def make_posterior_workflow(workflow,
samples_files,
config_file,
label,
rdir,
posterior_file_dir='posterior_files',
tags=None):
"""Adds jobs to a workflow that make a posterior file and subsequent plots.
A posterior file is first created from the given samples file(s). The
settings for extracting the posterior are set by the
``[extract_posterior]`` section. If that section has a ``parameters``
argument, then the parameters in the posterior file (and for use in all
subsequent plotting) will be whatever that option is set to. Otherwise,
the parameters in the posterior file will be whatever is common to
all of the given samples file(s).
Except for prior plots (which use the given inference config file), all
subsequent jobs use the posterior file. The following are created:
* **Summary table**: an html table created using the ``table_summary``
executable. The parameters to print in the table are retrieved from the
``table-params`` option in the ``[workflow-summary_table]`` section.
Metadata may also be printed by adding a ``print-metadata`` option to
that section.
* **Summary posterior plots**: a collection of posterior plots to include
in the summary page, after the summary table. The parameters to plot
are read from ``[workflow-summary_plots]``. Parameters should be grouped
together by providing
``plot-group-NAME = PARAM1[:LABEL1] PARAM2[:LABEL2]`` in that section,
where ``NAME`` is a unique name for each group. One posterior plot will
be created for each plot group. For clarity, only one or two parameters
should be plotted in each summary group, but this is not enforced.
Settings for the plotting executable are read from the
``plot_posterior_summary`` section; likewise, the executable used
is read from ``plot_posterior_summary`` in the
``[executables]`` section.
* **Sky maps**: if *both* ``create_fits_file`` and ``plot_skymap``
are listed in the ``[executables]`` section, then a ``.fits`` file and
sky map plot will be produced. The sky map plot will be included in
the summary plots. You must be running in a python 3 environment to
create these.
* **Prior plots**: plots of the prior will be created using the
``plot_prior`` executable. By default, all of the variable
parameters will be plotted. The prior plots are added to
``priors/LALBEL/`` in the results directory, where ``LABEL`` is the
given ``label``.
* **Posterior plots**: additional posterior plots are created using the
``plot_posterior`` executable. The parameters to plot are
read from ``[workflow-plot_params]`` section. As with the summary
posterior plots, parameters are grouped together by providing
``plot-group-NAME`` options in that section. A posterior plot will be
created for each group, and added to the ``posteriors/LABEL/`` directory.
Plot settings are read from the ``[plot_posterior]`` section; this
is kept separate from the posterior summary so that different settings
can be used. For example, you may want to make a density plot for the
summary plots, but a scatter plot colored by SNR for the posterior plots.
Parameters
----------
samples_file : pycbc.workflow.core.FileList
List of samples files to combine into a single posterior file.
config_file : pycbc.worfkow.File
The inference configuration file used to generate the samples file(s).
This is needed to make plots of the prior.
label : str
Unique label for the plots. Used in file names.
rdir : pycbc.results.layout.SectionNumber
The results directory to save the plots to.
posterior_file_dir : str, optional
The name of the directory to save the posterior file to. Default is
``posterior_files``.
tags : list of str, optional
Additional tags to add to the file names.
Returns
-------
posterior_file : pycbc.workflow.File
The posterior file that was created.
summary_files : list
List of files to go on the summary results page.
prior_plots : list
List of prior plots that will be created. These will be saved to
``priors/LABEL/`` in the resuls directory, where ``LABEL`` is the
provided label.
posterior_plots : list
List of posterior plots that will be created. These will be saved to
``posteriors/LABEL/`` in the results directory.
