Commit b1150690 authored by Colm Talbot's avatar Colm Talbot Committed by Paul Lasky

Add some post processing tools

parent 92239c57
......@@ -11,6 +11,8 @@ Changes currently on master, but not under a tag.
- Renamed `BBHPriorSet` to `BBHPriorDict`
- Renamed `BNSPriorSet` to `BNSPriorDict`
- Renamed `CalibrationPriorSet` to `CalibrationPriorDict`
- Added method to result to get injection recovery credible levels
- Added function to generate a pp-plot from many results to core/
- Fixed a bug which caused `Interferometer.detector_tensor` not to update when `latitude`, `longitude`, `xarm_azimuth`, `yarm_azimuth`, `xarm_tilt`, `yarm_tilt` were updated.
### Changes
from __future__ import division
import os
from distutils.version import LooseVersion
import numpy as np
......@@ -9,8 +11,9 @@ import matplotlib.pyplot as plt
from collections import OrderedDict, namedtuple
from . import utils
from .utils import logger, infer_parameters_from_function
from .prior import PriorDict, DeltaFunction
from .utils import (logger, infer_parameters_from_function,
from .prior import Prior, PriorDict, DeltaFunction
def result_file_name(outdir, label):
......@@ -136,6 +139,7 @@ class Result(object):
self.log_bayes_factor = log_bayes_factor
self.log_likelihood_evaluations = log_likelihood_evaluations
self.sampling_time = sampling_time
self.max_autocorrelation_time = max_autocorrelation_time
def __str__(self):
"""Print a summary """
......@@ -327,7 +331,7 @@ class Result(object):
elif k in self.parameter_labels:
'key {} not a parameter label or latex label'.format(k))
latex_labels.append(' '.join(k.split('_')))
return latex_labels
......@@ -399,6 +403,80 @@ class Result(object):
fmt(summary.median), fmt(summary.minus), fmt(
return summary
def plot_single_density(self, key, prior=None, cumulative=False,
title=None, truth=None, save=True,
file_base_name=None, bins=50, label_fontsize=16,
title_fontsize=16, quantiles=[0.16, 0.84], dpi=300):
""" Plot a 1D marginal density, either probablility or cumulative.
key: str
Name of the parameter to plot
prior: {bool (True), bilby.core.prior.Prior}
If true, add the stored prior probability density function to the
one-dimensional marginal distributions. If instead a Prior
is provided, this will be plotted.
cumulative: bool
If true plot the CDF
title: bool
If true, add 1D title of the median and (by default 1-sigma)
error bars. To change the error bars, pass in the quantiles kwarg.
See method `get_one_dimensional_median_and_error_bar` for further
details). If `quantiles=None` is passed in, no title is added.
truth: {bool, float}
If true, plot self.injection_parameters[parameter].
If float, plot this value.
save: bool:
If true, save plot to disk.
file_base_name: str, optional
If given, the base file name to use (by default `outdir/label_` is
bins: int
The number of histogram bins
label_fontsize, title_fontsize: int
The fontsizes for the labels and titles
quantiles: list
A length-2 list of the lower and upper-quantiles to calculate
the errors bars for.
