Re: Refcat Results

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

On Mon, 27 Jan 2003 14:43:06 -0600, Tom wrote:

> ...
>One notes that the V data is pretty symmetrical around zero as one would 
>hope.  The I data is not symmetrical, and is biased toward tass stars being 
>brighter at low magnitudes and fainter at high magnitudes.  (Readers are 
>warned that I always get signs wrong on such things.)

Just so!

The apparent bias at high mag is a simple selection 
effect and nothing to worry about. Stars seen by 
TASS as being brighter get included
while those that come out fainter do not.

The bias for bright I stars is, on the other hand,
something to worry about.

>
>OK, it is not yet clear to me that this is a big source of error.  While 
>the one star is 2 mags off the mean, there are 800 stars typically in the 
>solution, so I suppose we divide 2.000 by 800 to get a 0.0025 mag effect 
>depending on whether this one star is in the solution.  But there are a lot 
>of pretty large differences.  Comments are welcome.
>
>Tom Droege  

I use a median/probable error process rather
than mean/standard deviation; this automatically
loses these outliers. Taking my results and doing
some hand-waving over them - such as pretending that
the errors are Gaussian - I estimate the formal
uncertainty in the photometric calibration for DS24 as
0.004 mags rms in V (600 sources matched) and
0.010 mags rms in I (800 sources matched).

That's using all the Tycho2 sources. Yes - I know
I should filter them. I have not got round to it.
Filtering out the faints, reds and peculiars should
reduce this, but not by much because the numbers
of sources matched goes down too.

For comparison, after correcting for the average
spatial errors, I end up (after considerable manual
processing ... take results with a large pinch of salt)
with a residual image to image variation of
0.011 mags rms in V
0.008 mags rms in I
Since this error includes remaining uncorrected 
spatial errors and other sources of error, it should 
be an upper limit on the error introduced by the catalog.

This is also an upper limit on the improvement that
might be achieved if the referance catalog were
perfect. Something to consider before putting a
major effort into revision. For comparison, spatial
variations across one image run to 0.04 mags rms
and up and I am having problems correcting
them down to even 20% of this ... an uncertainty
similar to the total error contribution of
the existing reference catalog.

Andrew Bennett, Avondale Vineyard