Re: when to discard measurements

[aah@nofs.navy.mil]

Michael R. wrote:
> My rule when working with WZ Sge last summer was to look at the 
>scatter in the relative magnitude of a pair (or three) of comparison
>stars in the field: if there was a spike, or jump, or period of 
>increased scatter in those relative magnitudes, then I would discard
>the corresponding measurements of the program star.
  This was a good rule for this specific star, since the brightness
flickering of WZ Sge itself was of such short duration that you could
not tell from looking at the star whether a datapoint was suspicious
or not.  Michael's rules at least gave a consistent method of removing
points with likely error.  However, you have to be careful about
applying such a method in the general case.  For example, you need
to quantify the rejection criterion: what consitutes a 'spike'?  How much
is enough increase in the scatter to constitute rejection; why not just
assign larger error to the differential photometry?  That is why there
is still a need for basic experience and why you should not arbitrarily
reject an observation just because it lies outside one sigma of a mean curve.
Instead, I would study the data collection process and find out *why*
there were periods of spikes and jumps.
Arne