quickcat coefficients need to be updated

[moustakas]

The coefficients in desisim/bin/quickcat haven't been updated in more than two years, since the redmonster + zdc1 era. We desperately need to do this update so that quicksurvey can use our latest and greatest simulation inputs / outputs (see also https://github.com/desihub/desitarget/pull/307).

[moustakas]

@londumas Do you by chance have any of these coefficients handy (I can update the code)? (Note: we also need the redshift success rate as a function of observing conditions and apparent magnitude.)

#- redshift errors and zwarn fractions from DESI-1657
#- sigmav = c sigmaz / (1+z)
_sigma_v = {
    'ELG': 19.,
    'LRG': 40.,
    'BGS': 13.,
    'QSO': 423.,
    'STAR': 18.,
    'SKY': 9999,      #- meaningless
    'UNKNOWN': 9999,  #- meaningless
}

_zwarn_fraction = {
    'ELG': 0.14,       # 1 - 4303/5000
    'LRG': 0.015,      # 1 - 4921/5000
    'QSO': 0.18,       # 1 - 4094/5000
    'BGS': 0.01,
    'STAR': 0.05,
    'SKY': 1.0,
    'UNKNOWN': 1.0,
}

_cata_fail_fraction = {
   # Catastrophic error fractions from redmonster on oak (ELG, LRG, QSO)
    'ELG': 0.08,
    'LRG': 0.013,
    'QSO': 0.20,
    'BGS': 0.,
    'STAR': 0.,
    'SKY': 0.,
    'UNKNOWN': 0.,
}

[londumas]

Would you like these coefficients from desisim? We could take the results of desitest. Or you would like the results from the SDSS data?

[julienguy]

We would need to have the coefficients calibrated on desi simulations, and ideally based on CCD images simulations. We have such CCD image simulations and the redshift catalog, matched to the truth. This only possible limitation is the statistics. I will put a link to this data set today. It would be great if one of you can have a look at it.

[londumas]

@julienguy, I volunteer as tribute.

[moustakas]

For reference, the current set of quickcat coefficients are based on this study-- Redmonster run on ZDC1 sample

[julienguy]

Helion and others, you can have a look at the following redshift catalog /project/projectdirs/desi/datachallenge/redwood/spectro/redux/redwood/zcatalog-redwood-target-truth.fits It has been run with version 0.13.0 of redrock (and software release18.6 for the rest of the software) See email of Stephen on data system list for more details on the redwood prod. In brief, this is the most realistic set of sims we have for DESI.

[londumas]

@julienguy, Thanks, I am computing the different statistic now.

[londumas]

@moustakas, Here are the results and the code to get them:

• I computed sigma_v as the std of Delta_v, only for objects with ZWARN==0 and Delta_v<1000 km/s
• ZWARN fraction is the ratio: ( ZWARN!=0 ) / all object
• cata_fail_rate is the ration: ( ZWARN==0 & |Delta_v|>1000 ) / all object

Is that what you want or you need more?

import scipy as sp
import fitsio
import matplotlib.pyplot as plt
from redrock.fitz import get_dv

def main():
    path = '/project/projectdirs/desi/datachallenge/redwood/spectro/redux/redwood/'
    h = fitsio.FITS(path+'/zcatalog-redwood-target-truth.fits')
    dv = get_dv(h['ZCATALOG']['Z'][:],h['ZCATALOG']['TRUEZ'][:])

    ###
    for spectype in sp.unique(h['ZCATALOG']['TEMPLATETYPE'][:]):
        print('\n')
        w = h['ZCATALOG']['TEMPLATETYPE'][:]==spectype
        ww = w.copy() & (h['ZCATALOG']['ZWARN'][:]!=0)
        print(spectype, '_zwarn_fraction: ', 1.*ww.sum()/w.sum() )
        ww = w.copy() & (h['ZCATALOG']['ZWARN'][:]==0) & (sp.absolute(dv)>1000.)
        print(spectype, '_cata_fail_fraction: ', 1.*ww.sum()/w.sum() )
        ww = w.copy() & (h['ZCATALOG']['ZWARN'][:]==0) & (sp.absolute(dv)<=1000.)
        print(spectype, '_sigma_v: ', sp.std(dv[ww]) )

    return
main()

b'0.0       ' _zwarn_fraction:  0.901349659107
b'0.0       ' _cata_fail_fraction:  0.09684151941
b'0.0       ' _sigma_v:  516.932772619

b'BGS       ' _zwarn_fraction:  0.0240823783425
b'BGS       ' _cata_fail_fraction:  0.00332170735758
b'BGS       ' _sigma_v:  37.7034302508

b'ELG       ' _zwarn_fraction:  0.086735213275
b'ELG       ' _cata_fail_fraction:  0.0198853032
b'ELG       ' _sigma_v:  38.0252072725

b'LRG       ' _zwarn_fraction:  0.0074137035463
b'LRG       ' _cata_fail_fraction:  0.00184162062615
b'LRG       ' _sigma_v:  67.3787383731

b'QSO       ' _zwarn_fraction:  0.0203632361035
b'QSO       ' _cata_fail_fraction:  0.011557512383
b'QSO       ' _sigma_v:  182.164628444

b'STAR      ' _zwarn_fraction:  0.344952380952
b'STAR      ' _cata_fail_fraction:  0.0500952380952
b'STAR      ' _sigma_v:  51.5065606385

b'WD        ' _zwarn_fraction:  0.0939597315436
b'WD        ' _cata_fail_fraction:  0.0
b'WD        ' _sigma_v:  54.3481663135

[julienguy]

thanks. it would be nice to also recalibrate the rest of the redshift efficiency calculation which is a function of magnitudes and [OII] fluxes. see https://github.com/desihub/desisim/blob/2362b20825c73ee78b324a61cb5b21d86bca8346/py/desisim/quickcat.py#L133

for instance, does quickcat reproduces the following efficiency curve derived from the same redwood set of simulations:

[londumas]

@julienguy, I don't think I will have the time to do that part.

[londumas]

@moustakas, Should I update the values in https://github.com/desihub/desisim/blob/master/py/desisim/quickcat.py? from line 40 to 72? Idon't have results for UNKNOWN and SKY. And for stars, it is very bad for MW targets, but normally it is good for Standard stars.

[julienguy]

I am working on the quickcat calibration in the branch quickcat-calib. All is done in a notebook: https://github.com/desihub/desisim/blob/quickcat-calib/doc/nb/quickcat-calibration.ipynb

[michaelJwilson]

Thanks Julien!

M

On Thu, 27 Sep 2018 at 17:13, julienguy notifications@github.com wrote:

I am working on the quickcat calibration in the branch quickcat-calib. All is done in a notebook:

https://github.com/desihub/desisim/blob/quickcat-calib/doc/nb/quickcat-calibration.ipynb

— You are receiving this because you are subscribed to this thread. Reply to this email directly, view it on GitHub https://github.com/desihub/desisim/issues/355#issuecomment-425280960, or mute the thread https://github.com/notifications/unsubscribe-auth/AFwSkl1YUxNkCQts5X24tMZcxs8VabFnks5ufWmagaJpZM4TlEM5 .

-- Dr. Michael Wilson

Cosmology Postdoctoral Fellow

http://www.roe.ac.uk/~mjw/

Lawrence Berkeley National Lab 1 Cyclotron Road, M/S 50A5104 Berkeley, CA 94720, USA

Tel: (+01) 510-502-3448 (+44) 7742 373026

[moustakas]

Done in #430.