Re: Working on Data

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

On 09 Jul 2003 14:13:51 -0400, Tom Droege <tdroege2@earthlink.net>
wrote:

>So far I have concluded:
>
>1) Some of the data is clearly bad.
>2) One can identify and remove "bad" data.
>3) The result is only a slight improvement when measured by a sigma vs
>mag plot.

This contradicts my conclusions in TN94, where I got
quite large improvements if I threw away enough data -
more than half the images.

> ...
>I still think that work is needed on the reference catalog and that will
>provide the greatest improvement for the data.

The methods I used in TN94 use the reference
catalog only to select the ensemble used for
calibration: even a "perfect" catalog would have
only a tiny effect by adding and removing weaker
sources from the ensemble - consistent used of
weighted fits means that these weak sources have
a very small effect since the weighting depends on
what is observed rather than on the (wrong) catalog
value.
>
>Tom Droege

On Wed, 9 Jul 2003 18:24:36 -0400 Stupendous Man
<richmond@a188-l009.rit.edu> wrote

>  Note the clear systematic pattern in V-band residuals, and lack
>thereof in I-band residuals.

In TN94, I was already removing these systematic
gradients. In that case, I-band was worse than V-band.
Yes - this depends on the particular flat-fielding
that was used. This appeared not to have any effect in 
the final, corrected results in the previous data which
I analysed from scratch with different flat-fielding.

Andrew Bennett, Avondale Vineyard