Independent analysis of TASS MKIV data

[Andrew Bennett <andrew.bennett@ns.sympatico.ca>]

I am at long, long last reasonably confident of
my PSF-fitting analysis. Here is a preliminary
comparison with the TASS pipeline data in the
Database.

I have processed data files HV/IRA2604797 and
extracted the matching data from the Database
(JD 245605.79772). I have so far done a detailed
comparison of just 4 square degrees: RA78 to 79
degrees times the image width in declination. The
comparison is somewhat tedious so I probably
won't do the rest ...

**** Executive Summary Conclusions:
There are no serious discrepancies provided
one accepts the higher error estimates that I
recently posted rather than the very optimistic
estimates included in the Database.

**** Details:

In this area there are 384 TASS stars.
My PSF program resulted in
259 V-band which I will call ABV
889 I-band: ABI.
Obviously, my V analysis did not go as deep
as the TASS pipeline. I am too lazy to rerun it
going deeper.

Of the 384 TASS stars
  204 appear in both ABV and ABI
  164 in ABI
   16 in neither ABV nor ABI
This is pretty typical of survey comparisons
and nothing at all to worry about.

Of the remaining 55 ABV stars
  27 also appear in ABI
  28 don't
Since the V and I areas do not completely
overlap, one expects some no-shows. However,
most of the ABV orphans are not in the non-
overlap area and some are quite bright.
Presumably, some are very blue and so off
the bottom of the TASS pipeline I-band
analysis. Presumably some have been rejected
by the pipeline as being non-stars for one reason
or another. Some may be real stars found by the
superior (I hope) multiple star capabilities of my
new improved PSF code. I haven't yet looked
at the image. Yes - I should have.

Of the remaining 494 ABI stars, only a tiny fraction
are in the non-overlap area. Lots of them are quite
bright. These are presumably red and prime
candidates as variable stars. It is a pity these are
not getting into the Database!

Detailed comparison of the 204 common stars give
  dMV = 0.0496 mags (p.e.)
A large part of this scatter comes from variation of
the PSF across the image and is entirely to be
expected in a comparison of this kind. Fitting 2nd
order spatial terms gives
  dMVCorr =0.0189 mags (p.e.)
The scatter increases with magnitude and is quite
similar to my estimate of the measurement errors.
The scatter of the difference between the two 
methods of analysis is actually larger than the
official estimate of measurement error included
in the Database. I was initially dismayed at this
result but further analysis shows that the expected
scatter between the two methods is in fact of the
same order as the scatter in either. 

*** If anybody wants to see a proof of this  at first
sight startling result, I will write it up and post it! ***

If you accept my larger estimate of this scatter, 
there is not a serious problem!

  dMI = 0.0484 mags (p.e.)
Spatial variation of PSF is less in I-band so the
improvement on subtracting spatial terms is not
as great as with V-band. The remaining scatter
is larger than one might have hoped but not
alarmingly so.

  dRA, dDec 0.540 to 0.713 arc seconds (p.e.)
Probably of no consequence but far, far larger
than I expected. My ABV and ABI determinations
differ by only 0.317, 0.384 arc seconds (p.e.) and that is
using independent images, which one would expect
to give at least root two greater error. Dare I suggest 
that my least squares fitting is giving better positions?

Andrew Bennett, Avondale Vineyard, NS, Canada