TASS Database, Store ra,dec Sigma or not?

[Chris Albertson <chrisa@wavenet.com>]

Summary:

Some background first.  I am splitting the tass catalog
into two tables, one to contain mean ra, dec and errors.
The other for photometry.  I am working on the first table
now.  We disused this earlier and decided to to keep the
mean ra and dec and errors.  I am wondering now, do we need
sigma?  Isn't sum of squares enough?  Yes I know this
subject seems like it has been beaten to death but I want to
get it right this time.

Longer:

I just finished up writting a new routine to compute means,
and sigma squared for RA and DEC.  Looking at it, it seems
that you can always compute sigma from the sum of squares,
mean and count (see below).  So we could:

1) Not compute it.  And, more importantly not _store_
   sigma squared in the TASS Astrometric Catalog.  If it
   can be directly computed from the other columns we do
   store, the only reason we should precompute and store it
   is if we think we will be _searching_ on the sigma value.

2) Assuming we _do_ want to keep Sigma, Why store Sigma
   Squared?  Seems I could take the square root and store
   Sigma in a 32 bit float.

I see three options:
  1) Store (Sigma)^2 as a 64 bit float
  2) Store Sigma as a 32 bit float
  3) Don't store either #1 or #2.

If we do #3 then searches for sigma > xx will be very slow
so we had better make sure that we don't want to do these kinds
of searches in real time while someone waits on a web based
interface.  I'd say either do #2 or #3.  but I don't know if
searches based on ra_sigma or dec_sigma will be popular.


Just for reference here is how I am computing Sigma Squared:   
  ra_sig2 = (ra_sx2 - (count * ra_mean * ra_mean))
            / (count - 1);

-- 
   --Chris Albertson             home: chrisa@wavenet.com           
     Redondo Beach, California   work: chris@topdog.logicon.com