Hi all, I just saw that when the time-dimension of a panel data set is restricted, the 'sample' display at the bottom of the main menu does not show the right time indices. See the screenshot attached where I restricted time to the range 2<=time<=10 but the shown indices are still <Panel: sample 1:1 - 10:9>. Of course, the actual data set in the background is correctly restricted.

Am 11.11.2019 um 01:36 schrieb Allin Cottrell: > On Sun, 10 Nov 2019, Artur Tarassow wrote: >> I just saw that when the time-dimension of a panel data set is >> restricted, the 'sample' display at the bottom of the main menu does >> not show the right time indices. >> >> See the screenshot attached where I restricted time to the range >> 2<=time<=10 but the shown indices are still <Panel: sample 1:1 - 10:9>. >> >> Of course, the actual data set in the background is correctly >> restricted. > > Gretl's basic "panel first period" is always 1 by construction: the > first period in the sample range. If the dataset has, for example, a > "year" variable, then "period 1" may correspond to a different year > depending on subsampling. And if you've set observation markers of the > form "individual:date" these will look "right" after subsampling.

Interesting point about the obs markers. I'm wondering whether it could make sense to apply "semi-markers" automatically when the time dimension in the panel is defined. So after a command like: setobs 4 1990:1 --panel-time that might imply setting obs markers of the form "G1:1990-1" or something like that. (G for 'group' which I think we use in gretl for the cross-section dimension.) Of course only if no other markers are defined already.