Re: [Gretl-devel] kalman filter status?
by andreas.rosenblad＠ltv.se
svetosch(a)gmx.net @ INTERNET skrev 2008-04-28 12:52:49 :
> Hi,
>
> I think I've asked about it before, but the "big" feature I'm
> missing in gretl is the Kalman filter for state-space models.
The "big" feature I am missing in gretl is quantile regression.
Andreas
16 years, 7 months
translation time?
by Sven Schreiber
I know I can update the (German) translation anytime, but I was still wondering whether any kind of deadline is approaching. I would like to catch up before the release of 1.7.5, but of course I would like to do it rather later than sooner (standard time discounting ;-)
cheers,
sven
16 years, 7 months
Another quantile regression bug
by andreas.rosenblad＠ltv.se
I have found another bug for the quantile regression function (again using
the Windows snapshot from 2008-05-14):
1. Using the GUI, open the Ramanathan data file "data 4-1".
2. Choose Model > Robust estimation > Quantile regression...
3. Set price as dependent and srft, bedrms, baths as independent
4. .1 .2 .3 .4 .5 .6 .7 .8 .9 as desired quantiles
Output:
Model 2: Quantile estimates using the 14 observations 1-14
Dependent variable: price
With 90 percent confidence intervals
VARIABLE TAU COEFFICIENT LOWER UPPER
const 0.1 1.13915 -1.#IND0 -1.#IND0
0.2 -0.856990 -1.#IND0 296.952
0.3 82.4068 -1.#IND0 150.337
0.4 109.455 -66.0818 138.898
0.5 107.628 -71.7964 436.549
0.6 16.4441 -57.0029 402.913
0.7 133.469 14.3902 -1.#IND0
0.8 92.1248 18.5548 -1.#IND0
0.9 29.0854 -1.#IND0 -1.#IND0
sqft 0.1 0.106679 -1.#IND0 -1.#IND0
0.2 0.105390 0.0309084 0.310494
0.3 0.105390 0.0672471 0.217919
0.4 0.124895 0.0421558 0.240482
0.5 0.145085 0.0413997 0.236487
0.6 0.106017 0.0551595 0.230844
0.7 0.204062 0.0766459 0.238617
0.8 0.231369 0.0962831 0.249091
0.9 0.215295 -1.#IND0 -1.#IND0
bedrms 0.1 13.8757 -1.#IND0 -1.#IND0
0.2 14.3244 -329.464 -1.#IND0
0.3 -6.49152 -243.010 39.0050
0.4 -24.6772 -41.1064 39.8206
0.5 -6.96203 -73.5335 62.9730
0.6 12.4155 -68.6148 73.6048
0.7 -43.9708 -104.434 61.8842
0.8 -5.52340 -1.#IND0 23.2836
0.9 0.246797 -1.#IND0 -1.#IND0
baths 0.1 12.7737 -1.#IND0 -1.#IND0
0.2 14.1302 -75.9195 564.224
0.3 14.1302 -40.6269 154.674
0.4 17.9793 -63.6575 142.768
0.5 -20.3390 -97.1802 65.3846
0.6 20.6819 -90.0964 65.8094
0.7 -10.8474 -80.7260 41.8477
0.8 -68.8458 -218.410 33.5273
0.9 -33.8371 -1.#IND0 -1.#IND0
Median of dependent variable = 291.5
Mean of dependent variable = 317.493
Standard deviation of dep. var. = 88.4982
I guess -1.#IND0 should mean plus or min us infinity?
The graph is also weird (see attached PDF file).
(See attached file: graph.pdf)
Best regards
Andreas
16 years, 7 months
Bug in the new quantile regression function
by andreas.rosenblad＠ltv.se
First, thanks for the new quantile regression function in gretl. It looks
great.
However, I found a bug (using the Windows snapshot from 2008-05-14):
1. Using the GUI, open the Ramanathan data file "data 4-1".
2. Choose Model > Robust estimation > Quantile regression...
3. Set price as dependent and srft, bedrms, baths as independent
4. Set 0.25 0.50 0.75 as desired quantiles
5. Press OK: The program freezes
Best regards
Andreas
16 years, 7 months
new tracker for incompatible changes
by Sven Schreiber
Dear all,
as we discussed recently I have created a new sourceforge tracker named
"Backwards-incompatible changes", and have added a first item relating
to the changed use of the $vcv accessors. Please check it out, correct
it and/or expand it. (Allin and Jack, could you just drop some keywords
what other incompatible changes will be new in 1.7.5, if any. thanks)
I have designated the "group" field of the tracker to refer to different
gretl versions, so that later it will be possible to search for all
incompatible changes in a specific version. Currently there is only a
group "new in 1.7.5", but as time progresses there will of course be
more groups. (And there is a bogus group "changed in gretl version" that
I erroneously created and now can't get rid of anymore.)
