4:14 p.m.
Hi everybody,
even though the release candidate was posted on the users' list, I'm
sticking to devel for such very technical and narrow issues.
In the $xlist reference it is claimed that in a system of equations the
predetermined variables are included. This is not true, and for good
reason. (They are in $system.predet_vars and in some of $system.list1
... list<N>.)
I would delete the two words from the funcref if nobody objects.
cheers
sven
5:06 p.m.
Am 07.01.2021 um 17:14 schrieb Sven Schreiber:
In the $xlist reference it is claimed that in a system of equations
the
On a related note in terms of the $system accessor bundle, here's a
reminder that (only) after the 'var' command there is a $compan accessor
which does not exist as $system.compan.
Instead the identical matrix is stored as $system.A. Maybe this would be
the time to make $system.compan an alias for $system.A in that case, in
order to mirror the direct $compan accessor.
thanks
sven
10 p.m.
New subject: more system stuff (Re: Re: very minor $xlist documentation bug)
Am 07.01.2021 um 18:06 schrieb Sven Schreiber:
Am 07.01.2021 um 17:14 schrieb Sven Schreiber:
> In the $xlist reference it is claimed that in a system of equations the
On a related note in terms of the $system
And yet another thing in the equation system area; consider this example
script:
<hansl>
open denmark
system
equation LRM 0 LRY # 2 RHS terms
equation IBO 0 IBO(-1) seasonals() LRY(-2) # nominally 7 RHS terms
end system
estimate system method=ols # of course S4 is dropped from eq2
bundle bs = $system
eval $coeff # contains 2+7 elements, _last_ one zero!
eval bs.list2 # contains 7 RHS elements, S4 penultimate
</hansl>
OK, what do I mean? S4 is auto-dropped from eq2 due to collinearity,
fine. First question: should $coeff really contain a zero coefficient
for it? I guess it could, but OTOH the standard gretl printout does not
include dropped regressors showing zero coefficients. Also, in the plain
single-equation case gretl's $coeff does _not_ contain such an extra entry.
So I think the extra zero should be scrapped from $coeff in the system case.
Next, the auto-dropping of S4 is not reflected in the equation's
variable list ($system.list2, nor in .xlist or .instr_vars, by the way).
Again, I think it should be removed, for example for consistency with
ols and the plain $xlist there.
If for some reason these items were not removed, the ordering should be
matched, which is not the case currently. (See above, last entry vs.
penultimate entry.)
I hope that's it about issues with the system command :-)
cheers
sven
1:35 p.m.
Am 07.01.2021 um 17:14 schrieb Sven Schreiber:
Hi everybody,
even though the release candidate was posted on the users' list, I'm
sticking to devel for such very technical and narrow issues.
In the $xlist reference it is claimed that in a system of equations the
predetermined variables are included. This is not true, and for good
reason. (They are in $system.predet_vars and in some of $system.list1
... list<N>.)
I would delete the two words from the funcref if nobody objects.
This is now pushed to git (master).
thanks
sven
1471
days inactive
1475
days old
3 comments
1 participants
participants (1)
-
Sven Schreiber