[daip] [!7296]: AIPS - OMFIT and strange chi2 values

[Michael Bietenholz]

Michael Bietenholz updated #7296
--------------------------------

OMFIT and strange chi2 values
-----------------------------

           Ticket ID: 7296
                 URL: https://help.nrao.edu/staff/index.php?/Tickets/Ticket/View/7296
           Full Name: Michael Bietenholz
               Email: mbieten at yorku.ca
             Creator: User
          Department: AIPS Data Reduction
       Staff (Owner): -- Unassigned --
                Type: Issue
              Status: Open
            Priority: Default
                 SLA: NRAO E2E
      Template Group: Default
             Created: 22 September 2015 08:35 PM
             Updated: 23 September 2015 10:45 PM
                 Due: 25 September 2015 12:50 PM (1d 14h 5m)
      Resolution Due: 30 September 2015 08:35 PM (6d 21h 49m)



So ignore my previous post.  I had forgotten that I was fitting Stokes 'I' which has  a weight of 2.0 if the RR and LL weights are 1.0.

I think I have figured out what is happening:

1) Both the "Post-Fit rms" and NOISE(1) are the values *for weight=1*.  They have to be of course, but the documentation could make that explicit, since if your weights are not near 1.0, then both NOISE(1) and Post-fit rms can be a long way from the actual scatter.

2)  The documentation suggests that NOISE(1) <= 0 defaults to NOISE(1)=1.0.  However, using NOISE(1)=0 produces quite different values for "real chi2" etc than does using NOISE(1)=1.0.   What actually seems to happen with NOISE(1)=0 is that, after the fit, the values are adjusted to NOISE(1)=[post-fit rms] (running with NOISE(1)=0 and NOISE(1)=[post-fit rms] *does* get the same values).

However, this is bad because it means that, with NOISE(1)=0, a crappy model gets the same chi2 as a good model - because the scaling by the post-fit rms makes up for any difference in the chi2!  So I think the post-fit scaling for NOISE(1)=0 should get turned off again.


------------------------------------------------------
Staff CP:  https://help.nrao.edu/staff