VECM, Impulse response: gretl crashes
by Joachim Smend
Dear Allin, dear Riccardo, dear other contributors to the
development of gretl,
as a newcomer to the gretl user forum first of all I’d like
to say how impressive gretl already is – in its functionality, its
user-friendliness, but also in its flexibility to incorporate other programmes
or own functions through its own scripting language. For an average applied
economist, almost everything is there and easily available – and that as a
result of the efforts of just a few people. Congratulations and thank you very
much – you’ve really succeeded in creating a wonderful programme! I’ve started
using it at version 1.20 or so, and am absolutely amazed at the speed of its
development!
As for SVAR functionality, which would indeed be a great new
item: I’ve seen in a recent posting by Riccardo that there’s already a
collection of scripts available and that one may check how they work out by
oneself. May it be possible to receive these scripts?
And a question on GARCH models: is an addition of other
types of models (such as TGARCH, GARCH-M, where Lee Adkins provides a solution
by using the MLE in his ebook, or EGARCH, PARCH, CGARCH), and possibly an
increase in the possible order of GARCH models planned?
Maybe the latter has already been discussed – if yes, I’m
sorry for bringing it up again!
Thank you very much!
Best,
Joachim
(jsmend(a)hotmail.com)
_________________________________________________________________
http://redirect.gimas.net/?n=M1003xWin7Geschenk2
Hol dir das Gratis-Geschenkpaket von Windows 7 für deinen PC ab!
16 years
Multiple Imputation analysis
by Nathan Paxton
Hi all,
I've recently discovered gretl, having been a fairly elementary user of R and Stata for a few years.
I do work with multiple imputation datasets, and I generate them in R, using the Amelia package. But the sorts of tests I want to run aren't easily done using any of the more canned packages, so I end up having to write long code files to implement Rubin's (1986) combination rules for MI parameter estimates.
I don't need to use gretl to do the actual MI, but I wonder if there's any facility in gretl for doing analysis on separate MI-generated datasets and then combining the results according to the Rubin rules. That is, can gretl be told easily to run a procedure m times for the m existing datasets and then combine the results to get 1 parameter estimate?
Thanks in advance for your help.
Best, -Nathan A. Paxton
----------
Nathan A. Paxton, Ph.D.
Lecturer
Dept. of Government, Harvard University
napaxton AT fas DOT harvard DOT edu
http://www.fas.harvard.edu/~napaxton
========================================================
Stand up for hope, faith, love
But while I'm getting over certainty
Stop helping God across the road like a little old lady.
—U2
========================================================
16 years
Problem with Excel files
by Henrique Andrade
Dear Gretl community,
When I try to open a excel file (.xls) the following error message appears:
"First char of varname (0x0) is bad
(first must be alphabetical)
Please rename this variable and try again"
My data is quarterly (using the format 1980.1 for the first quarter of
1980) and my variables name are something like this: bcbXXXX (where XXXX is
a number).
Best,
Henrique
16 years
weird bootstrap behavior
by Summers, Peter
Folks,
I've just noticed some very strange behavior when I try to generate studentized bootstrap confidence intervals in a model with only a constant (this is problem 5.6 in Davidson/MacKinnon's ETM). The model is earnings data regressed on a dummy variable for a particular income group, so it's just estimating the mean.
Here are the results from running a studentized bootstrap:
For the coefficient on G3 (point estimate 27973.6):
Studentized 95% confidence interval = 16009.7 to 17410.2
Based on 999 replications, using resampled residuals
The graph of the sampling distribution, on the other hand, is centered roughly on the point estimate (see 'studentized.pdf', attached).
When I do a regular bootstrap confidence interval, I get
For the coefficient on G3 (point estimate 27973.6):
95% confidence interval = 42857 to 44843
Based on 999 replications, using resampled residuals
With the distribution in 'bootstrap.pdf'.
Things seem to work ok in a model with all three group dummies and no constant.
PS
===============================
Dr. Peter Summers
Assistant Professor
Department of Economics
Texas Tech University
===============================
16 years
Fitted, actual plot for TS data
by Jaroslaw Gramacki
Suppose we have a Time series Data in gretl generated as a simple TS:
z_t = p * z_{t-1} + u_t; -1 < p < 1
x_t = a + b * t + z_t;
Two scenarios:
1. We build an OLS model (no lags, static model)
2. We build Time series (say Cochrane-Orcutt) model (no lags, static model)
In scenario 1: If we make Graphs -> Fitted, actual plot -> Against time, we obtain "a
normal" regression line.
In scenario 2: If we make Graphs -> Fitted, actual plot -> Against time, we obtain in
fact a static 1-step ahead forecast plot.
I understand the difference beetwen time series and cross-sectional data but why two
different plots are obtained via exactly the same sounded menus?
Regards,
Jaroslaw Gramacki
16 years
cross-correlation output for several variables
by Artur T.
Good evening,
I've got a suggestion for improving the cross-correlation output. At the
moment it is only possible to get an output of the result for only two
variables under consideration. Maybe it is worth to implement the
possibility to set a variable x for which the cross-correlation with
several other variables is computed (similar to the multiple scatter
graph idea). The current output format can remain as it is at the
moment; one just could add more columns for the additional variables
under consideration.
Hopefully this is not hard to program and you also like the suggestion.
Best,
Artur
16 years, 1 month
