Seti *like*

[Tom Droege <tdroege2@earthlink.net>]

You all jumped on the wrong problem.  I said seti *like*.  The problem is 
to get the data analyzed in a uniform way.  The mail works just fine for 
sending out data.  I have sent out several hundred disks already and they 
were quickly reduced to almost nothing.  This as determined by measuring 
the feedback.  ;^)

I doubt that many of you want to feed a years run of 1500 CDs through a 
program.  But 10-20 of you might do 10 a month.  This gets us star 
lists.  These can be sent to the next processing point by mail.  I also 
assume that we will redo this several times before we get it right.  Each 
of the group might store a few hundred CDs.  When a new analysis version 
was sent out then they could be run through again.  That is just life.  We 
will learn how to do it right.

The next problem is to examine the star lists and to get the results sorted 
into "real" objects.  There is just too much data for anyone to look at the 
final star data.

In my own dumb way I have been thinking about the problem that Chris 
mentioned.  What to do about the "bad" data - the single event 
data.   There are lots of cosmic rays, meteors, satellite trails, and the 
like.

I would like not to require V/I pairs.  This is because I have been leaned 
on by someone who does not want me to throw away single 
observations.  Someone I respect.  Someone else that I respect wants to 
keep only the V/I pairs.  That is the dilemma of listening to experts.

OK, how about this.  We will have hundreds of observations of each 
star.  Suppose we reduce them to RA, Dec. position pairs.  Now if we plot 
these in RA, Dec space in a frequency plot, (the value at each RA, Dec. 
point is the number of times it was hit) we will have something that looks 
like a sky image.  Points hit most often will be "brighter".  They will 
have some RA, Dec. distribution that generates one PSF for real stars and 
something else for galaxies and other stuff.  Now we can search for "stars" 
using aperture photometry or some such.  (This assumes we have a way to 
keep track of the real data associated with each RA, Dec. point plotted.) 
We now collect all the hits in some aperture and call this a "star".  We 
can tell by the shape of the distribution what kind of an object it is.  We 
can place some requirements on the distribution to avoid blends of two 
stars.  i.e. we want a clean single peak.  The  "bad" data should now be 
just a uniform background noise.  I see no reason why it should not be 
uniform.  But clear thinking will tell.

OK, Chris will tell me that he can do this in a data base, and that is fine.

Tom Droege