Am 08.01.2016 um 23:43 schrieb juergen.malitte(a)t-online.de:
I would make a note respective the different behavior of the command
"setobs n 1:1 --stacked-cross-section" by comparison with the
appropriate functionality in the GUI. I wouldn't say it's a mistake, but
it evokes some confusion.
Yes, I agree, and even think that some mistakes may be involved, too.
(This is with a Dec 7th snapshot.)
Opening the file in the GUI you choose the data-structure
then "gestapelte Querschnitte" (stacked cross-sectional). The result
will be the following:
That looks good.and the representation of the identifier has the format
"t:i" as expected (because the first staple of cross-sectional data
belongs to the period 1 etc.).
What worries me here is that neither in the main gretl window, nor in
the show-data window, nor in the dataset info text do I see any
indication of the t:i ordering. Implicitly we are supposed to remember
that we declared the dataset as "stacked cross-section" back at import
time. This is a recipe for confusion down the road. Instead I would
suggest to convert the data to gretl's standard i:t panel internal
ordering immediately on import, so the "stacked cross-section" would
only apply to the import source file, not to the imported data. Perhaps
that was even intended, in which case it would be a bug.
Furthermore, for me the sample description line in the main window says
"Undated: Full sample 1:1 - 2:4", where instead I think it should say
panel data. So I'm even more inclined to think there is some bug there.
In general one of my feature requests for panel data has been for a
while that the current active sample should be displayed by mentioning
the two different data dimensions explicity, not just with pure index
numbers. But sample selection in panels is not trivial, given
non-balancedness and possibly non-contiguity.
thanks for the report,