[daip] aips_gripe

[Leonia Kogan]

Hi Lincoln,

>2) you iteratively discard points > 5*mean.  Do you then compute an RMS of
>what remains?

YES

>3) do you then compare each point in the first cube to the RMS that is
>relavant for the plane it is in, and choose to blank it or not?

I use the calculated RMS(I)*COEF as a threshold to exclude all points 
at the plane 'I' which are more than RMS(I)*COEF.
Actually I substitude the global threshold FLUX by RMS(I)*COEF.

>4) is the value of FLUX and CUT now in sigmas, rather than map units
>(e.g., Jy)?

The value of RMS(I) is in map units. The user can multiply the calculated 
RMS(I) by the undimensional COEF (COEF=FLUX at this case)

>I think that the best way to compute an RMS is to use the technique
>applied in JMFIT - to construct a histogram of the source and estimate
>the width of the counts' number vs flux profile centered around zero.

You mean IMEAN (not JMFIT). In the case of strong sources at the plane 
you have to find solution for thr source (fit gausian(s)) and then 
to calculate rms either by histogram or by direct calculation.

Otherwise you have to specify the box at the plane which you consider 
is free of a source.

I want to get rid of specifiing the off source area for each plane.
It can be annoying for the user.


Now let's brash aside an algorithm of RMS finding for each plane.

Suppose RMS(I) is evaluated correctely. 
Then is the cutoff described at the item 3 is what you want?

If it is I'l offer you another algorithm of RMS(I) calculating without drop off 
of the negative points.


Leonia

----- Begin Included Message -----

>From lincoln at play.harvard.edu Tue Nov 13 16:28 MST 2001
Date: Tue, 13 Nov 2001 18:28:08 -0500 (EST)
From: Lincoln Greenhill <lincoln at play.harvard.edu>
To: Leonia Kogan <lkogan at zia.aoc.NRAO.EDU>
Subject: Re: [daip] aips_gripe
In-Reply-To: <200111132223.PAA13572 at bonito.aoc.nrao.edu>
MIME-Version: 1.0
X-Lines: 66
Status: RO

Hi Leonia,

I was just getting to that letter (about XMOM).

First - no need to apologize about not understanding what I meant
about COMB.  Afterall, I sent my gripe in three letters, the second
of which said that I was wrong to write in the first place.  I am
not surprised there was confusion...and it was MY fault.

I do not think that the modifications you discuss re XMOM address the
problem that we face - but I may not fully understand the changes you
made.

Questions:

1) you read the second image, one plane at a time, and discard the
negative points.  Why?  That is not a proper technique for RMS
calculations...

2) you iteratively discard points > 5*mean.  Do you then compute an RMS of
what remains?

3) do you then compare each point in the first cube to the RMS that is
relavant for the plane it is in, and choose to blank it or not?

4) is the value of FLUX and CUT now in sigmas, rather than map units
(e.g., Jy)?

Suggestions:

I think that the best way to compute an RMS is to use the technique
applied in JMFIT - to construct a histogram of the source and estimate
the width of the counts' number vs flux profile centered around zero.

I'll presume that I am right that your modified code computes an RMS for
each plane, and drops pixels in each plane, based on the respective RMS.
This fix does not fully address the problem we have that there are
artifact-like patterns in the background of some of our planes (e.g.
ripples that look like a narrow ladder extending north and south of some
of the brightest emission).  We wish to apply a more stringent statistical
test to the blanking of pixels in the regions of planes that have this
"ladder".  One way to do this is to set a single theshold w/ FLUX and CUT
(e.g., 10 sigma, compare the flux of each pixel to a local RMS for the
part of the plane in which it lies, and drop pixels that are less than the
local RMS.  The user could set the size of the local region in which the
RMS is computed.


Regards,

Lincoln







Lincoln J. Greenhill      Harvard-Smithsonian Center for Astrophysics
Radio & Geoastronomy Division, 60 Garden St, Cambridge, MA 02138, USA
Telephone:  1 617-495-7194             FAX:  1 617-495-7345
Internet:  greenhill at cfa.harvard.edu   http://cfa-www.harvard.edu/~lincoln






----- End Included Message -----