IGCI: Information-geometric approach to inferring causal...

In DMIRLAB-Group/CANM: Causal Discovery with Cascade Nonlinear Additive Noise Models

Description

This is the implementation of the IGCI method for causal discovery.

Usage

IGCI(x, y, refMeasure, estimator)

Arguments

x	The observation of the cause.
y	The observation of the effect.
refMeasure	reference measure to use: 1: uniform 2: Gaussian
estimator	estimator to use:1: entropy (eq. (12) in [1]), 2: integral approximation (eq. (13) in [1]). 3: new integral approximation (eq. (22) in [2]) that should deal better with repeated values

Value

f < 0: the method prefers the causal direction x -> y
f > 0: the method prefers the causal direction y -> x

References

Janzing, Dominik, et al. "Information-geometric approach to inferring causal directions." Artificial Intelligence 182 (2012): 1-31.

Examples

set.seed(0)
x=rnorm(100)
y=exp(x)
f=IGCI(x,y,refMeasure=2,estimator=1)
if(f<0){
 print("X->Y")
}else{
 print("Y->X")
}

DMIRLAB-Group/CANM documentation built on May 31, 2019, 12:40 a.m.