Switching this from users to devel.
On Tue, 29 Jun 2010, Allin Cottrell wrote:
On Tue, 29 Jun 2010, Sven Schreiber wrote:
> Actually as a goal (2.0-ish?) I think it would be nice if sample
> restrictions worked better with panel data. For example, something like:
> smpl 1960 1980 --idtime=year
This idea looks quite attractive at first, but I'm afraid it gets
confusing fast. The trouble is that we have two modes of setting
the sample, which behave somewhat differently and have different
support behind the scenes...
Some more thoughts on this. By "behave differently" above I mean
(A). When you do
smpl <startobs> <endobs>
this always replaces any previous sample setting that was done in
the same way; and it always compounds (and never replaces)
sub-sampling done via --restrict. This may sound a bit awkward but
in fact I'm pretty sure it's the most intuitive behavior.
(B). When you use --restrict with smpl it compounds any
sample limitation that's already in place, unless you also give
the --replace option.
In effect, the two sampling modes are nested. Use of --restrict
carves out a subset of the data, and "smpl <startobs> <endobs>"
operates relative to the current subset.
So to achieve what you're talking about for panel data in such a
way that it's consistent with what happens with non-panel data
(and not confusingly inconsistent!), we'd need to emulate the
non-panel behavior of "smpl <startobs> <endobs>" quite exactly.
I guess this may be doable, but it won't be very easy.