Gretl project on the launchpad
by Ivan Sopov
Hello, gretl developers.
I'm trying to start a translation of help files into russian on
launchpad.net as it seems to be the most suitable tool to participate
for all familiars with econometrics but not with gettext, linux. cvs,
etc.
The problem is that there is already a project for gretl on launchpad
and it is strongly prohibited to start more than one project for a
single program. I cannot contact with Constantine Tsardounis for about
a month, so I think it is time to re-assign that project to someone
else. On the irc-channel of launchpad I was told that
Our admins can re-assign the project to new owners but we'd prefer to
hear from the upstream owners. can you get one of them to submit a
question here:
https://answers.edge.launchpad.net/launchpad
But if nobody from main developers wants to register and do something
at launchpad it is possible to assign this function to me and in that
case a letter in this list will probably be enough.
I have prepared a .po-file for genr_funcs.xml and gretl_commands.xml
with the help of po4a utility and got 1511 strings for translation
(strings a rather big).
Good luck, Ivan Sopov.
P.S. My previous letter about using launchpad for translation is
http://lists.wfu.edu/pipermail/gretl-devel/2009-November/002171.html
12 years, 6 months
gretl and openmp
by Allin Cottrell
As some of you know, we're currently experimenting with openmp in
gretl. When building from CVS, use of openmp is the default (if
openmp is supported on the host) unless you pass the option
--disable-openmp to the configure script. In addition the current
snapshots for Windows and OS X are built with openmp support
(using gcc 4.4.3 and gcc 4.2.4 respectively).
This note is just to inform you about the state of play, and to
invite submission of test results if people would like to do that.
Right now, we use openmp only for gretl's native matrix
multiplication. So it'll get used (assuming you have at least two
cores) if you do matrix multiplication in a script, or call a
function that does matrix multiplication (such as qform), or use a
built-in command that happens to call matrix multiplication. If we
decide it's a good idea, we could use openmp directives in other
gretl code (but as along as we rely on lapack for much of our
number-crunching, and as long as lapack is not available in a
parallelized form, the scope for threading will remain somewhat
limited).
In a typical current use situation, with gretl running on a
dual-core machine where there's little other demand being placed
on the processors, the asymptotic speed-up from openmp should be
close to a factor of two. However, it takes a big calculation to
get close to the asymptote, and we've found that with small to
moderate sized matrices the overhead from starting and stopping
threads dominates, producing a slowdown relative to serial code.
This is similar to what we found with regard to the ATLAS
optimized blas; see
http://ricardo.ecn.wfu.edu/~cottrell/tmp/gretl_speed.html
Anyway, in case anyone would like to test I'm attaching a matrix
multiplication script that Jack wrote. Right now this is mostly
useful for people building gretl from source, since you want to
run timings both with and without MP, which requires rebuilding.
But if you're currently using a snapshot from before yesterday
(build date 2010-03-21 or earlier) you could run the script, then
download a current snapshot and run it again.
Allin
13 years, 9 months
Conference program
by Sven Schreiber
Hi,
does anybody know at what times the gretl conference starts and ends?
Since it is not trivial to get to Torun, I guess many people need that
information for their travel arrangements.
Today seems to be the last day for early payment at a reduced fee -- it
would be good to put that information about the conference schedule
online, too (or did I just miss it? -- I'm not talking about the
detailed program with paper titles, just the outline of the schedule).
And regarding the payment: The bank transfer account information is
provided on a normal unsecured web page. I hope that somebody checks
that page regularly that it hasn't been hijacked and that the account
information is still correct. Which also leads me to ask the following
question: If I miss the early registration anyway, can I pay the
(higher) fee at the conference, in cash or whatever?
Thanks for your help,
sven
13 years, 12 months
Re: [Gretl-devel] Some Bugs
by Allin Cottrell
On Fri, 18 Mar 2011, Hélio Guilherme wrote:
> I was experimenting with a small (cross-section) dataset, and I found some
> bugs:
>
> 1- Variable Deaths is treated as continuous, with disabled "Treat this
> variable as discrete" check-box.
