[daip] [Fwd: DBCON/IMAGR question]

[Eric Greisen]

The important thing is to have data weights represent 1/uncertainty**2
with an arbitrary scale.  When combining two data sets, those scales
could be different so one would want to reweight one or both.  Modern
FILLM reads VLA data with weights suitable for calibration and then
they should be on the same scales already when combining
configurations or different runs.  In the old bad days, FILLM weighted
solely by integration time and so the scales in different runs could
be quite different.  If I were to write DBCON today, it would probably
not have the rescale option.

Of course, we take no responsibility for weights from non-VLBI and
non-VLA observations and depend on the programs that bring those data
into AIPS.

Eric Greisen