[daip] JMFIT/IMEAN booby trap

[Eric Greisen]

On 07/09/2014 10:21 PM, R. Craig Walker wrote:
> I just stumbled on a booby trap for the unwary (like me).  Nothing is
> formally being done wrong, but in practice, people could be led badly
> astray.  Consider the following sequence, which is very common for me:
>
> IMEAN, full image, DOCAT=1
>     Puts correct ACTNOISE in header.
> IMEAN, tight window to measure flux density, DOCAT=1 because I'm not paying
>     attention to it - never did until very recently.
>     Replaces ACTNOISE with a much higher value
> JMFIT.  Uses bad ACTNOISE to calculate errors.
>
> I just did this.  Fortunately I noticed the statement about ACTNOISE being
> used and knew I had likely measured a flux density with a tight box.  I
> reran
> IMEAN with a wide open box, replacing ACTNOISE.  Then JMFIT came up with
> much smaller errors.  My worry is that people will blunder into the bad
> case very easily.  I suggest protecting against it somehow.  Is there any
> reason not to default to recalculating the noise with a histogram fit?
> Unfortunately RADIUS doesn't really have room for another option.


I changed IMEAN to avoid writing ACTNOISE and ACTMEAN unless the area is
at least 20% of a plane and the volume 10% of a cube.  DOCAT=2 says do
it anyway.

This should help.  But the fitting-task message about RMS remains 
significant.

Eric