> 2- When plotting "X-Y scatter with factor separation" using variable Team
> as factor:
> a) The data points labels only appear for the Team = 0 (when mouseover
> data point).
> b) Selecting "All data lables" plots lines connecting data points
> (attached screenshot).
Thanks for the report. These things should now be fixed in
CVS/snapshots.
Allin Cottrell
14 years
Some Bugs
by Hélio Guilherme
Hi Allin, Jack and the rest of the world :),
I was experimenting with a small (cross-section) dataset, and I found some
bugs:
1- Variable Deaths is treated as continuous, with disabled "Treat this
variable as discrete" check-box.
2- When plotting "X-Y scatter with factor separation" using variable Team
as factor:
a) The data points labels only appear for the Team = 0 (when mouseover
data point).
b) Selecting "All data lables" plots lines connecting data points
(attached screenshot).
This was seen in both Windows XP Professional and Open SuSE 11.3 x64.
Snapshot and CVS builds.
I have a suggestion:
How about having the plot markers before the legend? I am always mouseover
those to see its data labels :).
Best Regards,
Hélio
14 years
Re: [Gretl-devel] things to test
by Allin Cottrell
On Mon, 14 Mar 2011, Artur Tarassow wrote:
> Thanks for your reply and the new cvs, Allin.
You're welcome.
> I've got an issue left. Unfortunately, I do not have access to the Harbo
> et al article. I am just a little bit confused by what is meant by
> "exogenous"?
> In the case one has weakly exogenous I(1) variables, I used to take the
> critical values (for all 5 cases) reported in "Pesaran, M.H., Shin Y.,
> Smith, R.J. (2000). Structural analysis of vector error correction
> models with exogenous I(1) variables. Journal of Econometrics 97:293-343."
>
> Are those critical values provided by Harbo et al. also applicable for
> this scenario with weakly exogenous I(1) variables, or are they only
> useful for unrestricted I(0) variables (for the cases were one can find
> nuisance-free asymptotic critical values)??
In my understanding, yes, they are applicable for weakly exogenous
I(1) variables. However, I'm not totally clear on the apparent
contradiction between Harbo et al and Pesaran, Shin and Smith.
That is, it seems at first sight that PSS gaily produce tables of
critical values for cases that Harbo et al showed to be infested
with nuisance parameters. I need to read the PSS article more
carefully; I think the resolution (if it really is a resolution)
is that PSS mean something different by their definition of the 5
"cases" from what Johansen meant.
I can say this much for sure: Harbo et al are thinking of a setup
(their "partial system") where the extra variables are added (a)
in levels form, restricted to the cointegration space, and also
(b) in first-differences, unrestricted. As Sven has noted, this
corresponds to the Johansen-standard treatment of deterministic
terms (where, for example, the presence of a restricted trend
implies the presence of an unrestricted constant).
Allin
14 years
Re: [Gretl-devel] things to test
by Allin Cottrell
On Sun, 13 Mar 2011, Artur Tarassow wrote:
> Thank you for the new features.
And thanks for testing.
> I've got a question regarding the new critical values which gretl
> reports now. I estimate this vecm:
>
> -------------------
> open denmark.gdt
> coint2 4 LRM LRY ; IBO IDE --rc
> -------------------
>
> and obtain the following output:
>
> -------------------
> Johansen test:
> Number of equations = 2
> Lag order = 4
> Estimation period: 1975:1 - 1987:3 (T = 51)
> Case 2: Restricted constant
> Exogenous regressor(s): IBO IDE
>
> Log-likelihood = 402.981 (including c: 258.25)
>
> Rank Eigenvalue Trace test 80% 90% 95% 99%
> 0 0.40735 36.958
> 1 0.18250 10.277
> --------------------
>
> for which no critical values are reported.
