[daip] [!6528]: aips - OMFIT (31DEC15)

Michael Bietenholz do-not-reply at nrao.edu
Thu Apr 30 21:54:45 EDT 2015


Michael Bietenholz updated #6528
--------------------------------

OMFIT (31DEC15)
---------------

           Ticket ID: 6528
                 URL: https://help.nrao.edu/staff/index.php?/Tickets/Ticket/View/6528



I tested the AIPS version of OMFIT.  The AIPS version to produce much the same fit values as my modified version of Ketan's version (from which some of the present AIPS code is derived) except it produces uncertainties which seem to be sqrt(2) larger.  I made  a simple model data set (UVMOD pt source at origin, unity uncertainties, 2 polns), with FLUX = 0.1, so the uncertainty per visibility per poln. should be 0.1 (Jy).  The uncertainty on the fitted I-polnt flux density should then just be 0.1/sqrt(Nvis*2)

Fitting just a point source ('DOT' in OMFIT), with only the flux density being variable, I found that UVMOD and my version of OMFIT and the 0.1/sqrt(Nvis) all agree, while the AIPS version of OMFIT gets an uncertainty which is sqrt(2) larger.  I conclude that it is OMFIT which is coming up with the wrong value here.

Also, the .HLP file for OMFIT  after this part:
----------------------------------------------------------------
OMFIT
Task:  OMFIT fits a source model to the UV data directly.
       Optionally, this task can simultaneously self-calibrate
    ....

Is out of sync with the first part of the .HLP file, particularly for APARM.  E.g., APARM(1) is first described as  "(1) % change in Chi-squared", and then later as   APARM(1)...Estimated thermal noise per requested Stokes visibility.

While we're at it, the OMFIT output claims:
Post-Fit Variance     =        0.10133614 Jy/Vis

What it gives here is not the variance, which would usually be in Jy**2, but rather the sqaure root of the variance, so the text should probably be changed to maybe "post fit rms" (in Jy), or give the variance, ie. the square of the number presently printed out, which would be the actual variance.


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



More information about the Daip mailing list