[daip] question to rescaling in IMAGR

[Eric Greisen]

Eva Schinnerer writes:

 > I have a question regarding the proper noise estimate for a rescaled image 
 > produced by IMAGR. Fabian and I have been discussing this, but we cannnot 
 > come to a conclusion.
 > 
 > As I understand the rescaling option in IMAGR [IMAGPRM(5) - IMAGPRM(7)], 
 > this is a way to take into account that the dirty beam does have low level 
 > wings when using data from multiple arrays with similar integration times. 
 > So, for example, for our 2.5hr D array + 3hr C array observations, I get a 
 > scaling factor of ~ 0.4 when using a robust parameter of 0. I interpret 
 > that as an adjust of the units, as the area of the dirty beam is larger 
 > than the area of the clean beam used for the the convolution of the CLEAN 
 > components. Is that correct?
 > 
 > The problem is now how and where to estimate the correct rms:
 > (2) in the rescaled CLEANed image. 
 > or
 > (1) in a CLEANed image that has not been rescaled, however, was CLEANed in 
 >     the same way as the rescaled one.
 > 
 > I would think that # (1) is correct, as the rescaling is just a change of 
 > units. However, Fabian argues that # (2) is the correct way, but he does 
 > not know why.
 > 

I do not know the correct answer here either and few people really use
the rescaling option.  The question of the rms of a Cleaned image is
quite an issue.  If you Clean very deeply and use no Clean boxes (or
rather use a very large one), then you can reduce the apparent noise
to arbitrarily low levels.  You do this of course by moving noise flux
into apparent, but spurious, objects while also taking real flux from
the real sources and putting that too into spurious objects.  The use
of tight boxes or the options to eliminate weak isolated components
(in the IMAGRPARMs also) helps to avoid this.  I would normally quote
the rms in Jy/beam and then define that beam.  After a rescale from
beam = "dirty beam" to beam = "clean beam" you should quote in the
latter unit or option (1) above.

Eric Greisen