Measuring the correlations

First we set up the configuration for the correlation measurements. Here we want to measure correlations corresponding to a transverse angular diameter distances between 0.5 and 1.5 kpc. Additionally, we instruct the code to compute the correlation functions in 22 linearly spaced bins of redshift in a range of 0.1 to 1.2:

config = yaw.Configuration.create(
    rmin=500.0,  # can also be a list of lower scale limits
    rmax=1500.0, # can also be a list of upper scale limits
    # unit="kpc"  # defaults to angular diameter distance, but angles and
                  # comoving transverse distance are supported
    # rweight=None,     # if you want to weight pairs by scales
    # resolution=None,  # resolution of weights in no. of log-scale bins
    zmin=0.1,
    zmax=1.2,
    num_bins=22,
    # method="linear",
    # edges=None,  # provide your custom bin edges
)

Next we measure the autocorrelation amplitude, which is a measure for the galaxy bias of the reference sample. Afterwards we measure the cross-correlation amplitude, which is the biased measure of the unknown redshift distribution. Typically, this is the most expensive operation in the workflow:

cts_ss_list = yaw.autocorrelate(
    config,
    cat_reference,
    cat_ref_rand,
    progress=True,  # shows a progress bar, default: False
)

cts_sp_list = yaw.crosscorrelate(
    config,
    cat_reference,
    cat_unknown,
    ref_rand=cat_ref_rand,
    unk_rand=cat_unk_rand,
    progress=True,  # shows a progress bar, default: False
)

The measurement functions above always return a list of correlation pair counts. Since we configured a single measurement scale, the lists contain just a single item, which are the pair counts that we are interested in. We can save them to a HDF5 file for later inspection or avoiding to recompute the pair counts every time.

cts_ss = cts_ss_list[0]
cts_ss.to_file("w_ss.hdf5")

cts_sp = cts_sp_list[0]
cts_sp.to_file("w_sp.hdf5")
# restored = yaw.CorrFunc.from_file("w_sp.hdf5")

Correlation pair counts are stored as yaw.CorrFunc objects and are very flexible. They can be sampled to an actual correlation function using a correlation estimator,

w_ss = cts_ss.sample()  # creates a CorrFunc object
w_ss.plot()  # automatic plot

or inspected (e.g. by indexing along the redshift bin or patch axis) to investigate individual pair counts:

cts_sp.patches[3:6]  # subset with all pair counts involving patches 4 to 6
cts_sp.bins[:5]  # subset with all pair counts of the first 5 redshift bins

dd = cts_sp.dd  # access stored reference-unknown pair counts
dd.get_array()  # array with shape (num_bins, num_patches, num_patches)