dpi: int
Dots per inch resolution of the plot
figure: matplotlib.pyplot.figure
A matplotlib figure object
"""'Plotting {} marginal distribution'.format(key))
label = self.get_latex_labels_from_parameter_keys([key])[0]
fig, ax = plt.subplots()
ax.hist(self.posterior[key].values, bins=bins, density=True,
histtype='step', cumulative=cumulative)
ax.set_xlabel(label, fontsize=label_fontsize)
if truth is not None:
ax.axvline(truth, ls='-', color='orange')
summary = self.get_one_dimensional_median_and_error_bar(
key, quantiles=quantiles)
ax.axvline(summary.median - summary.minus, ls='--', color='C0')
ax.axvline(summary.median +, ls='--', color='C0')
if title:
ax.set_title(summary.string, fontsize=title_fontsize)
if isinstance(prior, Prior):
theta = np.linspace(ax.get_xlim()[0], ax.get_xlim()[1], 300)
ax.plot(theta, Prior.prob(theta), color='C2')
if save:
if cumulative:
file_name = file_base_name + key + '_cdf'
file_name = file_base_name + key + '_pdf'
fig.savefig(file_name, dpi=dpi)
return fig
def plot_marginals(self, parameters=None, priors=None, titles=True,
file_base_name=None, bins=50, label_fontsize=16,
title_fontsize=16, quantiles=[0.16, 0.84], dpi=300):
......@@ -433,57 +511,39 @@ class Result(object):
figures: dictionary
A dictionary of the matplotlib figures
if isinstance(parameters, dict):
plot_parameter_keys = list(parameters.keys())
truths = list(parameters.values())
elif parameters is None:
plot_parameter_keys = self.search_parameter_keys
truths = None
plot_parameter_keys = self.injection_parameters.keys()
truths = [self.injection_parameters.get(key, None) for key
in plot_parameter_keys]
plot_parameter_keys = list(parameters)
truths = None
truths = [self.injection_parameters.get(key, None) for key
in plot_parameter_keys]
labels = self.get_latex_labels_from_parameter_keys(plot_parameter_keys)
if file_base_name is None:
file_base_name = '{}/{}_'.format(self.outdir, self.label)
file_base_name = '{}/{}_1d/'.format(self.outdir, self.label)
if priors is True:
priors = getattr(self, 'priors', False)
elif isinstance(priors, (dict)) or priors in [False, None]:
elif isinstance(priors, dict) or priors in [False, None]:
raise ValueError('Input priors={} not understood'.format(priors))
figures = dict()
for i, key in enumerate(plot_parameter_keys):
fig, ax = plt.subplots()
ax.hist(self.posterior[key].values, bins=bins, density=True,
ax.set_xlabel(labels[i], fontsize=label_fontsize)
if truths is not None:
ax.axvline(truths[i], ls='--', color='orange')
summary = self.get_one_dimensional_median_and_error_bar(
key, quantiles=quantiles)
ax.axvline(summary.median - summary.minus, ls='--', color='C0')
ax.axvline(summary.median +, ls='--', color='C0')
if titles:
ax.set_title(summary.string, fontsize=title_fontsize)
if isinstance(priors, dict):
theta = np.linspace(ax.get_xlim()[0], ax.get_xlim()[1], 300)
ax.plot(theta, priors[key].prob(theta), color='C2')
fig.savefig(file_base_name + key)
figures[key] = fig
return figures
if not isinstance(self.posterior[key].values[0], float):
for cumulative in [False, True]:
key, prior=priors[i], cumulative=cumulative, title=titles,
truth=truths[i], save=True, file_base_name=file_base_name,
bins=bins, label_fontsize=label_fontsize, dpi=dpi,
title_fontsize=title_fontsize, quantiles=quantiles)
def plot_corner(self, parameters=None, priors=None, titles=True, save=True,
filename=None, dpi=300, **kwargs):
......@@ -702,7 +762,7 @@ class Result(object):
if all(~np.isnan(self.posterior.log_likelihood)):'Plotting maximum likelihood')
s = model_posterior.ix[self.posterior.log_likelihood.idxmax()]
s = model_posterior.iloc[self.posterior.log_likelihood.idxmax()]
ax.plot(xsmooth, model(xsmooth, **s), lw=1, color='k',
except AttributeError:
......@@ -777,6 +837,49 @@ class Result(object):
= priors[key].prob(self.posterior[key].values)
def get_all_injection_credible_levels(self):
Get credible levels for all parameters in self.injection_parameters
credible_levels: dict
The credible levels at which the injected parameters are found.
if self.injection_parameters is None:
raise(TypeError, "Result object has no 'injection_parameters'. "
"Cannot copmute credible levels.")
credible_levels = {key: self.get_injection_credible_level(key)
for key in self.search_parameter_keys
if isinstance(self.injection_parameters[key], float)}
return credible_levels
def get_injection_credible_level(self, parameter):
Get the credible level of the injected parameter
Calculated as CDF(injection value)
parameter: str
Parameter to get credible level for
float: credible level
if self.injection_parameters is None:
raise(TypeError, "Result object has no 'injection_parameters'. "
"Cannot copmute credible levels.")