Oof, they were being reported on stdout, not where they should be
going. I didn't notice because I was using gretlcli, where
everything goes to stdout. Fixed in CVS; new snapshots available
shortly.
> Secondly, the following model is estimated with the corresponding
> p-values; but still with this note at the end:
>
> --------------------
> coint2 4 LRM LRY ; IBO IDE --ct
>
> Johansen test:
> Number of equations = 2
> Lag order = 4
> Estimation period: 1975:1 - 1987:3 (T = 51)
> Case 5: Unrestricted trend and constant
> Exogenous regressor(s): IBO IDE
>
> Log-likelihood = 408.691 (including c: 263.959)
>
> Rank Eigenvalue Trace test p-value Lmax test p-value
> 0 0.45555 43.161 [0.0000] 31.007 [0.0002]
> 1 0.21205 12.154 [0.0005] 12.154 [0.0005]
>
> Note: in general, the test statistics above are valid only in the
> absence of additional regressors.
> --------------------
That result is expected. Since we can't find nuisance-free
asymptotic critical values for "Case 5" plus exogenous vars, we
fall back on reporting the no-exogenous-variables p-values with a
warning. Maybe we could do better; in fact we might be better just
saying that we can't do this test.
Allin
14 years
copy and paste data
by Sven Schreiber
Hi,
a colleague of mine who has just started to use gretl was
asking/suggesting that new data which lives in a spreadsheet could be
pasted into gretl (presumably into the data editor). He was mentioning
Stata's capabilities.
I think something like this was dicussed before, but I thought I should
bring it up again. It's a more "visual" approach than importing the
data the usual way, and I think it could be attractive for many users.
cheers,
sven
14 years
things to test
by Allin Cottrell
There are various new things going on that would be nice to have
tested as thoroughly as possible. I'll mention 3 points.
1) The build for MS Windows: I've updated my cross-compiler
to gcc 4.5.2 and have built new blas and lapack DLLs using
lapack 3.3.0. I've also updated the gnuplot build to current
CVS. Let me know if this has broken anything, please.
2) Tobit, intreg, heckit: We've re-implemented the tobit command
and we now support user-specification of the limit(s) for left-
and/or right-censoring. In addition Jack has coded the analytical
Hessian for these models and we're experimenting with
Newton-Raphson iteration rather than BFGS for finding the MLE. If
you work with models of this sort, please try out the new code.
3) Johansen cointegration test in the presence of exogenous
regressors. For a long time we've reported the standard asymptotic
p-values for the trace and lambda-max tests, while issuing a
warning that these are not valid when exogenous regessors are
included. We now (after a useful correspondence with Soren
Johansen) try to do better: we report critical values for the
trace test based on Harbo, Johansen, Nielsen and Rahbek,
"Asymptotic Inference on Cointegrating Rank in Partial Systems"
(Journal of Business and Economic Statistics 16/4, October 1998).
The critical values were re-simulated in gretl with 50,000
replications.
There remain some doubts about what to do (a) if the exogenous
variables are just dummies, and (b) in the cases where the
highest-order deterministic term is unrestricted (i.e.
unrestricted constant, or constant plus unrestricted trend). Harbo
et al show that the distribution of the trace test contains
uneliminable nuisance parameters in the latter cases (and Johansen
is not confident that CATS is doing the right thing).
Allin
14 years
new forecast plot
by Allin Cottrell
Talha asked me a while back about getting the sort of forecast
plot you see in many of the textbooks -- with the confidence
interval for mean Y widening, the further you get from mean X.
This is now implemented in CVS/snapshots, for simple OLS
regressions on cross-sectional data. In the GUI forecast dialog,
select the option of showing the confidence interval for mean Y to
see the systematic variation in the confidence band clearly.
(In many cases this variation is hard to detect if you view
the confidence interval for actual Y, since the residual
variance tends to swamp the effect of parameter uncertainty.)
Allin
14 years