This tracker is already visible on the sourceforge tracker page. I'm not
sure if regular users are able to submit new items (which they normally
shouldn't do), but at least you need to be logged in before doing that,
and submitted items can always be deleted later.
Before I announce this on the users list I would like to wait a bit for
feedback, however.
cheers,
sven
16 years, 7 months
Problems building and installing on Fedora 8
by Max Pyziur
Greetings,
I've been trying to get gretl running on my Fedora 8 Core 2 machine.
The pre-compile rpm fails with:
error: Failed dependencies:
libreadline.so.4 is needed by gretl-1.7.4-1gtk2.i586
even though readline is installed:
# rpm -qa | grep readline
readline-devel-5.2-10.fc8
readline-5.2-10.fc8
readline-5.2-10.fc8
readline-devel-5.2-10.fc8
I've created a symlink to libreadline.so.4
libreadline.so.4 -> libreadline.so.5.2
but that doesn't move the process forward.
I've tried compiling an rpm from source by taking
gretl-1.7.4/redhat/gretl.spec.in and modifying it and placing it in
/usr/src/redhat/SPECS, and the binary into /usr/src/redhat/SOURCES and
issueng the following command:
rpmbuild -bb --clean gretl.spec (when in the /usr/src/redhat/SPECS
directory)
I get the following two sets of errors:
1)
checking for LAPACK... no
*** Could not run LAPACK test program, checking why...
*** The test program failed to compile or link. See config.log for the
*** exact error that occured. This may mean LAPACK was incorrectly
installed
*** or that you have moved LAPACK since it was installed.
LAPACK is installed:
# rpm -qa | grep lapack
lapack-devel-3.1.1-2.fc8
lapack-devel-3.1.1-2.fc8
lapack-3.1.1-2.fc8
lapack-3.1.1-2.fc8
2) The rpmbuild fails with the following:
mkdir .libs
gcc -o .libs/gretlcli gretlcli.o complete.o ../lib/.libs/libgretl-1.0.so
-ldl -L/usr/local/lib -lz -lxml2 -lglib-2.0 -lgmp -lfftw3 -lm -lreadline
-ltermcap
../lib/.libs/libgretl-1.0.so: undefined reference to `dtrcon_'
../lib/.libs/libgretl-1.0.so: undefined reference to `dgels_'
../lib/.libs/libgretl-1.0.so: undefined reference to `dspsv_'
../lib/.libs/libgretl-1.0.so: undefined reference to `dpptri_'
../lib/.libs/libgretl-1.0.so: undefined reference to `dpptrf_'
../lib/.libs/libgretl-1.0.so: undefined reference to `dsycon_'
../lib/.libs/libgretl-1.0.so: undefined reference to `dpocon_'
../lib/.libs/libgretl-1.0.so: undefined reference to `dsyev_'
../lib/.libs/libgretl-1.0.so: undefined reference to `dgeev_'
../lib/.libs/libgretl-1.0.so: undefined reference to `dtrtri_'
../lib/.libs/libgretl-1.0.so: undefined reference to `dgetrf_'
../lib/.libs/libgretl-1.0.so: undefined reference to `dgelss_'
../lib/.libs/libgretl-1.0.so: undefined reference to `dorgqr_'
../lib/.libs/libgretl-1.0.so: undefined reference to `dsytri_'
../lib/.libs/libgretl-1.0.so: undefined reference to `dpotrf_'
../lib/.libs/libgretl-1.0.so: undefined reference to `dsytrf_'
../lib/.libs/libgretl-1.0.so: undefined reference to `dpotri_'
../lib/.libs/libgretl-1.0.so: undefined reference to `dgecon_'
../lib/.libs/libgretl-1.0.so: undefined reference to `dgeqrf_'
../lib/.libs/libgretl-1.0.so: undefined reference to `dpotrs_'
../lib/.libs/libgretl-1.0.so: undefined reference to `dgesvd_'
../lib/.libs/libgretl-1.0.so: undefined reference to `dgetrs_'
../lib/.libs/libgretl-1.0.so: undefined reference to `dgetri_'
collect2: ld returned 1 exit status
make[1]: *** [gretlcli] Error 1
make[1]: Leaving directory `/usr/src/redhat/BUILD/gretl-1.7.4/cli'
make: *** [cli] Error 2
error: Bad exit status from /var/tmp/rpm-tmp.19815 (%build)
RPM build errors:
Bad exit status from /var/tmp/rpm-tmp.19815 (%build)
It seems that these errors have been encountered before
(http://aur.archlinux.org/packages.php?ID=13752), but there is no solution
offered.