if parameter in self.posterior and\
parameter in self.injection_parameters:
credible_level =\
sum(self.posterior[parameter].values <
self.injection_parameters[parameter]) / len(self.posterior)
return credible_level
return np.nan
def _check_attribute_match_to_other_object(self, name, other_object):
""" Check attribute name exists in other_object and is the same
......@@ -890,3 +993,45 @@ def plot_multiple(results, filename=None, labels=None, colours=None,
if save:
return fig
def make_pp_plot(results, filename=None, save=True, **kwargs):
Make a P-P plot for a set of runs with injected signals.
results: list
A list of Result objects, each of these should have injected_parameters
filename: str, optional
The name of the file to save, the default is "outdir/pp.png"
save: bool, optional
Whether to save the file, default=True
Additional kwargs to pass to matplotlib.pyplot.plot
Matplotlib figure
fig = plt.figure()
credible_levels = pd.DataFrame()
for result in results:
credible_levels = credible_levels.append(
result.get_all_injection_credible_levels(), ignore_index=True)
n_parameters = len(credible_levels.keys())
x_values = np.linspace(0, 1, 101)
for key in credible_levels:
plt.plot(x_values, [sum(credible_levels[key].values < xx) /
len(credible_levels) for xx in x_values],
color='k', alpha=min([1, 4 / n_parameters]), **kwargs)
plt.plot([0, 1], [0, 1], linestyle='--', color='r')
plt.xlim(0, 1)
plt.ylim(0, 1)
if save:
if filename is None:
filename = 'outdir/pp.png'
return fig
......@@ -8,6 +8,7 @@ import matplotlib.pyplot as plt
from bilby.core.likelihood import GaussianLikelihood
from bilby.core.prior import Uniform
from bilby.core.sampler import run_sampler
from bilby.core.result import make_pp_plot
from bilby.hyper.likelihood import HyperparameterLikelihood
outdir = 'outdir'
......@@ -31,8 +32,7 @@ fig1, ax1 = plt.subplots()
fig2, ax2 = plt.subplots()
# Make the sample sets
samples = list()
evidences = list()
results = list()
for i in range(Nevents):
c0 = np.random.normal(true_mu_c0, true_sigma_c0)
c1 = np.random.uniform(-1, 1)
......@@ -47,11 +47,10 @@ for i in range(Nevents):
result = run_sampler(
likelihood=likelihood, priors=priors, sampler='nestle', nlive=1000,
outdir=outdir, verbose=False, label='individual_{}'.format(i),
save=False, injection_parameters=injection_parameters)
ax2.hist(result.posterior.c0, color=line[0].get_color(), normed=True,
alpha=0.5, label=labels[i])
......@@ -72,6 +71,8 @@ def run_prior(data):
return 1 / 20
samples = [result.posterior for result in results]
evidences = [result.log_evidence for result in results]
hp_likelihood = HyperparameterLikelihood(
posteriors=samples, hyper_prior=hyper_prior,
sampling_prior=run_prior, log_evidences=evidences, max_samples=500)
......@@ -85,3 +86,4 @@ result = run_sampler(
use_ratio=False, outdir=outdir, label='hyper_parameter',
verbose=True, clean=True)
result.plot_corner(truth=dict(mu=true_mu_c0, sigma=true_sigma_c0))
make_pp_plot(results, filename=outdir + '/hyper_parameter_pp.png')
......@@ -11,8 +11,9 @@ import os
class TestResult(unittest.TestCase):
def setUp(self):
bilby.utils.command_line_args.test = False
priors = bilby.prior.PriorSet(dict(
priors = bilby.prior.PriorDict(dict(
x=bilby.prior.Uniform(0, 1, 'x', latex_label='$x$', unit='s'),
y=bilby.prior.Uniform(0, 1, 'y', latex_label='$y$', unit='m'),
......@@ -230,5 +231,15 @@ class TestResult(unittest.TestCase):
with self.assertRaises(ValueError):
def test_get_credible_levels(self):
levels = self.result.get_all_injection_credible_levels()
self.assertDictEqual(levels, dict(x=0.68, y=0.72))
def test_get_credible_levels_raises_error_if_no_injection_parameters(self):
self.result.injection_parameters = None
with self.assertRaises(TypeError):
if __name__ == '__main__':
Markdown is supported
0% or .
You are about to add 0 people to the discussion. Proceed with caution.
Finish editing this message first!
Please register or to comment