Wednesday, March 3, 2010

Improving Luminosity Function

Had a long chat with Joe Hennawi today. He basically went through his Quasar Monte Carlo (that generates the QSO catalog for the likelihood) and showed me where to make improvements.

He agrees that what we are doing with variability and putting the QSOs in the denominator of the likelihood calculation is correct. He thinks we have problems we need to first address are:

1. Because we remove all variable objects from the everything catalog, we basically don't have low redshift quasars anywhere in our code, and so we need to change the QSO catalog to include these (but only put them in the denominator, and keep the numerator to only be QSOs in our redshift range).

2. The luminosity function isn't working well enough for low redshifts to do (1) so I need to fix this by using Richard '06 function instead -- this is coded up already, I just need to implement it.

Joe also described to me his work at NYU. I need to get in touch with Jo Bovy and get a description of his "extreme deconvolution" method for a "future possibilities" slide for my Utah talk.

Here are (very rough) notes from my talk with Joe today, basically for my own use:

get luminosity function working on joe's monte carlo
change likelihood so that numerator is z, 2.15 - 4 and denominator, is all quasars of all redshifts)
hopkins/richards paper

make same plot as on blog with coadded photometery

send joe the coadded stipe-82 and median seeing.

jo bovi

file for monte-carlo -


dndz file is generated (need to fix it with different luminosity function)
randomly select objects from dndz distribution.
qso_mag_rand - given limitis generate random magnitudes
qso_fakephoto -matches the redshifts and imagnitude
if brighter than some magnitude limit, then uses real data, if less than

luminosity function code
computes the number counts of quasars in dn/dz so integrated over dz would get number count between

don't change the cosmological parameters.

needs to be modified
try to implement the Richards '06



n(i_faint,i_bright, z_min,z_max) = int_z_min^z_max dz qso_dndz(i_faint,i_bright,z)

generates the distribution
change so it goes from 0.5 to 5.0

Files I'm using:
[3/3/10 1:13:26 PM] Joseph F. Hennawi: /home/jhennawi/Projects/QSOClustering/pro/k_correct/
[3/3/10 1:25:51 PM] Joseph F. Hennawi: n(i_faint,i_bright, z_min,z_max) = int_z_min^z_max dz qso_dndz(i_faint,i_bright,z)
[3/3/10 1:28:45 PM] Joseph F. Hennawi: /home/jhennawi/Projects/QSOClustering/pro/k_correct/
[3/3/10 1:30:22 PM] Joseph F. Hennawi: /home/jhennawi/Projects/QSOClustering/pro/k_correct/
[3/3/10 1:32:11 PM] Joseph F. Hennawi: /home/jhennawi/Projects/QSOClustering/pro/k_correct/shen_hopkins
[3/3/10 1:41:52 PM] Joseph F. Hennawi: /home/jhennawi/Projects/QSOClustering/pro/data/sdss/

No comments:

Post a Comment