On Fri, 8 Jan 2016, juergen.malitte(a)t-online.de wrote:
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.
Suppose you have a dataset "Umsatz_Panel.txt", which contains the sales of
i =4 units over t = 2 periods, for example:
1980
2110
2200
2312
2455
2380
4200
4560
Opening the file in the GUI you choose the data-structure "panel" and then
"gestapelte Querschnitte" (stacked cross-sectional). The result will be the
following:
1:1 1980
1:2 2110
1:3 2200
1:4 2312
2:1 2455
2:2 2380
2:3 4200
2:4 4560
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.).
If you open the file "Umsatz_Panel.txt" in a script with the commands:
open "Umsatz_Panel.txt"
setobs 4 1:1 --stacked-cross-section
then you get the result
1:1 1980
1:2 2455
2:1 2110
2:2 2380
3:1 2200
3:2 4200
4:1 2312
4:2 4560
This is not the same representation as in the GUI, Gretl has changed the
format "t:i" into "i:t" !! I know, that "i:t" is the
preferred and
standard representation for panel data in Gretl. If you store the above
representation "t:i" (received over the GUI) in a gdt-file and load this
again, you will then get the same result, will say the representation
"i:t".
First I was a little bit confused concerning the different behavior in the
GUI and in the script. But i think it should work identically !!
In respect of the command "setobs n 1:1 --stacked-time-series" the behavior
is the same as in the GUI !
Thanks for drawing this to our attention. The behavior of the GUI
"Dataset struture" option is unambiguously wrong for the case where
imported data are in stacked-cross-section form.
As you point out, the "i:t" form is the "preferred and standard"
representation of panel data in gretl (in fact, the only form
recognized as valid), so the effect of declaring imported data as
"stacked-cross-section" should be to reorganize the data into the
standard form, as is done by the "setobs" command.
This bug should be fixed soon.
Allin Cottrell