SACC (Save All Correlations and Covariances, an utterly crappy acronym inspired by usually equally bad attempts by David A) is a format and reference library for general storage of 2-dimensional power spectra and correlation functions and their covariance matrices in the HDF5 format. It is very loosely inspired by Joe Zunz's 2point.
Install by saying
./setup.py install
For local installation might need to add --user to that.
You can create a fake datasets by
./examples/create_sacc.py
which you can reload using
./examples/load_sacc.py
and finally run
./examples/split_sacc.py
to load the dataset created by create_sacc.py, split it into three files and reload it again for test.
To describe a generic 2-point correlation function or power spectrum measurements, one needs several ingredients:
- tracer describes a set of tracers in one photometric bin. The tracer description contains the distribution N(z) of tracers and possibly some uncertainty in N(z) in terms of templates to marginalise over. A different photometric bin will be a different tracer, but one can link tracers of the "same kind" (i.e. LSST galaxies) by a common ID root in tracer name. We also allow for external "tracers" such as CMB kappa measurement.
- binning describes a binning of the power spectrum. In short, it is a list of measurements, where each measurement is defined by an ell (or separation in case of configuration space), the pair of tracers it refers to, the actual quantity measured (e.g. shear or numebr density) and the window function. The binning specifies what is measured, but not the actual numbers. We can measure auto and cross power with different binnings.
- mean is the vector of measurements. It has the same number of entries as binning
- precision is the inverse covariance matrix corresponding to mean measurements
The sacc is essentially a container for the tuples of tracers,
binnings, mean vector and precision matrix.
Thus python module should allow reading anbd writing files in sacc
format. If you need to read these files in other context, please see
documentation in rtd and
here.