I'm wondering why 'restrict' isn't possible after GMM. Performing
Wald-type tests should not pose extra problems once we have the
estimator and the covariance matrix, right? Or am I missing something?
(with gretl build 22 Jan on Linux)
In a script window the execute button always seems to apply to the
selected region if something is selected, instead of the whole script file.
Is this a bug or a feature? I can imagine both, just want to make sure.
Yes, I know that sesampling rows of a matrix is something one can do easily
with the existing gretl tools. See the discussion here:
However, I still think it should be included natively in gretl by extending
the resample() command so that it can be used with mxn matrices too. It
would be easier, more natural and less error-prone to use B = resample(A)
instead of writing special code every time.
cottrell(a)wfu.edu @ INTERNET skrev 2008-01-29 04:19:50 :
> On Mon, 28 Jan 2008, andreas.rosenblad(a)ltv.se wrote:
> > I think that at least some of feature requests should be >
> targeted before the release of a new version. At least feature >
> requests 1854919 and 1855731, I hope.
> Re. 1854919: Resampling rows of a matrix is something you can do
> easily with the existing gretl tools, there's no need for a special
> function. (We're systematizing the documentation for all gretl
> functions, so hopefully this sort of thing should become more apparent.)
> For example:
> nulldata 5
> scalar T = 20 # rows of original matrix
> scalar R = 20 # resampled rows: you choose
> matrix A = mnormal(T,2)
> matrix rsel = int(muniform(R,1)*T+1)
> matrix B = A[rsel]
> print A B rsel
> One small thing: the line "matrix B = A[rsel]" works fine, but it
> might be more transparent if it read
> matrix B = A[rsel,]
> Unfortunately the latter variant doesn't work in the current release
> because of a stupid copy-n-paste bug in the code for parsing matrix
> sub-slices, but that's fixed in CVS.
> Gretl-devel mailing list
Not much is happening recently on the gretl lists and trackers (which is
not a bad sign).
Are there any further features in development right now? If not, would
it be an option to make another release, after leaving some time for
translators? A couple of bugs have been fixed since the last release,
including at least one crash (panel-related). Or should some of the
feature requests be targeted?
>> Comment By: Allin Cottrell (allin)
> Date: 2008-01-28 22:22
> Logged In: YES
> Originator: NO
> Sorry, not worth the bother. Confidence bands are shown in the GUI
> graph for corrgm.
Fair enough. Before I create the follow-up request in the tracker, let
me ask here:
What about a '--quiet' or '--no-graph' option to the corrgm command that
would omit the (ASCII) graph and only print the numbers (with
significances). The graphs tend to clutter my script outputs without any
real value added (given that they are not PACF but ACF).
svetosch(a)gmx.net @ INTERNET skrev 2008-01-27 16:54:20 :
> Or should some of the feature requests be targeted?
I think that at least some of feature requests should be targeted before
the release of a new version. At least feature requests 1854919 and
1855731, I hope.
I'm back on Linux (let's see for how long...), and the first thing I did
(well, almost) was to check out the gretl source from cvs to update
de.po and compile. It seems to me that some new strings relating to the
recent working directory changes are not marked for translation; I
noticed some strings in de.po that contain "working dir", but still it
doesn't seem to cover everything. Or am I missing something?
Recently I have proposed to add a new feature to gretl in this thread:
Basically I have thought about a "project" manager to get a better script
handling, automatic parameters loading, fast script running. This idea has
several links with function packages and sessions concepts, that's why I
decided to post It here to get any remarks.
At the start gretl should read a file like this:
<descr>This is cool</descr>
to allow some improvements such as:
- fast switching among projects;
- quick loading of several runs configuration for the same projects
- running the default master script from any other file of the project;
- no include statement needed
- no nulldata/open statement needed
- easy export to function package of the entire project
Regarding to the concept of sessions Sven talked about in the thread
previously posted, I don't see any dangerous clashes cause the session
would still behave as a snapshot, eventually It should contain the reference
of the project.
What do u think about this? Useful? Useless? Let me know
futurdorko has made some good suggestions in regard to gretl's
built-in script editor lately, and I've made some changes in CVS.
If people could beat on the new stuff and expose any bugs that
would be great.
The main thing is new handling of Tab and Enter. I've tried to
make these keys behave pretty much as they do in, say, Emacs' C
Tab: when the Tab key is pressed at the beginning of a line (or in
white space to the left of any non-space characters on a given
line), instead of simply entering a Tab character, try to
determine what the logical indent is for the current line, and
enter enough space characters to go to that indent.
Enter: when Enter is pressed, check the indentation of the line
being completed, and if it's not correct, adjust it.
So, for example, if you type
<Tab>if x > 3<Enter>
<Tab>print "x > 3"<Enter>
<Tab>print "x <= 3"<Enter>
if x > 3
print "x > 3"
print "x <= 3"
This assumes the default of 4 spaces per "Tab", but if you
right-click in the editor window you get a configure option in the
popup menu that lets you set this value from 2 to 8 spaces.
You also get the option of turning off the "smart" Tab behavior
if you don't like it; and if you have an existing script open in
the editor you get the option of applying "canonical" indentation
to the whole thing.
Is it because I am too tired already, or should p.37 of the user guide
say "n>=m" instead of "n<=m" in the context of explaining the fdjac
function? (With respect to the delta method, meaning that you cannot get
out a covariance matrix which is higher-dimensional than the one you
On the same page, n is also used as the sample size, but I'm not sure if
that really causes confusion for anybody.