[daip] model-fitting in AIPS

[Andy Biggs]

Hi Eric. A few bugs/problems/suggestions about model-fitting in AIPS:

UVFIT:

This seems to be fundamentally broken in 31DEC12 and 31DEC13. It works 
okay if I do not fit image positions, but as soon as these are included 
in the fit, I get very strange answers - the fluxes are hardly varied 
from their start values and the error bounds are very large. 31DEC11 on 
the other hand seems to work perfectly.

Minor problem: if FITOUT is not set, one gets the following:

localh> UVFIT1: ZTXOP2: using translated file name =
localh> UVFIT1: ZTXOP2:
localh> UVFIT1: ZTXOP2: FORTRAN OPEN ERROR =     10
localh> UVFIT1: ZERROR: ON FILE
localh> UVFIT1: ZERROR: IN ZTXOP2 ERRNO = 10 (No child processes)
localh> UVFIT1: ERROR OPENING TEXT FILE

One thing about UVFIT is that it fits Gaussians and there's no mention 
anywhere about the ability to fit point sources. It seems though that 
setting IMSIZ=0 is effectively a way of forcing a point-source fit. This 
makes perfect sense I suppose, but the documentation insists that this 
parameter *MUST BE GIVEN* which certainly puts me off leaving it blank. 
If IMSIZ=0 is tantamount to a point-source fit then it would be very 
helpful if the documentation would say this. Certainly, IMSIZ=0 gives 
very similar answers to the DOT option in OMFIT.

OMFIT:

I like to use this because it gives me an explicit point-source option. 
It definitely has its idiosyncrasies, but it seems to work quite well. 
There are though a few problems:

* Each component has to be given the same component identifier AND this 
has to be "2". I can't quite believe that no-one else has mentioned 
this, but there is simply no way that a multi-component model will fit 
unless I do this.

* Setting the OUTFIL parameter causes OMFIT to crash with the following 
error:

forrtl: severe (27): too many records in I/O statement, unit -5, file 
Internal Formatted Write

However, this is not a huge problem as OMFIT does actually write the output.

* OMFIT refuses to work on two of my datasets, issuing the following:

localh> OMFIT1: *****  OMSIZE FINDS NO COMPONENTS ******
localh> OMFIT1: Problem with input adverbs or INFILE

I'm fitting to >40 files and all work fine apart from these two. UVFIT 
also works well on them. However, if I split the source I'm fitting to, 
OMFIT suddenly works fine.

If you'd like to investigate these issues, I have put a tar file on my 
ftp site that includes two FITS files, input model files and minimum 
input parameters:

ftp://ftp.eso.org/pub/general/abiggs/modelfit.tar

Cheers,

Andy

-- 

Andy Biggs
ARC Astronomer, ALMA Regional Centre
European Southern Observatory
Karl-Schwarzschild-Straße 2
D-85748 Garching
Germany

tel. +49 89 3200 6471
fax. +49 89 3200 6898