News and Updates‎ > ‎

catalog reconciliation

posted Feb 4, 2013, 7:08 AM by Paul M. Sutter   [ updated Feb 4, 2013, 7:13 AM ]
There have been a few questions raised recently about apparent inconsistencies among the three (July, August, and November) releases. We would like to take this opportunity to explain the differences in more detail. 

If you are not already doing so, please use the November release!

The catalog inconsistencies can be explained by:

1) A lower radius threshold for including voids in the November release. Previously, it was two times the mean galaxy separation. We now include voids all the way down to the mean galaxy separation. Also note that we miscalculated the mean galaxy separation for the dim1 sample; this has been corrected. Use the smallest voids at your own risk; you may wish to impose an arbitrary cutoff (e.g., 2*r_min or 4*r_min) for your analysis.

2) Different handling of the high-redshift cap of each subsample. In July and August, we were placing mock particles along the cap and treating voids there just like voids at the mask boundaries (i.e., these would appear as "edge" voids). However, this procedure was giving us difficulty at higher redshift, so we switched to a simpler approach: we no longer place mock particles at the redshift cap and remove all voids that might intersect the upper redshift boundary of the subsample.
 
3) A bug fix in the central density calculation. Previously, all voids were using a fixed radius to determine the central density. This is not what we intended. We fixed this to calculate central densities using 1/4 of the effective radius, as we state in the paper. So some voids that were excluded are now included and vice-versa.

4) The August release (but not the July release) contained a bug in the barycenter calculation. Centers were still inside the voids but slightly offset from the true barycenter. This has been fixed for November. 

5) There was a bug in the catalog release pipeline script which switched the "central" voids with the "edge" voids. This has been corrected for November: "central" voids are now actually central voids!

Changes 1) and 2) generally explain the different void populations in the "all" sample. The redshift cap handling especially affected the lrgbright sample. If you were using the "all" sample, then change 5) didn't affect you. However, if you are using the "central" sample, then you will see a completely different catalog in November.

Thank you for your understanding!
Comments