The contents of my gretl.spec file is appended below. Any help is
appreciated. Apologies if this has been covered elsewhere; if there are
links showing information as to how to get through this, that would be
great.
Max Pyziur
pyz(a)brama.com
######################
# cat gretl.spec
%define name gretl
%define ver 1.7.4
%define rel mp
%define prefix /usr
Summary: econometrics package
Name: %{name}
Version: %{ver}
Release: %{rel}
License: GPL
Group: Applications/Math
Source: ftp://ricardo.ecn.wfu.edu/pub/gretl/gretl-%{ver}.tar.bz2
Buildroot: %{_tmppath}/%{name}-%{ver}-%{rel}-root
Packager: Allin Cottrell <cottrell(a)wfu.edu>
URL: http://gretl.sourceforge.net/
Vendor: Allin Cottrell <cottrell(a)wfu.edu>
Docdir: %{prefix}/share/doc
Requires: gtk+ >= 1.2.3
%description
gretl is a free econometrics package. It comprises a shared library,
a command-line client and a gui client that uses GTK. gretl offers
several least-squares based estimators. Besides reading data files
in its own format it also reads RATS 4 databases. It has a built-in
spreadsheet for editing data, and uses gnuplot for graphing. It can
output regression results in LaTeX format.
%prep
%setup -q
%build
%configure
make
%install
rm -fr %{buildroot}
%makeinstall
%post
/sbin/ldconfig
%clean
rm -fr %{buildroot}
%files
%doc COPYING README INSTALL
%{prefix}/lib/lib*
%{prefix}/lib/gretl*/*
%{prefix}/bin/gretlcli
%{prefix}/bin/gretl_x11
%{prefix}/bin/gretl
%{prefix}/share/gretl/*
%{prefix}/share/locale/*
%{prefix}/share/man/*
16 years, 7 months
RFC: $sigma & co.
by Riccardo (Jack) Lucchetti
Hi folks,
I've been working on the manual to update the matrix chapter. Only now I
realise that there is some inconsistency in the way dollar accessors are
used, with respect to variances/covariances after estimation.
Typically, after a model is estimated, we use $sigma for the residuals and
$vcv for the parameters. However, this is not always the case; some
exceptions are well justifed, some less well.
After estimating a system, $sigma holds the covariance matrix of the
residuals and $vcv the covariance matrix of the parameters of the
structural form; this is nice and consistent with, say, ols, but is _not_
consistent with VAR/VECM estimation, where $vcv holds the cov. mat. for
the residuals and $sigma is unused. Moreover, we have an accessor ($h)
only used for garch estimation, which does the same job as $sigma (which
is unused for garch models), when you take into account that the estimate
of the conditional variance is not a scalar but a series in these models
by their very nature.
In order to fix things here, some backward-incompatible changes are
needed. If you ask me, I'd use $sigma for VAR/VECM/GARCH models
throughout; I'd scrap $h; I'd implement $vcv as the covariance matrix of
parameters for VARs. Clearly, this would definitely break a few existing
scripts.
My question to you is: do you agree with the above? I won't touch the
source anyway before Monday, because what matters in the end is Allin's
opinion (that's what benevolent dictators are for). However, what you guys
think is very important too. So, even if you don't have a strong opinion
on this, please reply!
Riccardo (Jack) Lucchetti
Dipartimento di Economia
Università Politecnica delle Marche
r.lucchetti(a)univpm.it
http://www.econ.univpm.it/lucchetti
16 years, 7 months
