(the non-official) Sloan-Digital Sky Survey Luminous Red Galaxy Sample

This NON-OFFICIAL cite contains *my* selection of LRGs from their seventh and final data reslease (DR7). As a consistency check I compared my algorithm with Eisenstein et. al 2005 using the DR3 and found excellent agreement both in sample and in the two-point correlation function.
For details regarding the mask, fiber-collision weight, completeness etc., see the bottom. Also many of the details can be found in:

Kazin et al. (2010)

files for Ariel

=============== My Selection of LRGs from SDSS DR7======================
* The details of how the LRGs are selected are on the bottom.

Full DR7 LRG sample- z=(0.16,0.47), Mg=(-23.2,-21.2)

Dim Subsample DR7-Dim- z=(0.16,0.36), Mg=(-23.2,-21.2)
All

data fits

data ascii

Northern Cap only (used in Kazin et al (2009)

Bright Subsample DR7-Bright- z=(0.16,0.44), Mg=(-23.2,-21.8)
All

data fits

data ascii

Northern Cap only (used in Kazin et al (2009)

DR3 z=(0.16,0.47), Mg=(-23.2,-21.8)
All

=========== Using weight for Paircounting ======================
Details of constructions of weights are below, and in Kazin et al.(2009)
Weight options:
(1) sector_completeness
(2) radial_weight
(3) fiber_coll_weight
(4) fgotmain
relevant tag:
(5) n(z)*1e4
When coutning pairs you nominally add the value +1 to the relevant separation bin. To take into account radial and angular selection functions, which you should when dealing with observational data like SDSS here, you need to add +1*(radial weight)*(angular weight)

Steps used in Kazin et al.(2009):

RADIAL
n(z) is the comoving density of LRGs at redshift z.
Consider n(z)*1e4 as unitless
wr0=1/(1+4*n(z)*1e4), (notice the factor 4)
(radial weight)=wr0/mean(wr0)
wr0 is costumely used to take care of radial effects.
The normalization is not important (as 2pcf is normalized with random count), but yields more reasonable couple counts in the end.

ANGULAR
We use only sectors with sector_completeness>0.6
The sector incompleteness is taken care by subsampling the RANDOM points by using the completeness as the probablity of a point to survive the cut.
(angular weight)=fiber_coll_weight
( some multiply the latter also by sector_completeness, but in the case of the SDSS survey we do not expect impact on 2pcf)

======== details of selecting LRGs ===============
Here I explain in detail (IDL) the selections I preformed from lss_index.72.fits which may be obtained from /lss/dr72/.
If you find this section a bit confusing then first read our paper (above), and if that Details regarding .fits file tags I use below may be found in NYU VAGC.
From the vagc-dr7/vagc2 folder you will also need:
object_sdss_tiling.fits
object_sdss_imaging.fits
All three files have the same indeci.

If you have the idlutils package, you only need the PRIMTARGET column, so do:
TILING= HOGG_MRDFITS(TILINGFILE,1, NROW=28800,COLUMNS='PRIMTARGET',/silent)
The first selection I do is:
primtarget=32
itarget=where((tiling.primtarget and primtarget) gt 0 and $
(ulong(tiling.primtarget) AND ulong(2L)^31L) eq 0)

What this means:
primtargets=32 is the bit-mask for LRGs. For MAIN and QSO it will take a few more lines, either read the SDSS documentation (specifically Stoughton et al 2001, Table 27) , or email me and I can send it.
The AND ulong(2L)^31L was added by Blanton. It has something to do with designated LRGs in the 82 Stripe. That stripe had many runs, and redefinitions, so this line is necessary.

More Masks:
ztype eq 1 : only SDSS spectra
z : redshift of interest
icomb ne -1: Not star area
completeness ge 0.6: Completeness of 'sector' in which the LRG resides is more than 60% complete.
Mg : Absolute Magnitude range.

-- Calculating Mg --
The forumla I use is:
Mg=(mr)-(lf_distmod(indat.z,omega0=OmegaM,omegal0=OmegaL))-delta_g+gMINUSr-z_calibration

where: magnitude in r band:
mr=22.5-2.5*ALOG10(CAL.PETROFLUX[2])-CAL.EXTINCTION[2]
where calibration file cal is: object_sdss_imaging.fits (above)
This takes care of extinction.

lf_distmod calculates distance modulus given a redshift.
I assumed a cosmology of OmegaM=0.25 OmegaL=0.75
This fucntion for IDL is in the idlutils package.

delta_g and gMINUSr are Color and K-corrections. I obtain these by interpolating (by redshift) from that published in Table 1 of
Eisenstein et al. (2001) under the Non-Star-Forming column.

z_calibration=0.2 takes care of evolution

-- Calculating Sector Completeness --
The sky is divided into non-overlapping regions called 'sectors', which vary in size, but nominally 0.5 degree, but can be as large as a few degrees. Some objects that are seen in the sky in the imaging procecess, do not obtain spectra. This means the sector is not complete.
A simple way to calculate the sector completeness is to do:
completeness=(# targeted objects with spectra)/(# all targeted objects)
In the paper I define 'targeted object' as the collective of LRGs, MAIN and QSOs. In practice I do:
completeness=(effective # targeted objects with spectra)/(# all targeted objects)
By 'effective' I refer to the fact that I take into consideration correction of number of targeted objects due to fiber-collisions (below). The correction here is for all targeted objects (not only LRGs!).

In our analysis the sector completeness is used for:
(1) LRG selection (above): Only regions with 60% completeness were used.
(2) Random Point distribution: Random points were distributed in a Poisson fashion in the LRG area. To account for incompleteness of sectors, in each sector we diluted the points in a random fashion using the completeness as a probablity function.

-- Dealing with Fiber-Collisions --
Spectra is obtained by spectrographs that are linked by fibers to a holed plate. Due to the fact that fibers have a physical size two objects neighboring within 55'' may not be measured simultaniously. To account for this region were scanned with overlap. Nontheless, some targeted objects (~7% of LRG+MAIN+QSO and ~2% of LRG only) did not obtain spectra in the DR7.
We took this in acount in two aspects of our analysis:
(1) Sector Completeness (above) - here we define 'targeted object' as LRG+MAIN+QSOs
(2) Pair Counting - we assign weights to LRGs that neighbors to LRGs (not MAIN or QSOs!) that did not obtain spectra.
We use the standard method:
We group the targeted objects into groups of 55'' using a friends-of-friends algorithm. We count the number of objects with spectra N of the total T. The weight assigned to each object with spectra is T/N. E.g, in a group of 4 objects and only three get spectra, the weight is 4/3. This weight is used when counting the spectra objects (when pair counting and completeness calculation).

=======================
Other Related Pages:

Eyal Kazin's Large-Scale Structure Page (under construction)

Baryonic Acoustic Feature Page (under construction)

Redshift Distortions Site (still under construction)

Contact Information:
eyalkazin AT gmail dot com