ABrox is a python package for Approximate Bayesian Computation accompanied by a user-friendly graphical interface.

• Model comparison via approximate Bayes factors
  • rejection
  • random forest
• Parameter inference
  • rejection
  • MCMC
• Cross-validation

Note that ABroxonly works with Python 3.

ABrox can be installed via pip. Simply open a terminal and type:

pip install abrox

It might take a few seconds since there are several dependencies that you might have to install as well.

If you installed Python via MacPorts, the abrox-gui command after installation of abrox does not work. You can alternatively start the GUI via (assuming Python version 3.5):

cd /Library/Frameworks/Python.framework/Versions/3.5/lib/python3.5/site-packages/abrox/gui/
python3.5 main.py

Unfortunately, the installation under Windows is a bit cumbersome. We explain the relevant steps below.

If not already done, install a Python3 version from here.

Check the version of Python that is installed by typing python into the console.

Now, install Visual Studio Build Tools from:

1. here

Now visit the following page to install the Scipy wheel. Choose the link that fits your Python version (see picture above). cp should be followed by the actual version (e.g. cp36) while the last part of the link should match the bit-version (e.g. win32).

2. here

After the installation, open a console in the directory the wheel has been downloaded into and type:

python -m pip install #name_of_the_whl_file

Repeat the same steps for the Numpy wheel:

3. here

Now, open a terminal and type:

python -m pip install abrox

You are now ready to use ABrox!

After ABrox has been installed, you can start the user interface by typing abrox-gui. We provide several templates in order to get more familiar with the GUI.

If you are more comfortable with plain Python, you can run your project once from the GUI and continue working with the Python-file that has been generated in the output folder.

We provide a few example project files so you can see how ABrox works (here). Currently, we provide:

• Two-sample t-test
• Levene-Test

• Ulf Mertens
• Stefan Radev