Agenda And Minutes for 2/3 10am CST

Call-in Info

Call-in instructions:

1. Dial Toll-Free Number: 866-740-1260 (U.S. & Canada)
2. International participants dial:
Toll Number: 303-248-0285
Or International Toll-Free Number:
3. Enter your 7-digit access code, followed by “#”

Your access code: 7691719


  1. Photo-z Updates
  2. Psf non-homogenized coadds
    1. Tile dependence of star-galaxy classification
    2. Comparing to Homogenized Coadds
    3. Comparing Photometry to Homogenized Coadds
  3. Star/Galaxy Round 2
  4. Pixelization of masks
  5. Catalog Validation Tests redux...


Attending: Huan, Felipe, Scott D., Aurelien, William W., Anne B., Diego, Douglas T., Alex D-W, Nacho, +

First, we noted that SVA1 Gold 1.0 was available at the Barcelona database...I need to get the exact link and send it around. Gold will also be distributed via DESDM...

Photo-zs: Huan gave a summary of the discussions at the photo-z telecon. The training will be done with Gold photometry, and errors will be estimated based on the validation set. Huan is still finalizing the training since there are a few missing galaxies to track down. But this run should be done soon.

There is still the worrying note about the COSMOS magnitude offsets (though not color offsets) between the different SLR runs. (Probably related to the saturation of all the 2mass stars in the area.) This needs to be investigated further.

Star/Galaxy: There was discussion on the s/g challenge about the star/galaxy ratio in the training sets, and how we should account for the fact that cosmos doesn't represent everywhere in the survey. Nacho has sent out an e-mail with more details.

The validation sets will now be blinded, but there will be a separated "test" set for playing around with.

Finally, we discussed pixelization of the masks and how to scale this to "DES-scale". (We should also have discussed plans for SVA1...) Eli has added the fine pixel masks to the Pixelization of masks page, as well as new residual plots.

While Eduardo had noted that nside=8192 has sufficient resolution to resolve 100 kpc at z=1 (which is fine enough for most everything), there was worry that this would still leave residuals on a larger scale. It actually may not be so bad, since the correlations are on very narrow lines.

However, the real way to test this would be to generate fake galaxies with 0 correlation and measuring the correlation function using random points generated from various pixelization levels. Eduardo has sent out a note to this effect.