hplusminus

Copyright (C) 2020 Juergen Koefinger, Max Planck Institute of Biophysics, Frankfurt am Main, Germany

With contributions from Klaus Reuter, Max Planck Computing and Data Facility, Garching, Germany.

Released under the MIT Licence, see the file LICENSE.txt.

Reference

Powerful statistical tests for ordered data
Juergen Koefinger and Gerhard Hummer
Preprint: https://doi.org/10.26434/chemrxiv.13373351 (2020)

Requirements

• Python 3
• Jupyter (for Jupyter notebooks only)
• Python modules (can be installed with pip or conda):
  • numpy
  • scipy
  • mpmath
  • matplotlib (for Jupyter notebooks only)

Installation

The hplusminus package can be installed in the following ways:

pip installation

To install via pip please run:

pip install --user hplusminus

Installation from source

After downloading and unpacking the source tarball please run:

python setup.py install --user

Python script

hplusminus_tests.py

Python 3 script to evaluate test statistics for normalized residuals. This script contains the central functionality.

See help for more information and usage:

python hplusminus_tests.py -h

Example for alternative model

python hplusminus_tests.py ./examples/alternative_model_normalized_residuals.txt

Example for true model

python hplusminus_tests.py ./examples/true_model_normalized_residuals.txt

Jupyter notebooks

Notebooks in the directory ./ipynb/ serve to explore the capabilties of our statistical tests. See notebooks themselves for more details on purpose and usage.

hplusminus_tests.ipynb

Evaluate statistical tests.

Addtionally to the functionality of hplusminus_tests.py, the notebook provides plots of normalized residuals and signs and a bar-plot to visually compare the p-values of the various tests.

hplusminus_statistical_power.ipynb

Calculate statistical power for all tests and given model

generate_models_for_residuals.ipynb

Generate models for residuals, which can be used with the Python script hplusminus_tests.py and the Jupyter notebooks hplusminus_tests.ipyn and hplusminus_statistical_power.ipynb.

Python package hplusminus

tests.py

Python 3 module file containing functions for the convenient evaluation of the statistical tests.

io.py

Python 3 module file for input and output.

rld.py

Python 3 module file for the calculation of the Shannon information (neg. log-probabilities) of all test statistics (rld for Run-Length Distribution). Required by Python script hplusminus_tests.py and Jupyter notebooks hplusminus_tests.ipynb and hplusminus_statistical_power.ipynb.

sid.py

Python 3 module file for the calculation of p-values using the gamma distribution approximation of the cumulative Shannon information distributions (SID). Required by Python script hplusminus_tests.py and Jupyter notebooks hplusminus_tests.ipynb and hplusminus_statistical_power.ipynb.

Directories

./hplusminus/

Python module hplusminus

./examples/

Examples for normalized residuals generated with generate_models_for_residuals.ipynb.

./hplusminus/gsp/

Numpy binary files containing B-spline parameters (knots and coefficients) for gamma distribution parameters for all tests. Information is read from these files. No need for user interaction.

./ipynb

Directory containing Jupyter notebooks.

./ipynb/data/

Directory used by Jupyter notebooks for input/output.