[Date Prev][Date Next][Thread Prev][Thread Next][Date Index][Thread Index]

Re: long term variation and instrumental drift

In reply to Michael R.:
  Ah.  Your reply on 020607 seemed to imply that you were
calculating transformation coefficients on a frame-by-frame
basis.  Nightly values are somewhat better.
  I know you are aware of many of the issues involved,
but in my mind one big problem is that others are not.
These 'novices' are then using the output from your pipeline
and assuming the results are correct.  For example,
Michael K. is assuming the V,Ic are *raw*, and they are not.
Neither are they necessarily properly transformed.  I have
real concerns over creating transformation coefficients
from non-photometric nights.  As you mentioned, one way is
to create an 'init' file that includes the current best
transformation coefficients, so that you can hold them constant
from night-to-night.  Another solution is to only produce
raw files, and then assume some further step will apply
transformations and zeropoints, but it is always easier to
work with what appear to be fully processed files.
  My gut feeling is that the night-to-night jumps seen by
Tom and others in the reductions are caused by non-photometric
conditions or improper flatfielding.  For example, are
non-sensical transformation coefficients created from such
nights 'flagged', or just blindly applied to the datasets?
  Again, and this cannot be stressed enough, both the data
collection and the data processing have to be considered
'engineering' at this stage.  Be careful in analyzing the
data and in how you present those data to the
outside world.  This is not to fault anyone involved!