Post-nonlinear causal models

Algorithm Introduction

Causal discovery based on the post-nonlinear (PNL 1) causal models. If you would like to apply the method to more than two variables, we suggest you first apply the PC algorithm and then use pair-wise analysis in this implementation to find the causal directions that cannot be determined by PC.

(Note: there are some potential issues in the current implementation of PNL. We are working on them and will update as soon as possible.)

Usage

from causallearn.search.FCMBased.PNL.PNL import PNL
pnl = PNL()
p_value_foward, p_value_backward = pnl.cause_or_effect(data_x, data_y)

Parameters

data_x: input data (n, 1), n is the sample size.

data_y: output data (n, 1), n is the sample size.

Returns

pval_forward: p value in the x->y direction.

pval_backward: p value in the y->x direction.

1

Zhang, K., & Hyvärinen, A. (2009, June). On the Identifiability of the Post-Nonlinear Causal Model. In 25th Conference on Uncertainty in Artificial Intelligence (UAI 2009) (pp. 647-655). AUAI Press.