dscorr: Computing $\Delta \Sigma$ from weak lensing surveys

WARNING: This package is currently under development and has not been sufficiently tested for scientific applications.

This is a small package that allows the user to compute $\Delta\Sigma$ using a lens sample with individual redshifts (e.g. a spectroscopic or good photo-z sample) and a source sample where only the redshift distribution $n(z)$ is known.

The code uses Treecorr to do the underlying correlation functions and has options for boost factors and random subtraction

The Jackknife covariance produced by dscorr uses Treecorr's internal jackknife patching tools applied to the lens sample. If your lenses and sources do not have the same footprint, I recommend you cut the lens sample to the overlap region before running (if you want to use the jackknife covariance).

Install

To install, pull the repo and run:

cd dscorr/src/
pip install -e .[dev]

Example

We are still working on getting some examples up and running but the basic procedure is the following (assuming you have a lens catalog in a Table table_l and sources in table_s):

#set up correlation function config
corr_config = dscorr.DeltaSigmaConfig(
    min_r = 0.09,             # R bin minimum
    max_r = 3.0,              # R bin minimum
    nbins_r = 10,             # R bin minimum
    bin_slop = 0.01,          # A Treecorr input: Accuracy parameter. 0.01 is usually plenty good enough
    bin_type = "Log",         # A Treecorr input: Should the R bins be log spaced ("Log") or Linear ("Lin")?
    var_method = "jackknife", # A Treecorr input: Should the errorbars use Jackknife resampling or just the shot noise 
    nprocess = 4,             # How many process to you want treecorr to use
    njk = 50,                 # If var_method="jackknife", how many patches shoudl you use (uses treecorr's internal k-means JK patching )
    sigcrit_weights = True,   # Should the lenses each get an additional 1/sigma_crit weight (inproves S/N)
    comoving = False,         # Are distances defined as cosmoving or not
    flip_e1 = False,          # Does your input catalog need e_1 -> -1*e_1
    flip_e2 = False,          # Does your input catalog need e_2 -> -1*e_2
    z_fid = 2.0,              # Fiducial value for the source redshifts before the n(z) correction is applied. Can be any number larger than your lens galaxies
    fbias_nsamples = 1000,    # How many samples of the n(z) should be used when correcting for photo-z spread (larger->slower but less noisy correction)
    use_boost = False,        # Do you want to apply the "boost factor" correction (requires randoms)
    use_random_sub = False,   # Do you want to apply the "random subtraction" correction (requires randoms)
)

#prepare the Delta Sigma calculation
corr = dscorr.DeltaSigma(
        lens_ra = table_l['RA'],
        lens_dec = table_l['Dec'],
        lens_z =  table_l['z'],
        lens_w = None,              #A weight for each lens galaxy (optional)
        source_ra = table_s['RA'],
        source_dec = table_s['Dec'],
        source_e1 = table_s['e1'],
        source_e2 = table_s['e2'],
        source_w =  table_s['w'],
        nz_z=nz_z,                 # An array of redshifts at which to specify your source n(z)
        source_nz=source_nz,       # The source redshift distribution (evaluated at nz_z)
        rand_ra=None,              # If you want to use boost factors or random subtraction specify your random catalog columns here
        rand_dec=None,
        rand_z=None,
        rand_w=None,
        config = corr_config, #pass the configuration of your correlation function here
        )

#run the calculation
r, ds, err = corr.process()

Please see the Treecorr documentation for more details on the treecorr config parameters, and the dsigma documentation for a great primer on the different correction applied to Delta Sigma calculations

credit: Jack Elvin-Poole, Pierre Burger