"""
# the list of plots to go in the summary
summary_files = []
if tags is None:
tags = []
analysis_seg = workflow.analysis_time
# figure out what parameters user wants to plot from workflow configuration
# parameters for the summary plots
summary_plot_params = get_plot_group(workflow.cp, 'summary_plots')
# parameters to plot in large corner plots
plot_params = get_plot_group(workflow.cp, 'plot_params')
# get parameters for the summary tables
table_params = workflow.cp.get_opt_tag('workflow', 'table-params',
'summary_table')
# get any metadata that should be printed
if workflow.cp.has_option('workflow-summary_table', 'print-metadata'):
table_metadata = workflow.cp.get_opt_tag('workflow', 'print-metadata',
'summary_table')
else:
table_metadata = None
# figure out if we are making a skymap
make_skymap = ("create_fits_file" in workflow.cp.options("executables")
and "plot_skymap" in workflow.cp.options("executables"))
# make node for running extract samples
posterior_file = create_posterior_files(workflow,
samples_files,
posterior_file_dir,
analysis_seg=analysis_seg,
tags=tags + [label])[0]
# summary table
summary_files += (make_inference_summary_table(
workflow,
posterior_file,
rdir.base,
parameters=table_params,
print_metadata=table_metadata,
analysis_seg=analysis_seg,
tags=tags + [label]), )
# summary posteriors
summary_plots = []
for group, params in summary_plot_params.items():
summary_plots += make_inference_posterior_plot(
workflow,
posterior_file,
rdir.base,
name='plot_posterior_summary',
parameters=params,
plot_prior_from_file=config_file,
analysis_seg=analysis_seg,
tags=tags + [label, group])
# sky map
if make_skymap:
# create the fits file
fits_file = create_fits_file(workflow,
posterior_file,
rdir.base,
analysis_seg=analysis_seg,
tags=tags + [label])[0]
# now plot the skymap
skymap_plot = make_inference_skymap(workflow,
fits_file,
rdir.base,
analysis_seg=analysis_seg,
tags=tags + [label])
summary_plots += skymap_plot
summary_files += list(layout.grouper(summary_plots, 2))
# files for priors summary section
base = "priors/{}".format(label)
prior_plots = make_inference_prior_plot(
workflow,
config_file,
rdir[base],
analysis_seg=workflow.analysis_time,
tags=tags + [label])
layout.single_layout(rdir[base], prior_plots)
# files for posteriors summary subsection
base = "posteriors/{}".format(label)
posterior_plots = []
for group, params in plot_params.items():
posterior_plots += make_inference_posterior_plot(
workflow,
posterior_file,
rdir[base],
parameters=params,
plot_prior_from_file=config_file,
analysis_seg=analysis_seg,
tags=tags + [label, group])
layout.single_layout(rdir[base], posterior_plots)
return posterior_file, summary_files, prior_plots, posterior_plots
def make_diagnostic_plots(workflow,
diagnostics,
samples_file,
label,
rdir,
tags=None):
"""Makes diagnostic plots.
Diagnostic plots are sampler-specific plots the provide information on
how the sampler performed. All diagnostic plots use the output file
produced by ``pycbc_inference`` as their input. Diagnostic plots are added
to the results directory ``rdir/NAME`` where ``NAME`` is the name of the
diagnostic given in ``diagnostics``.
Parameters
----------
workflow : pycbc.workflow.core.Workflow
The workflow to add the plotting jobs to.
diagnostics : list of str
The names of the diagnostic plots to create. See
:py:func:`get_diagnostic_plots` for recognized names.
samples_file : (list of) pycbc.workflow.File
One or more samples files with which to create the diagnostic plots.
If a list of files is provided, a diagnostic plot for each file will
be created.
label : str
Event label for the diagnostic plots.
rdir : pycbc.results.layout.SectionNumber
Results directory layout.
tags : list of str, optional
Additional tags to add to the file names.
Returns
-------
dict :
Dictionary of diagnostic name -> list of files giving the plots that
will be created.
"""
if tags is None:
tags = []
out = {}
if not isinstance(samples_file, list):
samples_file = [samples_file]
if 'samples' in diagnostics:
# files for samples summary subsection
base = "samples/{}".format(label)
samples_plots = []
for kk, sf in enumerate(samples_file):
samples_plots += make_inference_samples_plot(
workflow,
sf,
rdir[base],
analysis_seg=workflow.analysis_time,
tags=tags + [label, str(kk)])
out['samples'] = samples_plots
layout.group_layout(rdir[base], samples_plots)
if 'acceptance_rate' in diagnostics:
# files for samples acceptance_rate subsection
base = "acceptance_rate/{}".format(label)
acceptance_plots = []
for kk, sf in enumerate(samples_file):
acceptance_plots += make_inference_acceptance_rate_plot(
workflow,
sf,
rdir[base],
analysis_seg=workflow.analysis_time,
tags=tags + [label, str(kk)])
out['acceptance_rate'] = acceptance_plots
layout.single_layout(rdir[base], acceptance_plots)
if 'mcmc_history' in diagnostics:
# files for samples mcmc history subsection
base = "mcmc_history/{}".format(label)
history_plots = []
for kk, sf in enumerate(samples_file):
history_plots += make_inference_plot_mcmc_history(
workflow,
sf,
rdir[base],
analysis_seg=workflow.analysis_time,
tags=tags + [label, str(kk)])
out['mcmc_history'] = history_plots
layout.single_layout(rdir[base], history_plots)
if 'dynesty_run' in diagnostics:
# files for dynesty run subsection
base = "dynesty_run/{}".format(label)
dynesty_run_plots = []
for kk, sf in enumerate(samples_file):
dynesty_run_plots += make_inference_dynesty_run_plot(
workflow,
sf,
rdir[base],
analysis_seg=workflow.analysis_time,
tags=tags + [label, str(kk)])
out['dynesty_run'] = dynesty_run_plots
layout.single_layout(rdir[base], dynesty_run_plots)
if 'dynesty_traceplot' in diagnostics:
# files for samples dynesty tyrace plots subsection
base = "dynesty_traceplot/{}".format(label)
dynesty_trace_plots = []
for kk, sf in enumerate(samples_file):
dynesty_trace_plots += make_inference_dynesty_trace_plot(
workflow,
sf,
rdir[base],
analysis_seg=workflow.analysis_time,
tags=tags + [label, str(kk)])
out['dynesty_traceplot'] = dynesty_trace_plots
layout.single_layout(rdir[base], dynesty_trace_plots)
return out