Crash using omit command on ordered probit
by Artur T.
Dear all,
I attached a sample script and my data. Using the "omit" command after
estimating a ordered probit results in a crash on both Ubuntu (64-bit,
current cvs) and Win 7(64-bit, version from last week or so).
Hope these information are helpful.
Best,
Artur
10 years, 10 months
default(null) values tests
by oleg_komashko＠ukr.net
Dear all,
Below there are several additional tests of the default/null values behavior both inside and outside functions. Some results looks strange.
1) With lists
(i) in a script
nulldata 1
list znull = null
isznull = exists(znull)
nznull = nelem(znull)
print isznull nznull
(ii) inside a function
function void listtestfun (scalar x, list z[null]) isz = exists(z) nz = nelem(z) print isz nz end function
listtestfun(1) We have
isznull = 1.0000000 nznull = 0.0000000
and
isz = 1.0000000 nz = 0.0000000
as expected
2) with strings (i) in a script
string snull = null issnull = exists(snull) nsnull = strlen(snull) print issnull nsnull
The behavior is the same as for lists and is as expected to be
Also, emptycheck = snull=="" print emptycheck
We see that ' string snull = null' actually creates the "" string
(ii) inside a function
function void strtestfun (scalar x, string s[null]) iss = exists(s) ns = strlen(s) print iss ns end function
strtestfun(1,"oleh")
OK!
strtestfun(1)
we have
The symbol 's' is undefined *** error in function strtestfun, line 2
It seems now the problem is not with exists() but is like this: a function does not defines a string variable in contrast as it does with default null lists
3) catch string
function void testfun3 (scalar x, string s[null]) catch string s err = $error print err end function
testfun3(1)
testfun3(1,"a")
Please, explain why we have "0" in the first case and "2" in the second
Oleh
P.S. I have tried to do something with formatting but I have nowhere to check. Please, inform me
10 years, 10 months
seasonal / periodic dummies in panel data
by Sven Schreiber
Hi,
I have a panel dataset which is monthly in the time dimension, and I
have done:
setobs 12 1991:1 --panel-time
which is recognized as the subsequent invocation of '$obsdate' shows.
However, the menu item "add/periodic dummies" is greyed out. Any reason
for that?
Also, the seasonals() function gives strange error messages; here's an
example script:
<hansl>
open abdata.gdt
setobs 4 1900:1 --panel-time # just an example
series seas = seasonals(1)
</hansl>
thanks,
sven
10 years, 10 months
impossible to enter values in new/empty dataset
by Sven Schreiber
Hi,
I'm starting to ask myself what is wrong with this particular system;
first the gretl crash (opening the Stata file, see yesterday's thread),
and now this:
Start a new panel dataset, with 3 units and 4 time periods; gretl asks
"do you want to enter data"; yes, and then there's a window "edit data";
I enter "5" in obs 1:1 and hit return; then I want to enter another
number like "3" for obs 1:2, but when I hit return the number vanishes!
Same with any further numbers.
Actually I wanted to report a problem with the pxsum() function in
subsampled datasets, but this here got in the way while testing...
Anyway, to be sure I uninstalled the old gretl version in the "official"
Windows way, and reinstalled gretl before I did this.
good night,
sven
10 years, 10 months
Ssuppressing the "The convergence criterion was not met" message
by Marcin Błażejowski
Hi,
I wonder if it would be possible to add option to the 'catch' command in
order to suppress messages like 'The convergence criterion was not met'
(example code below).
Regards,
Marcin
<hansl>
open bjg
set echo off
scalar maxP = 4
scalar maxQ = 4
loop for p=1..maxP --quiet
loop for q=1..maxQ --quiet
catch arima p 1 q ; lg --conditional --quiet
if ($error != 0)
print "Estimation problem"
endif
endloop
endloop
</hansl>
--
Marcin Błażejowski
GG: 203127
10 years, 10 months
modtest vs reset
by Marcin Błażejowski
Hi,
'modtest' has two "verbosity" options: --quiet and --silent, while
'reset' has only --quiet. Maybe 'reset' would have --silent option to?
Regards,
Marcin
--
Marcin Błażejowski
GG: 203127
10 years, 10 months
Translation questions
by Sven Schreiber
Hi,
there were a few accumulated changes because of my laziness, and I have
a couple of questions:
* "Wrote gfn file as\n" -- "wrote" as synonym for "saved"? (Perhaps
using "saved" also uniformly in English then?)
* Does it exceed the length in context to translate "FEVD" as "FEVD
(Prognosefehlervarianzzerlegung)", which by the way is a nice example of
German composita words...?
* And similar for "IRF" -> "IRF (Impulsantwortfunktion)"
* line 16486: "Full" meaning complete, total, no space left...?
* In "eigenvalues of X'X, largest to smallest", there is an isolated
single quote. I remember once that I had introduced something like that
in the German translation ("Gauß'sche") which made German gretl crash?
Is this safe here now?
* "note: variance proportions columns sum to 1.0" - "note" as a noun or
the imperative of the verb? (I guess both mean the same thing though...)
* "%s: invalid option argument": should it rather be "optional", or what
does this mean here (line 6206)
* line 8673: "filter": noun or verb?
* line 17450: "condition index" -- context? I'm clueless.
At the risk of being redundant, let me remind my fellow translators of
this multilingual statistical glossary:
http://isi.cbs.nl/glossary/
thanks,
sven
10 years, 10 months
