[daip] OMFIT ([#XAQ-182450]: Text installation of AIPS on Ubuntu 11.04...)

[Eric Greisen]

Michael Bietenholz wrote:
> Hi Eric
> 
> I ran a very cursory test of the new OMFIT.
> It gets the same results as my OMFT9 version except for:
> 
> 1) Nat. Wgtd. Map Noise mfb OMFF9:  Nat. Wgtd. Map Noise  =      
> 1.67769938 mJy/Beam
> aips OMFIT: Nat. Wgtd. Map Noise  =      3.67815219 mJy/Beam
>  (ratio = 2.206)
> 
> 2) All of OMFIT's uncertainties are higher than my version by sqrt(2).
> I'm not vastly confident of my values, so you may be right.  One
> should probably just do a wee Monte-Carlo test to figure out which is
> correct.
> 
> OMFIT does seem to get the same Chi2 and "Real Chi2", but with my data
> set the output here might benefit from using G format. (The problem is
> that the AIPS data weights are often a long way from being 1/sigma^2,
> so without any scaling, you wind up with reduced Chi2 values not very
> near 1.0.
> 
>  OMFIT1:  Real Chi-Squared    =                 0.08495237
>  OMFIT1:  Real Reduced Chi^2  =                 0.00000149
> 
> I would also just strip out the messages giving the internal Chi2 and
> Reduced Chi2, I think those will just sow confusion.  Or at least
> label them e.g., "working Chi2" and just "Chi2".
> 
>                      michael
> 

I am going to make the following changes in tonight's MNJ:

1. If NOISE(1) not given, the "Real" Chi2 will use the measured variance 
rather than the default NOISE(1) (which is -1).  This will make the 
value printed less dependent on the user's NOISE(1) adverb.

2. I will restore G format for all the Chi2 FORMATs.

3. Ketan's Chi2 display will be labeled as such and appear only if 
PRTLEV>3.  The Real Chi2 will appear at PRTLEV.1.

However the naturally weighted Map noise is now correct and was wrong in 
IMFT9.  Ketan fixed this one.  Yours uses CHI2 and sqrt(chi2) is needed.

I think I will fix the formatted write to text file - AIPS standards 
require use of ZTXIO...

Thanks for all the help.

Eric Greisen