3-19-2008
---------

Work on checking deconvln of stars by stars, all imcat.

Put half the 193 stars in the Step2 psfD starfield into one file, and
the other half into the other (every other star goes into the even,
the interleaved ones into the odd).

Conclusions: the constant fit seems to work the best, maybe not surprisingly, here.

The smallest dispersion in the psf-corrected ellip. is seen in the 0th order fit.

Not surprisingly, the best result (smallest bias and dispersion) happens for the best psf and the best stars.

############# Files:

* -- psfmodel.psfd.oddstars.order#.ps -- # = 1,2,3,4,5,8,10, took all
   the odd stars, fit a polynomial psf model to these, and plotted the
   results on a grid (used imcpsf.pro, spawning shell processes with
   all those orders).  the 0th order constant fit seems to work the
   best.

* -- es0.orig.ps, es1.orig.ps -- histos of orig sextractor ellips

-- origsexellip.ps -- es1 vs. es0

* -- e.afterpsfcorr.order#.ps -- # = 1,2,3,4,5,8,10, took all the odd
   stars, fit a polynomial psf model to these, corrected the ellip of
   all the even stars with this psf, then made plots of e1 vs. e0.
   the smallest dispersion is seen in the 0th order fit.


-- es1hist.besteven.ps
   es0hist.besteven.ps
   es0hist.oddstars.all.ps 
   es0hist.oddstars.best.ps,
   es1hist.oddstars.all.ps 
   es1hist.oddstars.best.ps -- histos of star ellip *before* psf correction, for all and best, odd and even.


-- histos of ellip *after* psf correction: all = all stars (around 97
   of them), best = cut on mag < 24 (biggest effect) and fwhm < 3.5,
   for the odd and even stars respectively, cuts each sample down to
   51 or so stars.

e0.alloddbesteven.ps
e0.allonall.ps
e0.bestoddalleven.ps
e0.bestoddbesteven.ps

e1.alloddbesteven.ps
e1.allonall.ps
e1.bestoddalleven.ps
e1.bestoddbesteven.ps


-- Zooming in on the region around zero:

e1.finerhist.alloddbesteven.ps  -- has a clear small positive bias
e1.finerhist.bestoddbesteven.ps -- no bias