Kernel‐based tests for joint independence

Python code of independence test algorithm proposed in

Kernel‐based tests for joint independence
Niklas Pfister, Peter Bühlmann, Bernhard Schölkopf, Jonas Peters
Journal of the Royal Statistical Society: Series B (Statistical Methodology) 80.1 (2018): 5-31.

Prerequisites

numpy (=1.13.3)
scipy (=1.1.0)
statsmodels (=0.10.1)

We test the code using python 3.6.8 on Windows 10. Any later version should still work perfectly.

Running the test

After installing all required packages, you can run demo.py to see whether dhsic could work normally.

The test code does the following:

generates 100 instances (a (100, 3) numpy array) from 3 jointly independent variables;
applies dhsic joint independence test on the data
changes variable 3 to be dependent on variable 1 and 2
applies dhsic joint independence test on the updated data.

Apply `dhsic` on your data

Usage

from dhsic import dhsic_test, dhsic

res = dhsic_test(X) # conduct joint independence test
res = dhsic(X)      # compute the dHSIC value

Description

Function dhsic_test()

Argument	Description
X	matrix of all instances, `(n_samples, n_vars)` numpy array
alpha (optional)	level of the test
method (optional)	method of the test

Output	Description
res	dictionary containing test statistic, critical value, p-value and the name of the test method

Function dhsic()

Argument	Description
X	matrix of all instances, `(n_samples, n_vars)` numpy array

Output	Description
dHSIC	dHSIC value

Author

Shoubo Hu - shoubo [dot] sub [at] gmail [dot] com

See also the list of contributors who participated in this project.

License

This project is licensed under the MIT License - see the LICENSE file for details.

amber0309/Kernel-based-tests-for-joint-independence