Invariant Causal Prediction for python
Project description
ICPy
This packages provides a simple python implementation of Invariant Causal Prediction (ICP) [1].
The source code for the actual algorithm resides in ./src/icp/ICP.py.
See also the original implementation in the R package InvariantCausalPrediction.
Installation
pip install ICPy
Usage
import icpy as icpy
import numpy as np
np.random.seed(seed=1)
n = 100
noise = 0.1
E = np.repeat([0, 1, 2], np.ceil(n / 3.0))[0:n] # "Environment"
A = np.random.normal(scale=noise, size=[n]) + np.equal(E, 1) # Node A
B = A + np.random.normal(scale=noise, size=[n]) / 3 + np.equal(E, 2) # Node B
C = B + np.random.normal(scale=noise, size=[n]) # Node C
# /--->---\
# E -> A -> B -> C
icpy.invariant_causal_prediction(X=np.column_stack((A, B)), y=C, z=E) # test if A or B are parents of C
Output
ICP(S_hat=array([1], dtype=int64), # Column 1 = Node B was (correctly) identified as parent of C
p_values=array([ 1.51508232e-01, 4.59577055e-37]), # error levels at which A and B would/are indentied as parent of C
p_value=0.16416488336322549) # p-value for testing against violation of the model assumptions (e.g. a direct effect of E on C)
News
v0.0.003 (2020-05-15)
- fix failing import (thanks to @lgmoneda, #1)
- fix issues when environments are not subsequent whole numbers starting at 0 (thanks to @lgmoneda, #1)
References
[1] J. Peters, P. Bühlmann, N. Meinshausen, Causal inference by using invariant prediction: identification and confidence intervals, J. R. Stat. Soc. Ser. B Stat. Methodol. 78 (2016) 947-1012. doi:10.1111/rssb.12167.
Project details
Download files
Download the file for your platform. If you're not sure which to choose, learn more about installing packages.
Source Distribution
Built Distribution
File details
Details for the file ICPy-0.0.6.tar.gz
.
File metadata
- Download URL: ICPy-0.0.6.tar.gz
- Upload date:
- Size: 5.6 kB
- Tags: Source
- Uploaded using Trusted Publishing? No
- Uploaded via: twine/5.0.0 CPython/3.9.19
File hashes
Algorithm | Hash digest | |
---|---|---|
SHA256 |
2f65df7f9d9be6f6bd52a77ffc8f0858aaa2c6dd4fc7958a278fb3f15cc2b538
|
|
MD5 |
09cff7133f00d06cfb39eb27bba88b12
|
|
BLAKE2b-256 |
c225bf3d0b8f852f71d2808c180dbb126c16e2abd00bd5bf8a9cb0fc505e046a
|
File details
Details for the file ICPy-0.0.6-py3-none-any.whl
.
File metadata
- Download URL: ICPy-0.0.6-py3-none-any.whl
- Upload date:
- Size: 4.6 kB
- Tags: Python 3
- Uploaded using Trusted Publishing? No
- Uploaded via: twine/5.0.0 CPython/3.9.19
File hashes
Algorithm | Hash digest | |
---|---|---|
SHA256 |
e92a09337721b06543af3c3d95ce119c3276aa004c156e19c1cf037bc7c5f83c
|
|
MD5 |
79fe46ab0db066bda89782ba9d6ebed2
|
|
BLAKE2b-256 |
f95de7ca0300393659771690a65437e8ad2321cc6c2da42fa5c87130d164529d
|