[Gretl-devel] Re: panel restriction trouble
On Sat, Jul 15, 2023 at 2:53 PM Sven Schreiber
<sven.schreiber(a)fu-berlin.de> wrote:
Hello panel devotees,
I believe I'm seeing a bug in cumulated sample restrictions in the panel
context [...]
Here's a minimal example of what you describe:
<hansl>
nulldata 32
# panel: T = 8, N = 4
setobs 8 1:1 --stacked-time-series
genr time
smpl time < 6 --restrict
# the following gives "No observations would be left!"
# smpl time == max(time) --restrict
# but the next two lines work
series d = time == max(time)
smpl d --dummy
print -o
</hansl>
The trouble is that in the first case gretl is evaluating "max(time)"
relative to the unrestricted dataset (which would indeed leave no
observations).
This should be fixable -- if we reckon it's truly a bug -- but it's
the sort of thing where one has to tread very carefully.
Allin