GWForge.population_fisher.compare
=================================

.. py:module:: GWForge.population_fisher.compare

.. autoapi-nested-parse::

   Fit a population block to the injections and report what came back.



Attributes
----------

.. autoapisummary::

   GWForge.population_fisher.compare.BLOCK_ROLE


Functions
---------

.. autoapisummary::

   GWForge.population_fisher.compare.compare_block


Module Contents
---------------

.. py:data:: BLOCK_ROLE

.. py:function:: compare_block(model, block, injections, free_parameters=None)

   Fit one block to the injections and format the comparison.

   :param model: Fitted **in place**: on return ``model.fiducial`` holds the fitted
                 values, so the Fisher that follows is evaluated at them.
   :type model: GWForge.population_fisher.base.PopulationModel
   :param block: ``"mass"``, ``"redshift"`` or ``"spin"``.
   :type block: str
   :param injections: Every injection, detected or not.
   :type injections: dict
   :param free_parameters: Vary only these, holding the rest at their config values. Defaults to
                           every parameter the model has.

                           The driver passes the *Fisher's* free list, so the fit varies exactly
                           what the forecast will vary. That matters when a parameter is pinned
                           because the data cannot constrain it: ``delta_m`` for a taper narrower
                           than the sampling grid has no maximum to find, so an optimiser given it
                           rails against whichever bound it was handed and drags the correlated
                           parameters with it. Since the catalogue was *generated* here, such
                           parameters are known exactly and there is nothing to fit.
   :type free_parameters: sequence of str or None

   :returns: Lines for the summary file.
   :rtype: list of str