O3b

The aim of the MR is for the changes needed between O3a and O3b to be collected in one space.

Will bump version to 0.3.0.

• gwpopulation_pipe/common_format.py
  • new minimum/mass arguments for smoothed mass distributions
  • new vt-ifar-threshold argument to allow differing thresholds to be used
  • format of the file has changed
    • this method was producing data frames that had too many columns
    • the new format splits things up into logical groups
    • @colm.talbot will document the new format
    • new format has presumably been accounted for in plotting (@shanika.galaudage)
• gwpopulation_pipe/data_analysis.py
  • some new imports
  • SmoothedMassDistribution -> SinglePeakSmoothedMassDistribution in GWPopulation
  • parser arguments moved to main parser
  • vt_ifar_threshold passed into vt_func
  • more specific instantiation to allow minimum/maximum mass to be passed
  • sampler kwargs parsing taken from bilby_pipe (reviewed)
  • assigning arbitrary attributes to Bilby objects was losing information, save in meta_data instead
  • names of hyperparameters constructed before sampling and used to downselect from single large prior
    • previously we had a prior file for each combination of models
    • now use a single big prior file to avoid duplicate prior definition
• gwpopulation_pipe/data_collection.py
  • lalinference_prior -> euclidean_distance_prior for inclusiveness
  • evaluate_prior has more complex logic based on passed arguments
    • currently handles only two redshift priors, the usual mass prior, and default (i.e., component) spins
  • load_posterior_from_meta_file allows preferred labels to be passed to access random run names
  • load_meta_file_from_json should not be used as all posteriors are passed as hdf5
  • load_samples_from_csv should not be used as all posteriors are passed as hdf5
  • new function load_o3b_events as a copy of load_o3a_events and modified parsing
    • this may need to change if all o3a and o3b files will be in the same directory
• gwpopulation_pipe/main.py
  • new parsing of arguments, kind of messy, probably easiest to just verify what the output looks like
• gwpopulation_pipe/parser.py
  • many new parameters based on new functionality
  • vt-parameters may be relevant, it says which parameters to include in sensitivity estimation
    • should be mass, redshift, spin for the default as we have full spin injections
• gwpopulation_pipe/post_plots.py
  • refactoring but no substantive changes
  • files are still written with deepdish. This should be removed, but needs me to understand h5py.
• gwpopulation_pipe/utils.py
  • KNOWN_ARGUMENTS is a hack to deal with the default spin model being defined in terms of beta distribution parameters but sampling in mean and variance for the prior parsing
• gwpopulation_pipe/vt_helper.py
  • a bunch of code migrated to GWPopulation
  • read injection files using h5py
  • will only work if mass, spin, and redshift are all being fitted
  • check jacobian for spins

cc/ @shanika.galaudage and @maya.fishbach as reviewer