Re: Data reduction methodology for V-I colors

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

On Wed, 11 Aug 2004 23:07:34 -0000, "Richard Miles"
<rmiles.btee@btinternet.com> wrote:

>Tom et al.,
>
>Am following TASS developments with interest.  It's a fascinating project.
> ...
>However, it occurred to me that TASS may not be making optimal use of some
>of the data, in particular those pertaining to V-I color index.

True.

I spent all last winter working on a form of ensemble 
correction using all nearby observations. both V and I
simultaneaously, that should take out some of the scatter 
in the image calibrations.

I just checked back to see what I wrote up and the
answer is "nothing." I have an early draft write-up
that doesn't match what I was actually doing at the 
point when I had to start pruning my vines. After the
draft write-up and before I quit, I was giving myself
a crash course in Bayesian statistics and had proved
that the results I got depended only on my a-priori
assumptions ... pretty much independent of the data.

Assume that some images are worse than others but no
stars are variable and you can measure how bad some of
the images are. Assume that some stars are variable as
well as some images being bad and ask how many stars are
variable and you get answers like "30% of stars vary
significantly on the scale detectable by TASS" which is
just not true. Look into it a bit more deeply and the
program replies " ... or your image errors are not
Gaussian" which one pretty well knew to start with.

So the short answer is that I failed to come up with
an ensemble correction that did one damned bit of
good. That doesn't mean that you couldn't do it!

Andrew Bennett, Avondale Vineyard