompanying files. I am also only
interested in using the static library part of gretl, is there
some usefull documentation on this (I was somewhat unsuccessfull
in my searches)
regards, Martin
ps.: I am new to gretl, but next to jmulti this looks really cool.
17 years

Multivariate Distributions
by Rob Tischer

Gretl Users,
I've had occasion to use the MLE routine to run some simple Probit
analyses. I was going to work with some bivariate Probit models, but am
I correct in my belief that Gretl doesn't have a bivariate normal cdf?
FYI. The biggest problems I had were moving between matrices and
variables, and trying to save variables with fewer than the number in
the initial dataset (other than 1x1).
Rob
17 years

Re: Gretl-users Digest, Vol 30, Issue 14
by Thiago Marzagão

Thank you for your reply.
The point is that I'd like to constrain the fitted values (and not the
observed ones) of the dependent variable.
The problem is that my depedent variable cannot "physically" (i.e., in real
world situations) extrapolate a certain range of values (thus all my
observed values lie within this range) but when I run the regression some of
the fitted values are, in fact, outside this range (although the general fit
of the model is good). So I need a kind of Tobit model, but I need to
specify not only the lower bound but also the upper bound of the range. The
Tobit command in GRETL constrains the lower bound to zero (whenever the
fitted values are negative), but it does not specify an upper bound (and it
does not allow me to specify a lower bound other than zero).
Best regards,
Thiago
2006/10/27, gretl-users-request(a)ricardo.ecn.wfu.edu <
gretl-users-request(a)ricardo.ecn.wfu.edu>:
>
> Message: 1
> Date: Thu, 26 Oct 2006 18:02:59 -0400
> From: James Wells <jlwells(a)panix.com>
> Subject: Re: [Gretl-users] censored regression
> To: Gretl list <gretl-users(a)ricardo.ecn.wfu.edu>
> Message-ID: <45413093.1090202(a)panix.com>
> Content-Type: text/plain; charset=ISO-8859-1; format=flowed
>
> Just a thought, but can transform your dependent variable so that the
> minimum is zero? E.g., suppose your observed values of Y are 3, 5, and
> 8 so in the console, do "genr newdepvar = Y - 3"
>
> Thiago Marzagão wrote:
> > I'd like to constrain my dependent variable (which is quantitative and
> > continuous) estimates to lie within a certain range (say, A and Z). Is
> > there any way to do that with GRETL? (Tobit command constrains the
> > lower bound to zero, but I think it does not allow me to specify an
> > upper bound or an alternative lower bound)
> >
> > Thank you,
> >
> > Thiago Marzagão
> > ------------------------------------------------------------------------
17 years, 1 month

Johansen
by Javier García

Then how is it that the nule hypothesis r = 0 and r = 1 are both accepted to
the two contrasts? the only consistent hypothesis with the results is that
both nule hypothesis is that the rank (pi) minor or equal to "r", because
the rank can't be 0 and 1 at the same time.
thank you
Javi
17 years, 1 month

censored regression
by Thiago Marzagão

I'd like to constrain my dependent variable (which is quantitative and
continuous) estimates to lie within a certain range (say, A and Z). Is there
any way to do that with GRETL? (Tobit command constrains the lower bound to
zero, but I think it does not allow me to specify an upper bound or an
alternative lower bound)
Thank you,
Thiago Marzagão
17 years, 1 month

Johansen
by javier garcia enriquez

Hellooo;
At this moment we are doing a cointegration analysis and a doubt has come up about the nule hypothesis that your programme uses in the Johansen test.We don´t know if the maximum eigenvalue uses the same nule hypothesis as the trace contrast( as for instance, Eviews does) , I mean, if both of them contrast the hypothesis that there are, at most, "r" cointegration vectors or, on the contrary, as we have read in some papers, this is only done by the trace one, while the maximum eigenvalue contrasts that there are exactly "r" cointegration vectors.
What we want to know exactly is if in your programme both of them use the same nule hypothesis ( at most "r" cointegration vectors) or the trace one uses that one, while the maximum eigenvalue uses the other one ( exactly "r" cointegration vectors).
Seeing the results we have got, we think that both of them use the same one cause, if it isn´t like that, the contrast of the maximum eigenvalue would say it exists 0 and 1 cointegration vectors at the same time.
Thanks
Javi
---------------------------------
LLama Gratis a cualquier PC del Mundo.
Llamadas a fijos y móviles desde 1 céntimo por minuto.
http://es.voice.yahoo.com
17 years, 1 month

Johansen
by Javier García

At this moment we are doing a cointegration analysis and a doubt has come up
about the nule hypothesis that your programme uses in the Johansen test.We
don´t know if the maximum eigenvalue uses the same nule hypothesis as the
trace contrast( as for instance, Eviews does) , I mean, if both of them
contrast the hypothesis that there are, at most, "r" cointegration vectors
or, on the contrary, as we have read in some papers, this is only done by
the trace one, while the maximum eigenvalue contrasts that there are exactly
"r" cointegration vectors.
What we want to know exactly is if in your programme both of them use the
same nule hypothesis ( at most "r" cointegration vectors) or the trace one
uses that one, while the maximum eigenvalue uses the other one ( exactly "r"
cointegration vectors).
Seeing the results we have got, we think that both of them use the same one
cause, if it isn´t like that, the contrast of the maximum eigenvalue would
say it exists 0 and 1 cointegration vectors at the same time.
Thanks
Javi
17 years, 1 month

hello
by javier garcia enriquez

Hello,
I am a student of Doctorate in Econometrics from the Basque Country University. I know Gretl since 2004 (when I studied Economics degree) and, at present, I use Gretl in my investigation (time series analysis) because I find it enough easy and useful. Thanks to all you, especially to its creator, Allin Cottrell.
Sincerely you
Javi
---------------------------------
LLama Gratis a cualquier PC del Mundo.
Llamadas a fijos y móviles desde 1 céntimo por minuto.
http://es.voice.yahoo.com
17 years, 1 month

Ljung-Box Q question
by adrian adermon

Hi,
I am testing the residuals from a regression for autocorrelation (for an
assignment at school), and the correlogram in Gretl gives me a different
value for the Ljung-Box Q than i'd expect. I'm using a maximum lag of 2,
and the autocorrelation coefficients are 0,7452 and 0,5167 respectively.
These are the same values that Eviews produces, but Gretl gives the
Ljung-Box Q as 17,5104, while Eviews gives 19,388, and my own
calculations gives 19,386, using the formula found in Gujarati's Basic
Econometrics, fourt edition, page 813. Does Gretl use a different
formula to calculate the Q value?
I'm using version 1.5.0, but i can't find anything about this in the
release notes for the later versions, so i'm assuming it might still be
a problem, if it is a problem and not some mistake or misunderstanding
on my part.
As i'm new to this list, i might as well briefly introduce myself. I'm a
student at Uppsala University in Sweden, currently studying economics
and statistics. Also, thank you Allin Cottrell for a great program!
/Adrian Adermon
17 years, 1 month