Welcome to BXA’s documentation!

About Bayesian X-ray Analysis (BXA)

BXA connects the X-ray spectral analysis environments Xspec/Sherpa to the nested sampling algorithm UltraNest for Bayesian Parameter Estimation and Model comparison.

BXA provides the following features:

  • parameter estimation in arbitrary dimensions, which involves:
    • finding the best fit

    • computing error bars

    • computing marginal probability distributions

    • parallelisation with MPI

  • plotting of spectral model vs. the data:
    • for the best fit

    • for each of the solutions (posterior samples)

    • for each component

  • model selection:
    • computing the evidence for the considered model, ready for use in Bayes factors

    • unlike likelihood-ratios, not limited to nested models

  • model discovery:
    • visualize deviations between model and data with Quantile-Quantile (QQ) plots. QQ-plots do not require binning and are more comprehensive than residuals. This will give you ideas on when to introduce more complex models, which may again be tested with model selection

BXA shines especially

  • when systematically analysing a large data-set, or

  • when comparing multiple models

  • when analysing low counts data-set with realistic models

because its robust and unsupervised fitting algorithm explores even complicated parameter spaces in an automated fashion. The user does not need to initialise to good starting points. The algorithm automatically runs until convergence, and slows down to sample carefully if complicated parameter spaces are encountered. This allows building automated analysis pipelines.

https://img.shields.io/pypi/v/BXA.svg https://coveralls.io/repos/github/JohannesBuchner/BXA/badge.svg Documentation Status Github repository

Who is using BXA?

  • Dr. Antonis Georgakakis, Dr. Angel Ruiz (NOA, Athens)

  • Dr. Mike Anderson (MPA, Munich)

  • Dr. Franz Bauer, Charlotte Simmonds (PUC, Jonathan Quirola Vásquez, Santiago)

  • Dr. Stéphane Paltani, Dr. Carlo Ferrigno (ISDC, Geneva)

  • Dr. Zhu Liu (NAO, Beijing)

  • Dr. Georgios Vasilopoulos (Yale, New Haven)

  • Dr. Francesca Civano, Dr. Aneta Siemiginowska (CfA/SAO, Cambridge)

  • Dr. Teng Liu, Adam Malyali, Riccardo Arcodia, Sophia Waddell, Torben Simm, … (MPE, Garching)

  • Dr. Sibasish Laha, Dr. Alex Markowitz (UCSD, San Diego)

  • Dr. Arash Bahramian (Curtin University, Perth)

  • Dr. Peter Boorman (U of Southampton, Southampton; ASU, Prague)

  • and you?


BXA’s documentation is hosted at http://johannesbuchner.github.io/BXA/


First, you need to have either Sherpa or Xspec installed and its environment loaded.

BXA itself can installed easily using pip or conda:

$ pip install bxa

If you want to install in your home directory, install with:

$ pip install bxa --user

The following commands should not yield any error message:

$ python -c 'import ultranest'
$ python -c 'import xspec'
$ sherpa

You may need to install python and some basic packages through your package manager. For example:

$ yum install ipython python-matplotlib scipy numpy matplotlib
$ apt-get install python-numpy python-scipy python-matplotlib ipython

BXA requires the following python packages: requests corner astropy h5py cython scipy tqdm. They should be downloaded automatically. If they are not, install them also with pip/conda.

The source code is available from https://github.com/JohannesBuchner/BXA, so alternatively you can download and install it:

$ git clone https://github.com/JohannesBuchner/BXA
$ cd BXA
$ python setup.py install

Or if you only want to install it for the current user:

$ python setup.py install --user

Supported operating systems: BXA runs on all operating systems supported by ciao/sherpa or heasoft/xspec. The support is systematically tested for every BXA release by Travis CI, but only for Ubuntu Linux.


In Sherpa, load the package:

jbuchner@ds42 ~ $ sherpa
Welcome to Sherpa: CXC's Modeling and Fitting Package
CIAO 4.4 Sherpa version 2 Tuesday, June 5, 2012

sherpa-1> import bxa.sherpa as bxa
sherpa-2> bxa.BXASolver?

For Xspec, start python or ipython:

jbuchner@ds42 ~ $ ipython
In [1]: import xspec

In [2]: import bxa.xspec as bxa

In [3]: bxa.BXASolver?

Now you can use BXA. See the documentation pages for how to perform analyses. Several examples are included.


See the code repository page

Citing BXA correctly

Refer to the accompaning paper Buchner et al. (2014) which gives introduction and detailed discussion on the methodology and its statistical footing.

We suggest giving credit to the developers of Sherpa/Xspec, UltraNest and of this software. As an example:

For analysing X-ray spectra, we use the analysis software BXA (\ref{Buchner2014}),
which connects the nested sampling algorithm UltraNest (\ref{ultranest})
with the fitting environment CIAO/Sherpa (\ref{Fruscione2006}).

Where the BibTex entries are:


  • We start with some notes on best practice in currently common Maximum Likelihood analysis (fitting) and Bayesian analysis.

The usage is similar in Sherpa and Xspec conceptionally:

Indices and tables