5:11 p.m.
Hi,
My name is Daniel Ventosa-Santaulària. During the last years I have noticed that many
practitioners perform unit-root tests, such as the ADF, without really studying the
deterministic components; they include constants and trends without being fully aware
that, in the presence of a unit root, the latter become a linear trend, and a quadratic
trend, respectively. It is important to recall that deterministic trends tend to
asymptotically dominate stochastic ones. Along with a colleague (M. Gómez), I developed a
test specifically designed to test for the presence of a drift (and a possible shift in
the drift) once there is evidence of unit root (recently published at the Journal of Time
Series Econometrics, :o) ). Such test could be considered as a companion of any unit root
test.
My proposal is the following: I would like/love to include the test in GRETL as a
Unit-Root companion test. I don't know if my proposal is too pretentious or plainly
unrealistic. Although the test has been revised by experts and published in a fine
journal, maybe the test is not important enough.
The fact is that I programmed the test using matlab code (the test is actually extremely
simple; it is a matter of simple LS regressions) but I am unable to write it in C
language. It is my first trial in a gnu/open source project and I would feel very proud
if I could collaborate in such a great program as GRETL. If there is a chance to do all
this, please follow this link to read the working paper version of the test. The
published version can be found here and a preliminary matlab code of the test can be found
here. (I prepared a much better one but I have not yet uploaded it). Many thanks in
advance for your attention; I look forward to hear from you,
Friendly,
Daniel
9:17 p.m.
IMHO the natural way would be to develop a so-called function package
which implements your test. Such a package can be uploaded to the
official gretl package server such that every gretl user can install
that package from within gretl. The code of function packages is written
in gretl's intuitive scripting and matrix language, and the packaging of
the package is done graphically from within gretl.
See section 10.6 of the gretl manual.
cheers,
sven
Am 07.02.2011 18:11, schrieb Daniel Ventosa S.:
Hi,
My name is Daniel Ventosa-Santaulària. During the last years I have
noticed that many practitioners perform unit-root tests, such as the
ADF, without really studying the deterministic components; they include
constants and trends without being fully aware that, in the presence of
a unit root, the latter become a linear trend, and a quadratic trend,
respectively. It is important to recall that deterministic trends tend
to asymptotically dominate stochastic ones. Along with a colleague (M.
Gómez), I developed a test specifically designed to test for the
presence of a drift (and a possible shift in the drift) once there is
evidence of unit root (recently published at the Journal of Time Series
Econometrics, :o) ). Such test could be considered as a companion of any
unit root test.
My proposal is the following: I would like/love to include the test in
GRETL as a Unit-Root companion test. I don't know if my proposal is too
pretentious or plainly unrealistic. Although the test has been revised
by experts and published in a fine journal, maybe the test is not
important enough.
The fact is that I programmed the test using matlab code (the test is
actually extremely simple; it is a matter of simple LS regressions) but
I am unable to write it in C language. It is my first trial in a
gnu/open source project and I would feel very proud if I could
collaborate in such a great program as GRETL. If there is a chance to do
all this, please follow this link
<http://econpapers.repec.org/paper/guawpaper/em200801.htm> to read the
working paper version of the test. The published version can be found
here <http://www.bepress.com/jtse/vol2/iss2/art3/>and a preliminary
matlab code of the test can be found here
<http://files.getdropbox.com/u/1307356/Arxius%20en%20la%20web/Matlab/GVS_t...;.
(I prepared a much better one but I have not yet uploaded it). Many
thanks in advance for your attention; I look forward to hear from you,
Friendly,
Daniel
_______________________________________________
Gretl-devel mailing list
Gretl-devel(a)lists.wfu.edu
http://lists.wfu.edu/mailman/listinfo/gretl-devel
4821
days inactive
4821
days old
1 comments
2 participants
participants (2)
-
Daniel Ventosa S. -
Sven Schreiber