4:21 p.m.
Hi devels,
it could be that I already reported/complained about some of these
earlier, but this is what I have noticed today:
1) Is there no way to add the $unit series to a panel dataset via the menus?
2) Also panel: The Beck-Katz standard errors (PCSE) fail for me with an
example (fixed-effects) model (using the example CLE.gdt data) when I
include time dummies. Now I'm aware that this could be just my ignorance
in the sense that it may not be possible econometrically -- however, ad
hoc I don't really see why, frankly.
3) The "Save as..." option in the menu really seems to do what other
(most?) programs call "save copy as...", namely the current file with
the old name remains active, not the new one with the new name.
THanks,
sven
8:47 p.m.
On Tue, 5 Nov 2013, Sven Schreiber wrote:
Hi devels,
it could be that I already reported/complained about some of these
earlier, but this is what I have noticed today:
1) Is there no way to add the $unit series to a panel dataset via the menus?
Not up till now, but it's now added in CVS: /Add/Panel unit index.
2) Also panel: The Beck-Katz standard errors (PCSE) fail for me with
an
example (fixed-effects) model (using the example CLE.gdt data) when I
include time dummies. Now I'm aware that this could be just my ignorance
in the sense that it may not be possible econometrically -- however, ad
hoc I don't really see why, frankly.
The "middle term" matrix in the PCSE is not guaranteed to be positive
definite in unbalanced panels, and it seems that including time dummies
increases the risk that it won't be p.d. I find that cutting back on the
number of time dummies can help, e.g. the following works:
<hansl>
open CEL.gdt
set pcse on
genr timedum
list Tdum = dt_4 dt_5 dt_6
# PCSE fails if you include dt_7
panel y const n s Tdum --robust
</hansl>
3) The "Save as..." option in the menu really seems to do
what other
(most?) programs call "save copy as...", namely the current file with
the old name remains active, not the new one with the new name.
Yes.
Allin
3:05 p.m.
Am 05.11.2013 21:47, schrieb Allin Cottrell:
On Tue, 5 Nov 2013, Sven Schreiber wrote:
> Hi devels,
>
> it could be that I already reported/complained about some of these
> earlier, but this is what I have noticed today:
>
> 1) Is there no way to add the $unit series to a panel dataset via the menus?
Not up till now, but it's now added in CVS: /Add/Panel unit index.
Thanks!
> 2) Also panel: The Beck-Katz standard errors (PCSE) fail for me with an
> example (fixed-effects) model (using the example CLE.gdt data) when I
> include time dummies. Now I'm aware that this could be just my ignorance
> in the sense that it may not be possible econometrically -- however, ad
> hoc I don't really see why, frankly.
The "middle term" matrix in the PCSE is not guaranteed to be positive
definite in unbalanced panels, and it seems that including time dummies
increases the risk that it won't be p.d. I find that cutting back on the
number of time dummies can help, e.g. the following works:
ok, I see.
> 3) The "Save as..." option in the menu really seems to
do what other
> (most?) programs call "save copy as...", namely the current file with
> the old name remains active, not the new one with the new name.
Yes.
:-)
what I meant to insinuate, why then not call it "save copy as..." ?
-sven
4063
days inactive
4064
days old
2 comments
2 participants
participants (2)
-
Allin Cottrell -
Sven Schreiber