[vsnet-chat 2082] Re: Plotting light curves

[aah@nofs.navy.mil]

Gianluca Masi presented a nice case of a 1000-image dataset
for the field of V603 Aql, and asked the question of how to
present the data -- raw, binned or smoothed?
  I rarely bin or smooth data.  First, there was a reason
why you were taking so many frames -- sometimes it is because
of tracking problems or saturation, but often it is to get
high time resolution.  Once you bin, you lose that feature.
Smoothing data, since it assumes knowledge of variations
prior to and subsequent to the current data point, has the
annoying habit of introducing features that are not real.
However, I would bin before I would smooth, especially for
the data that Gianluca has presented since his scatter is
quite large (several percent) and therefore his signal/noise
per data point quite low.  In this case, improving the
signal/noise at the expense of time resolution is the
right way to go.
  Binning and smoothing only work if you have uniformly sampled
data.  Gaps are always difficult to fit since you have
no a priori information as to what happens in the interim.
  In other words, how you present your data depends on
both the quality of your data and your intended audience.
If you want the data to be used in a scientific manner,
smooth if you must in your figure, but give the raw data
in some table.
  BTW -- did you really take 1000 images on a single night?
Amazing.  I can take about one image per minute or 600-700
images in a typical night if I stay on the same object, but
200-300 frames/night is more typical.  Granted, my 1024x1024
detector may take longer to read than your CCD, but it still
means you were working pretty hard!
Arne

VSNET Home Page

Return to Daisaku Nogami

---

vsnet-adm@kusastro.kyoto-u.ac.jp